FSP has unveiled its flagship PSU series, the 80 Plus Platinum certified Aurum 92+ HW, which will be available in capacities of 850 W (HW850), 1000 W (HW1000), and 1200 W (HW1200) and will offer a 92 percent efficiency level.

The Aurum 92+ HW features a single +12V rail design, compliance with ATX 12V v2.31 and EPS v2.92 that supports Intel’s 4th Generation processors’ low power states, LLC full-bridge topology design, and includes eight 6+2-pin PCIe connectors, 24-pin ATX and a “large number” of SATA / Molex ports.

FSP hasn’t released any information on the Aurum 92+ HW’s pricing or availability.

Manufacturers, Designers, And Labels

Between gobs of reader feedback and our own data compiled over many years, we've managed to put together a fairly comprehensive list of power supply brands and manufacturers. Despite the fact that it consists of more than 150 manufacturers, though, this list still doesn't reflect the entire market, which always seems to be in a state of flux. It can, however, be used as a guide to finding the difference between a bad deal and a bargain.

Who is Who?

Let’s start by dividing the manufacturers into three large groups so we can better understand the database and how these companies are connected:

1. The OEMs (Original Equipment Manufacturers)

OEMs manage all of their production internally. They either exclusively design and manufacture their own PSUs (like Enermax) or design and manufacture their own brands, as well as manufacture PSUs designed by other companies (such as FSP, HEC, and SeaSonic). Some of them focus heavily on worldwide exports and provide a range of models, which are later sold under different labels. It's common to find otherwise-identical models marketed under many different names and labels. The industrial areas around Shenzhen, China, are the cradle of the lowest-priced PSUs sold all over the globe.

2. Designers: Without Their Own Production

The second group of companies also develops and designs their own products. However, they have to outsource either some or all of the manufacturing to other companies. One example of this is Be Quiet. Those familiar with the brand noted how Be Quiet P7 models were suddenly much better than the disappointing P6. The answer was simply a manufacturer change, from Topower to FSP. Other examples of designers include SilverStone, Corsair, PC Power & Cooling, and Tagan.

3. The Labels: With or Without Any Technical Involvement

Arguably, this group could be subdivided. Some importers of foreign PSUs that resell models under their own labels have a certain influence over the quality and choice of components, while others simply bring in some very cheap products, relabel, and resell them.

This third group is the most interesting one for price-oriented customers, though also the most uncertain for quality. You're as likely to score a bargain by getting a relabeled high-quality product at a lower price as you are to be disappointed by being too tight-fisted. Some good examples of products to watch are new models from Aerocool, which are essentially the Cougar units from Compucase/HEC with a discounted price and completely restyled exterior.

After many tests and inspections of budget models (by us, our readers, and friendly computer stores), we would advise you to steer your piggy banks clear of the labels Rasurbo, Inter-Tech (Sinan Power, Coba), Tech Solo, LC Power, RaptoxX, Tronje, Xilence, Ultron, World Link, Q-Tec, etc. We were able to identify some of these models without looking at the UL number simply by having a look at the installed components. These were almost exclusively the simplest work of such manufacturers as Enhance, World Link, Andyson, Topower, Casing Macron, and Channel Well.

Lack of protection circuits, low efficiency, and bad build quality were major points of criticism. The lowest of the low was a European label called Hardwaremania24, targeted at OEM PCs. While still in standby mode, the PSU heated to about 176 degrees Fahrenheit, spent the next six hours billowing smoke, and finally made what might be described as a trumpeting sound before dying. The host computer was never even turned on. After analyzing the PSU, we found no protection at all save for a single slow fuse.

Firecracker Or Power Plant?

How do you identify a bad power supply before buying it?

• Extremely high wattage claims at a comparatively very low price are suspicious. There are simply no decent 750 W power supplies for $50. For every product class based on performance and features, there must be a minimum price. When a product is significantly below that price, be cautious. You can get a "400 W PSU" for $20, and such fire hazards are installed in budget PCs every day by unscrupulous companies that know exactly the risk they're handing off to buyers.
• Check the specifications. For example, if a PSU claims high performance on the 3.3 and 5 V rails while the 12 V rail numbers are low, then you know something is wrong.
• The manufacturer does not specify any combined maximum performance, but instead only shows the maximum load for each rail separately. This is done without specifying how much real power would be available if all rails are used at the same time. Avoid PSUs without this information.
• Be careful with juicy marketing expressions and commercial lingo: Super, Extreme, Gaming, Combat, etc. Using superlatives to describe something quite normal should arouse suspicion and have you double-check specification details.
• Passive rather than active Power Factor Correction (PFC) leads to lower power efficiency.
• Very few or short power connectors and cables might be an issue. A 750 W PSU usually has four PCIe connectors for graphics cards (2 x 6-pin and 2 x 6+2-pin), so think twice if a model only offers two (or at least consider your upgrade options).
• With cheap PSUs, the quality of the cable insulation may be poor, or the cables may not be insulated at all. The power cable grommet may also be insufficiently padded.
• Be careful if there are few or no indications of protection circuitry. If the PSU specification only says OPP (overload protection) or perhaps SCP (short circuit protection), this points towards a normal fuse. If the specification also says OVP (overvoltage protection), this probably means that it is equipped with a simple metal oxide variable resistor. These security measures by themselves are absolutely insufficient and cannot replace any kind of digital safety chip.

Unfortunately, you can't always tell at first glance whether you're dealing with a high-quality PSU or whether there's nothing but disappointment waiting behind the pleasant facade. Therefore, we decided to open up two budget PSUs representative of what you can find in many of today’s OEM PCs and illustrate the points and features you should be examining.

A First Look At the Inside: Primary Capacitor and PFC

First, look at the storage capacitors in the primary circuit. These act as buffers and help protect the PSU and computer from voltage fluctuations. The electrolyte used in them is key, because it evaporates or dries out through a combination of heat and time. As a general rule, capacitor lifetime is halved for each 10 degrees Centigrade increase in temperature over the specified normal load. Using higher-quality capacitors that can handle 105 degrees instead of 85 degrees (C) should almost double their lifetime, greatly contributing to the PSU's durability.

A PSU equipped with chokes (a kind of inductor) is a clear indicator of passive Power Factor Correction (PFC). Passive PFC plays a significant role in the efficiency of the PSU. Only active circuitry allows for factors close to the optimum value of 1, while passive components can reach 0.7 to 0.8 at best, meaning they only achieve 70% to 80% efficiency. PSUs with a passive PFC may be cheaper to buy in the short run, but poor efficiency can swallow savings over time in the form of higher electricity costs.

Protection Circuits

Even without opening the PSU, a data sheet can reveal some of the safety measures taken (or not) by manufacturers. A decent PSU should contain the following safety measures:

• OCP (Over Current Protection): protection against power spikes
• OVP (Over Voltage Protection)
• OPP (Over Power Protection): overload protection, sometimes called OLP 
• OTP (Over Temperature Protection): protection from overheating 
• UVP (Under Voltage Protection) 
• SCP (Short Circuit Protection) 
• NLO (No Load Operation): this isn’t exactly protection in the same sense as the other features, but it allows the PSU to power up and function normally, even with no load.

Without this information, you have to look inside the PSU to find out what you need to know.

We found no protection at all on this unit, except for a simple fuse. Sadly, this PSU is still available on the market under a couple of different labels.

Passive components do not guarantee sufficient protection. Without a digital security chip, the computer hardware is severely exposed to risks.

The security chip PS223 from Silicon Touch is popular, and you should avoid PSUs not using it or similar products, such as the PS332S.

Cables and Short Circuits

You can tell a lot about your PSU by looking at its internal wiring. A lack of heat shrink tubing, carelessly exposed solder joints, and components fastened with a glue gun are symptomatic of cheap and hazardous manufacturing. If unprotected cables are placed next to hot components, a PSU failure is nearly assured.

Boards

A final quality indicator is the circuit board material. Impregnated laminated paper (like the yellow boards in the pictures) is a sure sign of cost cutting. Fiber materials are much more durable and, perhaps more importantly, non-flammable.

How Do You Recognize The Manufacturer?

UL Numbers From Underwriters Laboratories

"Underwriters Laboratories, an independent firm working with product safety certification, has been active in the field of product testing and preparation of safety standards for more than a century. UL evaluates more than 19,000 types of products, components, materials and systems annually. Every year more than 20 billion UL marks are placed on products from 66,000 different manufacturers. The UL Group and its network of service providers include 68 testing and certification labs worldwide, serving customers in 102 countries."--from the About UL page at www.ul.com

Essentially, this means that all PSUs sold on the North American market must be marked with an UL number. This number should identify the actual manufacturer of a product. However, not every PSU has such a number. UL number omission on a North American product might indicate poor quality.

Step 1: Reading the UL Number

To find the UL number, there's no need to open your PSU or desolder anything. Simply open your PC and look at the power supply's label. The UL number usually begins with an "E" followed by a string of numbers. See the three examples below.

Online UL Number Query:

Now let's find out more about your PSU. Go to the UL Online Certifications Directory and enter the UL number in the UL File Number field. If the number exists in the database, you should immediately see the result. If not, the number is invalid, fake, or the manufacturer no longer exists.

Link: Online Certifications Directory

PSU Makers: 0-9 And A-B

Companies: 2the Max Asia Pacific, 3Y Power Technology, Ablecom, AcBel, Achme, AeroCool, Akasa, Amacrox, AMS, AOpen, Antec, Apex, Arctic Cooling, Aspire, Astec Int. Ltd., Asus, Athenatech, ATNG, Atrix, AXP, Be Quiet!, Bestec, BFG, Boser Technology Inc., Broadway

PSU Makers: C-D

Companies: Casing Macron, Channel Well, Chenbro, Chieftec, CompUSA, Coba, Codegen, Cooler Master, Coolmax, Cooltek, Corsair, Deer, Dell, Delta, Dongguan Zhangmotou, Dynapower

PSU Makers: E-H

Companies: Eagle Tech, Elwin Technology Ltd., Enermax, Enhance, Enlight, Etasis, Forepoint, Fortron, Foxconn, Foxlink, Fujitsu Siemens, Gigabyte, GPS, GTR, Hama, HEC, Hiper, Highscreen, High Performance Group, High Performance PC, High Power, Himere, Hipro, HP, Huntkey

PSU Makers: I-M

Companies: Impervio, Inter-Tech, In Win, I-Star, ICP, Jersey, JOU JYE, JPAC, Just PC, Key Mouse, Kingwin, L & C, LC-Power, Leadman, Levicom, Li Shin, Lian-Li, Lite-On, Linkworld, Logisys, Mad Dog, Mapower, Masscool, Mean Well, MGE/XG, Mitac, MSI, Mushkin

PSU Makers: N-R

Companies: Newton, Nexus, nMediaPC, NorthQ, NZXT, OCZ, Okia, PC Power & Cooling, Point of View, Powerman, Powertek, Powmax, Raidmax, RaptoxX, Rasurbo, Revoltec/Listan, Rosewill

PSU Makers: S-T

Companies: San Hawk, Scythe, Seasonic, Seventeam, Sharkoon, Shuttle, SilenX, Silverstone, Sinan Power, Sino Tech, Sirtec, Solytech, Soyo, Sparkle, Spire, Startech, Sunbeam, Sun Pro, Sunfone, Super Flower /SFC, Tagan, Taiwan Youngyear, Thermaltake, Topower, TTGI, Tyan

PSU Makers: U-Z

Companies: Ultra, Unitek, Vantec, Win-Tact, Wintech, Wiseframe, XClio, Xigmatek, Xilence, Xion, X-Spice, Yesico, Zalman, Zippy

Conclusion

We want to stress two things in particular:

A PSU is not necessarily bad because it is cheap. But the lower the price, the more you'll likely have to skimp on safety and performance features. A high-quality PSU contains more expensive components. If you buy cheap, you might have to buy twice--or more.

A PSU is not automatically inferior because it was made by a contractor with which you aren't familiar. These companies make all kinds of products, from luxury brands to pure garbage. Again, it's a question of hitting different production targets. Is the product built for quality or aggressive cost reduction? The responsibility always lies with the buyer to recognize and reject products that are obviously unsuitable.

We will keep collecting data and update this list every once in a while, and we’d love to include your findings. Please use our feedback form to forward information that isn’t available in our listing!

Less Can Be More

Editor's Note: If you want more background on how power supply testing plays out here at Tom's Hardware, I recommend that you check out Soon At Tom's Hardware: Full-Scale Power Supply Testing before reading this piece.

Special attention gets paid to the power supply unit (PSU) these days, primarily in the high-performance and gaming community. No wonder, as the requirements of top-shelf PCs with multiple processors and modern 3D graphics cards are constantly increasing. Such computers will suck 800 W or more from the power supply under load, and for short periods that number can be substantially higher.

On the other hand, power supplies are often forgotten in offices and for everyday multimedia applications. Truth be told, compact and mid-range PCs don't require a high-performance power supply. However, investing in a high-quality power supply can pay off in several ways.

First, a good PSU guarantees a reliable and consistent supply of power to all components. This can help extend component longevity. Second, a highly energy efficient PSU helps reduce the energy consumption of a home or office. Additionally, other aspects, such as ergonomics, cable modularity, and cable length can simplify or complicate PC assembly. PSUs with good build quality also prevent nicked fingers or a less-than-optimal fit during installation. Trust us. We've seen it all.

We gathered a few mid-range models for this roundup. A top street price of $150 was our only prerequisite, and we settled on four 500 W models from Cooler Master, Enermax, FSP, and SilverStone. The coveted 80 PLUS logo is emblazoned on each box, except for SilverStone's.

Cooler Master Silent Pro M500

Perhaps best known as a manufacturer of enthusiast-class computer cases, Cooler Master has played an important role on the PSU market for several years. The Silent Pro M500 we received for this roundup has been available in stores for several weeks, but that's no reason not to take a closer look at this power supply, especially considering its 80 PLUS certification.

As soon as you unpack the PSU, it becomes clear that the "Silent" added to the product name is not idle boasting. Silicon-covered mounting frames are included to stop vibrations from transferring to the computer case. This is a simple but effective tactic. In general, the housing of the Cooler Master Silent Pro M500 is very well made. No sharp edges or imperfect fit ruin the impression. A large 135 mm fan covers the entire base of the power supply.

The cables are flat in order to provide improved airflow inside the case, and while the lengths are sufficient, you might need some luck (Ed.: alternatively, you could do some research on component placement) to reach everything in large chassis. The ATX (20 + 4 pins) and CPU (4 + 4 pins) cables are firmly attached to the PSU, but all other cables are modular. You get plenty of cables as well: two PCIe, six SATA, and five four-pin Molex connectors.

Cooler Master Silent Pro M500: Measurements

Efficiency in accordance with the 80 PLUS specification:

Efficiency under different load profiles:

We had no issues while measuring the Cooler Master Silent Pro M500. The PSU meets 80 PLUS certification requirements with some margin. In fact, it comes quite close to qualifying for the Bronze logo. Efficiency only takes a small hit under full load at 115 volts; 81% is still okay. At the other end of the scale, at 25 W and 50 W, the Cooler Master PSU performs handsomely. At 50 W, it still achieves an efficiency of nearly 82% and falls only slightly under a 25 W load to 74 percent. The long hold-up times of 18 ms at 230 V and 19.2 ms at 115 V are also noteworthy.

The Cooler Master Silent Pro M500 is above average thanks to its efficiency at low loads, and its silicone-dampened housing make it especially suitable for use in compact computers in the office or living room. Additionally, the Cooler Master Silent Pro M500 is a reliable component, with good performance ratings for multimedia systems. Heavy users should consider that Cooler Master only provides one 12 V rail, though.

Enermax Pro87+ 500 W

As the only PSU in the roundup with an 80 PLUS Gold certification, the Enermax Pro87+ makes an immediate impression. One must search long and hard to find a PSU with such outstanding energy efficiency in the sub-$150 range, although we found that this product isn't currently available in all markets.

The Enermax Pro87+ is also a visual treat. Enermax chose to go with a gold-colored theme on both the packaging and the PSU itself. Even the fan blades have an aureate sheen.

Speaking of the fan, a nifty feature of the Enermax Pro87+ is that the fan continues to spin for 30 to 60 seconds even after the computer is shut down. This is supposed to increase the lifespan of sensitive components by evacuating residual heat from the computer case. We also like the Cord-Guard safety device that prevents you (or more likely your pet/kid/mother) from accidentally disconnecting the power cord. This is simple, but reliable.

The Enermax Pro87+ deliberately refrains from any sort of cable management or modularity. Thanks to the included Velcro cable ties, at least any unused cables can be tied up a bit so they don't impede air flow as much. Users who value the modular cable concept must resort to the Modu87+ series instead.

At least the cable length here is generous. The number of connectors is standard. In addition to the 24-pin ATX power supply and CPU connectors, you get two PCIe, six SATA, and five four-pin Molex connectors.

Enermax Pro87+ 500 W: Measurements

Efficiency in accordance with the 80 PLUS specification:

Efficiency under different load profiles:

It becomes clear during testing that Enermax has earned its 80 PLUS Gold certification. We saw impressive results in the 87% to 91% range in our energy efficiency measurements, both at 115 as well as 230 volts. This very solid showing was confirmed in other load measurements, with efficiency falling somewhat only at very low power loads. The overload test shows that Enermax's Pro87+ still has plenty of reserves. Even with an overload of more than 20% of the power supply's stated output, it still provides stable voltages. A small weakness was revealed when testing the hold-up time, though. Here, the power supply even falls short of the ATX specification requirements. This is an unnecessary drawback given the otherwise very high quality.

In the sub-$150 price class, the Enermax Pro87+ is a genuine highlight. The high energy efficiency under mid- and full-power loads qualifies the power supply for use in rather high-end PC systems. The Enermax Pro87+ is less suitable only for very low-power computers, as the energy efficiency diminishes under very light loads. However, for such PCs, this PSU would be overkill, anyway.

FSP Everest 85 PLUS 500

The FSP Everest 85Plus 500 stands out immediately, due to its blue color. No other special features are noticeable at first glance, though. Instead, the blue PSU convinces through its internal qualities. This is hinted at by the packaging's 80 PLUS Bronze logo.

FSP plays it safe with the number of cables and connections. The ATX and CPU cables are hardwired, but all other cables—two PCIe, six SATA, and five four-pin Molex—are modular. The FSP power supply even comes with two floppy connectors, which surprised us.

You can't go wrong with the FSP Everest 85Plus 500 W power supply. Both in efficiency and equipment, Everest fares well, although it doesn't particularly excel in any way, either, as our results on the following page will show.

FSP Everest 85 PLUS 500: Measurements

Efficiency in accordance with the 80 PLUS specification:

Efficiency under different load profiles:

Our test measurements confirm the 80 PLUS certification. At 230 V, the energy efficiency values of 86% to 88% under loads of 20, 50, and 100 percent of FSP's maximum rated power are very good. The results are a bit weaker at 115 V, as expected, but still in order. The respectable efficiency values maintain over almost the entire load distribution curve, except under very low loads where efficiency falls.

At 25 W load, only 70% power efficiency is achieved, but all in all, the results are still compelling. The FSP Everest's other tests were also good. Because the FSP Everest has four 12 V rails, power can be optimally distributed over the individual rails, preventing excessively strong amperages that are more difficult to keep stable (and safe for that matter).

SilverStone Strider SST-ST50F-230

The SilverStone Strider SST-ST50F-230 is the only power supply in this roundup without an 80 PLUS certification logo. Instead, the packaging says that the PSU can achieve an energy efficiency "of up to 80 percent." However, the Strider is very attractively priced at around $65—and the product's aesthetics match its price.

Of course, boring looks are usually irrelevant when it comes to power supplies. More decisive is the fact that this power supply is designed for 230 V operation-only (making it useless in North America; rather, this one is for our international audience). An 80 PLUS certification is not possible under these circumstances, as the tests mandate 115 V compatibility.

Despite the low price, SilverStone didn't skimp on cable management. All cables, even the ATX and CPU, are modular and round. The number of connections is satisfactory, too: two PCIe, six SATA, and six four-pin Molex connectors.

SilverStone Strider SST-ST50F-230: Measurements

Efficiency in accordance with the 80 PLUS specification:

Efficiency under different load profiles:

As expected, the SilverStone Strider didn't quite keep up with the 80 PLUS-certified products in these tests. The power supply completed the relevant measurements with energy efficiency values of at least 80% and maintained this number down to the comparatively low load of 85 W. When the load dropped further, efficiency nose-dived. At 25 W, the power supply only managed a weak 62 percent. The SilverStone Strider completed the tests to verify the inrush current, the behavior at overload, short circuit, and power outages without abnormalities.

The Strider is an unspectacular, but reliable power supply at a very reasonable price. It's suitable for budget-conscious PC users who are not dependent on 115 V support or particularly concerned about efficiency under low loads.

Test Configuration, Hold-Up Time, Inrush Current, Peak, And Short Circuit Tests

Hold-Up Time, Inrush Current, Peak, and Short Circuit Tests:

Efficiency According To 80 PLUS Spec., Temperature, And Standby Overview

Efficiency according to the 80 PLUS specification:

Standby and temperature measurements:

Efficiency Under Different Load Profiles

Results And Purchase Recommendations

The most important conclusion of this comparison test is that one does not have to spend more than $150 on a very good mid-range power supply. All tested models convinced us with high electrical quality and staying true to their 80 PLUS performance levels. But despite all power supplies achieving their advertised 500W and more, they showed individual strengths and weaknesses, making them more or less suitable for certain targeted applications.

Enermax Pro87+

In this roundup, the Enermax Pro87+ clearly and confidently takes home a test victory. With this PSU, everything is just right. The high build quality, excellent electrical quality, and equipment satisfy. In view of the 80 PLUS Gold certification, the price of around $140 isn't unreasonable, but it may still be too expensive for many users.

Cooler Master Silent Pro M500

The Cooler Master Silent Pro M500 is conspicuously inconspicuous. Thanks to very good efficiency at low loads, flat cables, and the silicon dampening frames, we recommend the Cooler Master especially for compact PC systems optimized for low power and low noise. Despite being a good fit for low-demand machines, the Cooler Master has enough reserves even to supply reliable power in high-performance multimedia systems.

FSP Everest 85Plus 500

The FSP Everest 85Plus 500 power supply is hot on Cooler Master's heels, but it doesn't quite manage the same high and flat efficiency curve. Furthermore, Cooler Master has more ideas on noise management and higher build quality. The gap between FSP and Cooler Master, however, is narrow.

SilverStone Strider SST-ST50F-230

For those who look primarily at price (and live overseas), both the FSP Everest as well as the SilverStone Strider SST-ST50F-230 are interesting. The latter power supply operates with good efficiency and has no weaknesses and a very clear price/performance focus. It's a shame it isn't usable in North America.

Four Cores For The Power Consumption Of Two

Intel’s Core i7 processor family has been available for a few months now, but despite being the undisputed performance champion, it is not yet ready for the true mainstream. Processors, and especially motherboards, are still too expensive; and the existing models deliver their impressive performance at equivalently impressive power consumption figures. But users looking for the best balance between performance and efficiency have an interesting new option now: Intel has released its first 45 nm Core 2 Quad processors that stay below a specified power consumption of 65 W--the Core 2 Quad S series.

Bringing Power Consumption Down

Although processors have reduced their power consumption by more than 50% since the good old Pentium 4 powerhouse, the addition of processing cores typically increases power consumption, at least during high load conditions. While we believe it is absolutely acceptable to live with a certain power requirement for high performance parts, everything should be done to keep power consumption reasonable while the processor is idle or running at a low load level.

Features Versus Technology

All modern processors implement power saving features. Cool’n’Quiet by AMD and SpeedStep from Intel allow the operating system to lower the clock speed and processor voltage to decrease power consumption. In addition, processors are capable of switching into halt states, which means they may interrupt execution for very short periods of time. Finally, some sections of modern processors can be completely shut down when they are not needed; this has traditionally applied to caches, but is increasingly being used for processor cores as well.

Manufacturing technology keeps improving, and processor manufacturers use these advances to introduce minor improvements to their products. Our article AMD’s Athlon Stepping Improvements, which we published in March of 2008, is a great example. We took four different processors that are all named “Athlon 64 X2 5000+” but are all based on different silicon steppings, and compared them.

Improvements at the silicon level can be used to accelerate processors within particular thermal envelopes, or to decrease processor power. Intel did the latter here, introducing additional Core 2 Quad processors named the Q8200S, Q9400S, and Q9550S. These are as fast as the regular versions, but they don’t require up to 95 W of power. All three models stay below the mainstream thermal envelope of 65 W, which finally makes Intel quad cores interesting for low-power systems.

Check prices for Intel's Core 2 Quad Q9400S

The Core 2 Quad S-Series

The new processors are all based on Intel’s 45 nm manufacturing process, and they basically utilize the same silicon as the regular, non-S models, centering on the Core 2 Duo 'Wolfdale' design, utilizing two processing cores. Intel puts two of these onto one physical processor to create a Core 2 Quad processor, which it refers to as Yorkfield. Supporters of monolithic processor design have criticized this approach, saying it would not result in a true quad-core. And indeed, the disadvantage is increased front side bus traffic due to inter-core communication. However, we believe that the real judge should be the benchmarks, and the results there have been nice: the Core 2 Quad delivers high performance at an acceptable power consumption level.

S-Series Models

The S-series currently consists of three processor models. All have exactly the same technical characteristics and performance as the corresponding regular models, but they are rated at a maximum thermal design power of 65 W. The Core 2 Quad Q8200S runs at 2.33 GHz and has two 2 MB L2 caches; the Core 2 Quad Q9400S is a 2.66 GHz part with two 3 MB caches; and the Q9550S top model runs at 2.83 GHz and has the full 6 MB L2 cache for each dual core die, resulting in a total of 12 MB L2 cache.

Power Matters

Having four cores run at only 65 W is a significant step, as this has been the thermal design limit for Intel’s Core 2 Duo processors since their introduction. This means that virtually any Core 2 Duo system that was designed to be energy efficient will now be able to work with these quad cores as well, effectively doubling the performance reserves.

However, Intel is not the first company to deploy a 65 W quad core processor for desktops, as AMD presented its Phenom X4 9100e in May 2008. Credit belongs to AMD as the first, but the 9100e wasn’t a high-performance part at only a 1.8 GHz clock speed. Not only is the Core 2 Quad faster when compared at the same clock speed, but the new S-series models at 65 W are available at mainstream clock speeds of up to 2.83 GHz. The higher frequencies mean that they are able to provide substantial performance even for traditional applications that may not be able to benefit from multiple processing cores.

Click here to view more images of Intel's Core 2 Quad and its vital stats in CPU-Z.

Check prices for Intel's Core 2 Quad Q9400S

Efficiency: Pay Attention To The Power Supply!

A key issue for creating a truly efficient system is the power supply. There are more and more really efficient units on the market, and the way to recognize them is an 80plus certification. This designation means that the power supply stays above 80% power efficiency (output power vs. input power) at 20%, 50% and 100% load. In addition, there are 80plus Bronze, Silver and Gold labels, which require even better efficiency levels:

Most power supplies without 80plus specification fall between 65% and 75% power efficiency at their given loads. Hence, it makes sense to look for a truly efficient device. Cost is the only limit: be prepared to fork out $200 for the nicest 80plus Gold and Silver products. The 80plus Bronze devices are more affordable, but they easily reach $100 or more.

Efficiency At The Lowest Loads

Another factor to consider is efficiency outside or below the specified 20% to 100% range. If you take a 1,000 W power supply and run a power efficient system such as the ones we used for this review, they will require only 40 W when idle. This means the power supply runs at a load of only 4%. In that case, even the best-designed PSUs start to become extremely inefficient. The result may be a 40 W system requiring 50-60 W when idle, and then it doesn’t make any sense to purchase efficient components anymore.

Power Supplies Used

We used three different power supplies to emphasize the importance of efficient PSUs. The first was the one we used for testing the Core 2 Quad S-series processors, a Fortron FSP220, which is a low-power, high efficiency PSU (220 W). We added an Enermax Pro82+ (PR425AWT, 425 W), which is an 80plus Bronze device that can be considered a very good choice for upper mainstream systems. Finally, we used another Enermax power supply, the EG565P-VE, which is roughly three years old and not as efficient as the others. Check out the benchmark section called Power Supply Impact for details.

The Enermax EG565P is a rather old device. We used it to show the impact of an inefficient power supply unit on an efficient PC configuration.

Check prices for Intel's Core 2 Quad Q9400S

Test Setup And Synthetic Benchmark Results

Synthetic Benchmark Results

SiSoftware Sandra 2008

PCMark Vantage

Check prices for Intel's Core 2 Quad Q9400S

3D And Game Benchmarks

3DMark Vantage

Game Benchmarks

Check prices for Intel's Core 2 Quad Q9400S

Audio/Video Benchmark Results

Audio Benchmarks

Video Benchmarks

Check prices for Intel's Core 2 Quad Q9400S

Application Benchmark Results

Check prices for Intel's Core 2 Quad Q9400S

Power Consumption And Power Supply Impact

The Core 2 Quad Q9550S has the full 12 MB L2 cache, while the Q8200S only has 4 MB total cache memory. The Core 2 Duo E8500 runs a total of 6 MB L2 cache. In the end, the idle power difference is insignificant. All three processors, including the 2.83 GHz Q9550S, are very efficient when they idle.

The dual-core requires the least power at peak load, while the entry-level quad-core isn’t too far behind. The Core 2 Quad Q9550S boosts the power consumption from 39 W when idle to 106 W at CPU peak power, a 67 W difference. Keep in mind that the power supply adds to the power consumption with its efficiency of 82%, so the Intel low power quad core clearly stays within its power specifications.

Power Supply Impact

Here are the power consumption results measured with the three different power supplies. The low power 220 W FSP220 power supply by Fortron is the most efficient, as it was designed for low power systems. Enermax’s Pro82+ also does very well, requiring only a little more idle power—it is even more efficient with the Core 2 Quad Q9550s once you throw CPU load at it.

Beware of older PSUs, though. The EG565P-VE by Enermax is a product that is three years old. Efficiency wasn’t on the priority list at that time, which results in 6 to 11 W of additional power consumption if you use an older device. It clearly makes sense to purchase an efficient and suitable power supply if you want a truly efficient system.

Check prices for Intel's Core 2 Quad Q9400S

Power And Efficiency With 3DMark And PCMark Vantage

3DMark Vantage--Power and Efficiency

We divided the 3DMark score by the total power used in watt-hours, giving a nice efficiency score in performance per watt, which the dual-core dominates. Four cores don’t give much of a benefit here, especially since the dual-core runs at faster clock speeds.

PCMark Vantage--Power And Efficiency

Next, we did the same with PCMark Vantage. Interestingly, the Core 2 Duo E8500 and the Core 2 Quad Q8200s show exactly the same performance per watt in this benchmark. The Q9550s is superior, as it delivers more performance at a slight increase in power consumption.

Check prices for Intel's Core 2 Quad Q9400S

Conclusion

Intel has introduced low-power, 45 nm quad core processors for the desktop market. Although AMD was first to release such a product, the Intel processors are exactly as fast as the regular 95 W versions.

The only real difference besides the power consumption level is their pricing, which is higher as a result of the lower energy use. While the Q8200 costs $163, the Q8200S has a $245 price tag. The price difference between the Core 2 Quad Q9400 and the Q9400S is $101. The Q9550S top model is only $52 more expensive than the regular version.

We believe this is a lot, as Intel has been offering low power 65 W Xeon quad-cores for a while now, and hence does have some experience with these parts. On the other hand, their new S-series models represent a unique selling proposition. After all, AMD currently does not offer quad-core processors at upper mainstream clock speeds and mainstream power envelopes.

The 8200S may not make a lot of sense when compared to a regular Core 2 Duo E8500 or E8600, but the Q9400S and Q9550S deliver significant performance within the mainstream 65 W TDP specification. Here’s hoping that AMD will be able to offer a low-power Phenom II soon as well.

Check prices for Intel's Core 2 Quad Q9400S

Overdrive: A Quick Recap

Chris: Hopefully everyone is enjoying a couple of days off with family and friends during this end-of-year holiday season. From everyone here at Tom's Hardware, here's hoping the break is treating you well.

Now that the Overdrive overclocking competition is over and done with, we wanted to give you a more in-depth look at what went down during the weekend when Team USA earned its first-place finish. In case you didn’t know, we held the world finals two weekends ago, from December 13th to 14th, with participants from five countries: France, Germany, Italy, Taiwan, and the United States.

The finals of the competition—whose participants were chosen from among the readers of our sites—was held in our Paris office using a test rig based on Intel's Core i7 and an X58 platform manufactured by MSI.

The Event's Vitals

The contest actually began on Friday, December 12 with a welcoming ceremony for the participants. What followed was comprised of more than 16 hours of benchmarks, 10 monitors, more than 10 TB of hard drive space, enough graphics power to make Pixar jealous (Ed.: OK, maybe not), 13 contestant, and six Tom's Hardware staff to supervise and validate scores. It also featured over 200 GHz in cumulative clock speed between the execution resources involved: five Core i7s, each with eight logical cores, overclocked to more than 5 GHz. Also on hand were 450 liters of liquid nitrogen at -196°C, 10 representatives sent by our sponsors, more than 10 kilowatts of power (if you were to total the output of the PSUs in play), and over 40 liters of beer.

If you'd like to see the blow-by-blow of how the competition transpired as it was happening, check out our event landing page, which hosts all of the news, picture stories, and editorial content.

Preparing For The Contest

Let’s start at the very beginning, with the preparations. In order for the contest to succeed, it needed to center on hardware proven easy to overclock and able to take punishment from some of the most talented overclockers in the world. After all, a C7 and a budget 200 W power supply with S3 graphics isn’t is very exciting. So, unlike the semi-finals, where Core 2 Duos were used, we shifted focus to today's fastest processor available: the Core i7 Extreme Edition (965). Obviously, an X58-based motherboard was used (MSI's Eclipse), along with a 1,010 W power supply furnished by FSP.

Two 1 TB hard disks were used (one as backup), with 3 GB of Kingston RAM and Windows Vista Ultimate (32-bit) as the OS. Added to that were a Logitech G5 mouse and G15 keyboard, the latter of which wasn’t to everybody’s liking, incidentally, since it was a QWERTZ model (though the Germans didn’t complain at all).

The Overclockers Arrive; We Dine

The overclockers arrived all through the day on Friday the 12th, and oddly enough, the one person who didn’t make it that day was a Frenchman: Marmott. Why? He lives in England and missed his flight, and so had to find another means of transportation to the Parisian office.

To welcome the overclockers, we decided to give them a taste of French cuisine, so off we went to a restaurant. The waitress deserves an honorable mention for being a good sport. She didn’t quite understand what overclocking meant, and it seems she had our group figured for a bunch of crazy, if harmless, weirdoes.

The rapport among the different teams was very good and some even realized they’d met before: one member of the Taiwanese team had covered the MSI contest in Taipei, where Florian_76 competed.

Once dinner was over, the participants headed to get some needed rest in preparation for what promised to be two very busy days.

Day 1: Delays And Water Cooling

The first day had its share of trials and tribulations. Part of the fun of an overclocking competition is watching an American drink liquid nitrogen (temperature -196° C, remember). Even if you’re fitted with an American stomach, please don’t try this at home (Ed.: Pierre, I'm hoping the LN2 was never actually swallowed).

Running Into Trouble

All of the teams ran into problems, but the French took the prize in that department. As mentioned earlier, Marmott missed his plane and had to spend the night in the station in London before catching the first train to Paris. As a result, he didn’t arrive until 11 AM, and Florian had to hold down the fort until then on his own. On top of that, the French overclocking expert is also a tad absent-minded: while chatting with a sponsor (Kingston), he didn’t realize that he was installing a 64-bit version of Windows Vista when he was supposed to be using the 32-bit version, so he had to install his operating system over again.

Among other problems that cropped up, our Taiwanese friends thought they had a defective DIMM, then finally realized (at the end of the contest) that the RAM was working fine, and the memory controller built into the processor seemed to be the source of the problem. Their setup would operate only in dual-channel mode (using two DIMMs instead of three).

Water Cooling, Air Cooling

Before chilling the processor with liquid nitrogen, the majority of the teams did preliminary tests on the components with standard air cooling—this is a way of checking to see that the processor, motherboard, and graphics card are not defective. It was a good thing that they did this, too. The Americans, for example, realized that the heatsink on their northbridge was improperly installed.

The Taiwanese decided to go for qualifying scores quickly, and decided to water-cool the graphics card the first day. They also chilled the water with liquid nitrogen, using salt (a request that surprised us a little) to keep the water liquid at sub-freezing temperatures (avoiding the formation of ice in the circuit).

Varnish

In addition to the stability tests, another important task is taken care of the first day: preparing the equipment. Liquid-nitrogen cooling has inherent risks and obviously the hardware needs to be insulated—computer components don’t really deal with condensation well. Each team had their little tricks in this department, such as erasers for the Germans, unpleasant-smelling varnish for the French, Vaseline for the Americans, and so forth.

End-Of-Day results

At the end of the day, the Taiwanese managed to run the majority of the 3D tests, whereas the other teams had run only the CPU tests. Among the little problems that befell contestants: one of the German team's results had to be canceled because they’d used the supplied graphics card for installing Vista instead of the GeForce GTX 280 to be used in the tests. The first card was only there so that the OS could be installed without using the main graphics card, enabling it to be prepared for the benches.

Day 2: An Interesting One

The second day began with the teams in high spirits. Everyone was there—even the French! But the good vibes were not to last. From defective multi-meters (for the Germans), to results that didn’t scale at all with clock speeds and forced Windows to be installed a third time (for the French home team), to problems with Turbo Mode—all was not rosy.

A Word About Turbo

The problem with Turbo Mode—a unique feature to Core i7, but also available in a less-evolved form (IDA) on the Core 2 Duo—is that the overclocker doesn’t really control the clock frequency. Simply put, if your processor operates at 4 GHz at a voltage of 1.5 V in standard mode, certain single-threaded benchmarks could be run at 4,200 MHz, for example, and cause crashes. Remember that Turbo Mode increases the multiplier coefficient when the number of cores in use is low (1 or 2 of the 8 logical cores). Activating Turbo also activates EIST (SpeedStep), which can cause problems with stability. EIST reduces the voltage at idle, but an overclocked processor often needs a higher voltage. In practice, the easiest approach for overclockers is to disable Turbo mode, even if certain protections do remain active on the MSI X58 Eclipse.

Liquid Nitrogen And Oxygen

Our overclockers were all hungry for a win, and so the liquid nitrogen flowed like water. The teams were chilling not only the CPUS with the stuff, but the GPUs too. For once, however, the chipsets weren’t fitted with their own nitrogen canisters; the X58 is in fact nothing but a PCI Express controller, so the usual components of the northbridge are in the processor package with the Core i7.

One interesting safety issue is that using so much liquid nitrogen can cause a problem with the oxygen concentration in the air. Whereas the normal oxygen content in the atmosphere averages between 20.5 and 21%, by mid-afternoon we were measuring only 19.2%. That’s a little low, so we had to air out our offices to bring the oxygen level back up.

The Winners

In the end, the Americans won the contest: Jake Crimmins, Jeremiah Allen, and Ton Khowdee posted the best scores with their Core i7 965 Extreme and MSI X58 Eclipse motherboard. The race was long and ended in a photo finish, with everybody’s scores improving right until the last minute.

Two Interesting Overclocking Results

We have a little surprise for the overclocking lovers reading this report. In addition to the Core i7, our contestants were able to try out a Phenom II, from the next generation of 45 nm Phenoms expected in January. 

4.95 GHz for the Phenom

Though AMD claims they’ve reached close to 6 GHz, we were able—with the help of JMax—to reach “only” 4.957 GHz. We should point out, though, that the factory speed of the processor in question, a Phenom II X4 940 BE, is only 3 GHz. We used the processor on a Gigabyte AMD 790GX motherboard, with DDR2 RAM (as required by the AM2+), and a liquid-nitrogen cooling system. This processor may have the potential for higher clock rates, but it had the unfortunate tendency to freeze as soon as the temperature dropped below -70°C.

To see what the processor was able to do, we ran a few tests at a stable frequency (4 GHz). The processor was able to perform a SuperPi 1M in 17.769 sec. During other tests using samples, the processor was able to boot up even when chilled to close to -200°C!

(Ed.: In preparation for benchmarking Phenom II, I asked AMD about the overclocking results garnered for our Phenom II sample under the influence of LN2. The answer was that the HyperTransport link was limiting scalability. If you're going to take the chip to extreme frequencies, you'll want to drop down to 1 GHz or so on the HT interconnect rather than its default HT3 speeds. On air, however, HyperTransport shouldn't be a limiting factor.)

5.28 GHz For Core i7

With a limited amount of time left—the contest was over and the other competitors were packing their gear—and with a limited quantity of nitrogen, we asked the French team to overclock the Core i7 for maximum frequency. They managed to log 5.287 GHz with the Core i7 965 test sample, and even went as high at 5.3 GHz (though not stably). The cooling system used liquid nitrogen and the motherboard was an MSI Eclipse, obviously using the X58 I/O hub. The memory was DDR3 supplied by Kingston, and no problems were encountered with it. In fact, with a frequency of 5.287 MHz, our processor placed in the Top 10 worldwide, copping ninth place (according to Ripping.org). It’s a result our overclockers can be proud of.

A Final Word Of Thanks

We’ll end by thanking everybody, starting with our sponsors. We’d like to thank MSI for the Eclipse motherboards (and the P45D3s during the qualifications) and for the army of engineers they sent in to help us out. We also want to thank FSP for the power supplies (1010 W powerhouses) and also for sending us 1200 W models, since the 1010s posed a problem for certain overclockers. Again, Kingston supplied the DDR3-2000 DIMMs used during the contests, Samsung the hard disks, and Logitech the keyboards and mice. We’d also like to thank LDLC and Gainward for their invaluable aid during these finals.

Then There’s You And The Overclockers

We also want you thank you, our readers. You sent in comments, complaints, and expressions of support; you offered to participate; and you even chose the eventual participants via our forum. You’re the ones we organized the contest for, and you deserve the credit for it happening.

And last but not least, we want to thank the overclockers themselves—for their good humor, and their ability to deal with all of the little problems that are inherent in this kind of competition.

A big “thank you” to all. Again, happy holidays!

After 16 hours of intense overclocking over two days, we have a world champion. Team IRONMODS from the United States - Jeremiah Allen, Jake Crimmins and Ton Khowdee, achieved the best scores during our Overdrive event, overclocking Core i7 Extreme 965 systems on X58 Eclipse motherboards by MSI. The race was long, but it also was almost neck and neck, as the scores kept improving almost every minute.

The Taiwan team CBB had a head start, as they quickly posted scores during the event. France started a bit late, with one team member being delayed on day one. Italy and Germany were always very active and the US team took some time to get up to speed. But their efforts were worth it: they were fastest in two 3DMark benchmarks, they dominated in wPrime 1024M and 32M. These four wins, combined with good scores on the other benchmarks, assured IRONMODS win.

Italy was close behind, with two wins in SuperPI 1M and 3DMark 05. Germany reached third place, with the strategy being to not fall too far behind the others in any benchmark. France was fourth and Taiwan fifth, as they had issues running the Core i7 in triple channel memory mode.

Final results:

As you can see, all the teams except Taiwan, which had those pesky hardware, were able to reach top scores in at least two benchmark categories. Even France, which took fourth place, reached two Overdrive records.

Congratulations to the US team, IRONMODS, which proved to be a magnanimous winner, thanking all the other guys for being excellent competitors and great sportsmen. We were glad to have the opportunity to host all these guys, and we'll provide a summary article next week.

Introduction

WR Team France had a late start, as one team member arrived late. However, they got some support by the Kingston representatives and they managed to get top scores in SuperPI 32M and PIFast - both are still valid at Sunday 11 AM. But there are still five more hours, so the race remains tight.

Italy: Turning Frustration Into A Challenge

The Italian team seemed to be frustrated. They didn't reach the clock speeds they were looking for, and as a result, they weren't hitting very impressive scores. But they kept trying, and things are looking better for them, as they still hold the records in SuperPI 1M, wPrime 1024M, and 3DMark01.

Blowing Off Steam

Whether you're successful or not, you need to take a break once in a while. Although it doesn't look like it, overclocking on liquid nitrogen is an intense job.

Maintaining The Ideal CPU Temperature

Massimilliano from the MemoryExtreme Team Italy has an eye on the CPU temperature, which is extremely important. Cooling the processor down too much might result in a so-called cold bug, which causes the CPU to literally freeze, and it will stop working for a while. As you'd be losing time, you want to avoid this.

Core i7 is much more sensitive than Core 2 it seems, so a 20°C temperature difference might be outside of the ideal temperature operating range. According to various teams, the ideal operating range is at around -95°C and -100°C.

Taiwan: Get Us Some Salt!

Taiwan's team CBB had a rather strange request. They asked us to get some salt. The reason behind this can be seen on the photo. As we mentioned yesterday, the guys utilize a water cooling solution for the graphics card, and they have been pouring liquid nitrogen into the water reservoir to reduce the water temperature. However, the water would of course start to freeze once it reaches 0°C. Adding salt is a simple and effective way to move the freezing temperature further down. This introduces additional margins for cooling.

Tom's Hardware Tracking Scores

Julien and Tristan from Tom's Hardware France are discussing the results, which we tracked on paper and on a notebook. We projected them to one of the office's walls so that everybody is up to date throughout the competition.

Pizza And...Fruit?

Almost everyone likes pizza, so we decided to keep it simple and ordered a nice little stack of them. We added some healthier snacks to the mix as well. Everyone knows you can't overclock well without your daily allotment of vitamins.

Liquid Nitrogen Refills

We had multiple 25-liter canisters, which we used to refill liquid nitrogen from a larger tank located close by.

Italy: Results!

Italy shoots, Italy scores. As you can see, Massimiliano is happy now.

Day 1 Evening Scores

At the end of day 1, only the Taiwanese team submitted results in all benchmark categories. As you can see, the playing field is rather even.

The Top Overclockers Meet In Paris

Group photo: There are all the participants next to the hardware, just some minutes before the event started.

France--One Guy Was Late; They're Catching Up

The French team consists of two overclockers. Florian_76 arrived in Paris on Friday, like most of the other guys. However, his teammate Marmott missed his flight on Friday and had to come in Saturday on morning. As a result, the WR Team from France had to start 90 minutes later. Unfortunately, they had another issue, as they accidentally installed the 64-bit version of Windows Vista Ultimate. Since we need the same conditions for everyone we had to insist on using Vista-32.

Frankreich--Catching Up

The French WR Team checked the hardware first, as they wanted to be sure they wouldn't run into issues. There will be plenty of time to use liquid nitrogen. They decided to insulate the motherboards using varnish.

Taiwan--LN2 After Only Two Hours

The Taiwanese team CBB decided to focus on the processor. Their graphics board is cooled by a liquid cooling solution.

Taiwan--Liquid Cooling The Graphics Card

The Taiwanese team also was the first to clone its Windows Vista installation using Norton Ghost. They decided to do this so that they can restore the entire system installation within minutes in a worst-case scenario. Otherwise you'd have to reinstall Windows and drivers, which takes at least half an hour even on this hardcore systems.

The liquid cooling solution on the bottom of the photo looks home brew, but in fact it seems to be working great.

Germany--Diligently Preparing Hardware

The German team BenchBros started with carefully insulating the graphics card using rubber from Faber Castell. Usually this material is being used as rubber to remove pencil writing. Clearly, simple tools seem to be most efficient.

Germany--Steady...

BenchBros from Germany takes a lot of time to prepare and to fix potential issues. Let's see if their scores will prove them right.

USA--Fixing Possible Issues Before They Arise

The US team IronMods had to spend some time to reattach the northbridge heat sink, which wasn't mounted well. It did not lay flat on the X58 chipset, which could have caused insufficient cooling at high clock speeds.

USA--Professional Test Bench

Team IronMods installed their test system onto an aluminum test bench, which they brought to the event. This is one of the most professional test benches we've seen, as it also include a 92 mm fan to cool the motherboard from the rear side of the PCB.

Italy--Thorough Component Checkups

MemoryExtreme Team Italy's Massimiliano spent time to check components carefully, then he installed Windows. The two team members found solder points to measure voltage of the key components on the motherboard. Finally, they did extensive test runs with the Kingston memory to find out the best combination of timings and clock speed ahead of the actual race for results.

Italy--Test Benches by DimasTech

The Italian team brought a test bench by DimasTech, which is a helpful tool for test engineers or overclockers.

MSI--We Wanna Try, Too!

Our sponsors from MSI set up their own system and tried their hands at overclocking, too.

FSP--More Powerful PSUs

Since the Everest 1010 PSU units weren't powerful enough for extremely overclocked GeForce GTX 280 cards due to unstable power on the 12 V rail (remember that the specified 1,000 W output are the total output), FSP brought modified units and some 1,200 W backup units. Clearly, the company takes the event very serious.

Kingston--Excellent Results With HyperX DIMMs

Kingston also sent two representatives to welcome our overclockers. There was no reason to complain, as everyone was very happy with the HyperX DDR3-2000+ DIMMs. Stephane from Kingston (left) took the opportunity to announce the new HyperX T1 DIMMs, which come with a larger heat spreader. The guy on the right is Florian Charpentier, our editor in chief for the French Tom's Hardware site.

Tom's Hardware, Hard At Work

Tom's Hardware's Overdrive Overclocking Championship would not work without Tom's Hardware staff. Florian Charpentier and Matthieu Lamelot from our Paris team took care of local organization and coverage.

Introduction

On November 29th and 30th, we hosted the Italian leg of Tom’s Hardware’s overclocking contest. You can read more about the competition that saw the American team selected right here. We hosted four teams, each taking its hardware as far as it’d go using liquid nitrogen.

In order to finish the contest, the teams had to complete ten tests. Five of these were aimed at assessing the overclocking capabilities of CPU, motherboard, and memory. Those included PiFast, SuperPI 1M, SuperPI 32M, Wprime 32M, and Wprime 1024M. The other five factored i3D performance into the equation with Aquamark, 3DMark01, 3DMark03, 3DMark05, and 3DMark06.

The scoring system granted 100 points for first place, then 80, 60, and 40 points, respectively, for second, third, and fourth place finishes. The sum of each team’s points decided the winner.

In addition to taking the hardware to its limit, overclockers also had to pay heed to how aggressive they were being. Any burnt component resulted in a 5% penalty to the final points total.

Faced with the limitations already experienced by our US and Germany competitions, the teams were forced to make some compromises and tough choices in order to complete the battery of tests.

If you’d like a closer look at the hardware used in the Italian trials, check out The Hardware Behind Overdrive.

Give Us The Liquid Nitrogen, Now!

After a week spent preparing the venue, the hardware, and the liquid nitrogen, Saturday morning arrived and the Italian team was ready to host the group of competitors. Despite ugly weather, all of the competitors made it on time for a 10:00 AM start.

Only a few minutes were devoted to the introduction and other pleasantries, as the competitors were eager to get down to business. Very quickly thereafter, they got their hands on 230 liter liquid nitrogen canisters, filling their 10, 30, and 50-liter dewars. First, GiorgioPrimo and Leghorn, followed by F.O.G.N.A, C1rcu1tburn3r, and Principino84. Then Hiwa-Zilla-Qballe took his turn, and finally Wasky and Marco74 got their fill of liquid freeze.

Although identical components were used for each team, there are always differences at the transistor level affecting how far a given part will scale. These are imperceptible for most users, but very much apparent when you overclock competitively. The distribution of hardware, therefore, was handled by pure luck, using numbered cards and a draw.

The guys then began to assemble their machines, install the operating system, and load various utilities and benchmarks, which took us to mid-afternoon. Optimizing the software was a very important step, because a streamlined system allows you to focus all the computing power on the test, then obtain low times or high scores, depending on the benchmark.

Before moving to the benchmarks, the teams took a few minutes to mod their hardware, another essential step without which it would be impossible to achieve good results. Soldering iron in hand, they made the changes needed in order to deliver more voltage to important components, like the CPU and GPU. This sets the stage for increasing clock speeds, reducing the timing of memory, and making other changes necessary to get better performance.

Start The Benchmarking

After entertaining the "online spectators" with some photos posted on the forum, and in an initial photo gallery presentation, it was time for the real fun to begin.

Memory Extreme was the first team to go from modding to benchmarking, closely followed by team-Zilla Hiwa-Qballe.

Unfortunately, Extreme Hardware lost a lot of time due to a problem with the operating system’s replication software, forcing them to delay the first benchmark. Replicating the operating system with software such as Acronis TrueImage or Norton Ghost is very important, because an excessive overclock can corrupt the operating system. We certainly cannot afford to lose the time that would be necessary to reinstall everything in such a case, especially after having thoroughly changed the system settings.

Team Next Hardware instead took a slower, more calculated approach, postponing the system start-up until late afternoon, and taking their time to modify and tweak the system.

Memory Extreme got off to a great start in the first leg, registering some great scores in the SuperPI 1M metric. The tests that followed, unfortunately, brought to light some hardware problems with the motherboard we were using, while the Intel E8600 CPU proved to be a high-quality component.

Team Hiwa-Zilla-Qballe followed Memory Extreme, gained some points in the 2D test, and then improved on those scores even further.

Disaster!

From the back of the room comes a desperate scream, followed by two words: "it’s gone!" And with that, GiorgioPrimo declares the premature death of a motherboard. Some attempts are made at resuscitation to avoid suffering the 5% penalty for burnt components, but nothing more can be done. Another "tombola" for Leghorn, who draws the number to assign them a second motherboard, and Memory Extreme gets to work installing the system for a second time. That’s going to leave a mark.

Meanwhile, Extreme Hardware started putting up some scores, checking in a thousandth of a second behind Memory Extreme in SuperPI 1M. A thousandth of a second could just be called bad luck, however this is a clear sign of how aggressive the challenges are in extreme overclocking. The teams compete against each other at a high level—so high that even one thousandth of a second can make all of the difference in the world.

Next Hardware also starts to put some liquid nitrogen into its CPU pot and slowly tries to increase the maximum level of performance in order to put some scores up on the board.

We start to celebrate—thinking, "only one motherboard burnt, what luck!"—when, from the other side of the room, comes another cry. This time, it’s C1rcu1tburn3r and his colleagues F.O.G.N.A. and Principino84. Unfortunately, we have another motherboard gone. Like Memory Extreme, Extreme Hardware also seeks to recover the board to avoid the penalty, but the team is forced to draw another number.

More scores and many liters of liquid nitrogen later, the bell rings. That’s all for today. Everyone back to the hotel and we’ll re-convene tomorrow.

Friends, Together

Extreme overclocking is not just a hobby or a pastime—it is a real passion, which leads its enthusiasts to spend full nights in front of a monitor (next to a dewar of liquid nitrogen, even). Our competitors are no exception. All of the teams, in unison, ask to continue through the evening, allowing themselves only a break for dinner.

Interestingly, the same thing happened here in the States. None of the overclockers wanted to break for the night. We ordered pizzas and went for as long as was allowed.

The group of Italians would have liked to concede, but was forced to decline, as going on at night would have represented a violation of the regulations. All of the other nations’ teams had the same amount of time at their disposal. Then there’s the safety issue; we are dealing with dangerous materials, and so despite their great expertise, we preferred them to have had a few hours of sleep behind them before returning to the test bench.

Instead of burning the midnight oil, we spent a wonderful evening at dinner, where the rivalry between the teams was set aside and the discussions branched out beyond overclocking. We were happy to discover that an atmosphere of friendship and mutual respect reigns. All the guys who participate in our event, before being rivals, are overclockers and friends, linked by the same enthusiasm for a common effort, which always pushes them to get in the game and improve, thousandth of a second after thousandth of a second.

Wake Up And Coffee

At 7:00 AM Sunday morning, all the teams were ready to get going again. Extreme Hardware and Memory Extreme struggled with their new motherboards, while Hiwa-Zilla-Qballe and NextHardware made some system adjustments before powering up and starting their testing anew.

The liquid nitrogen cooling systems had to be rebuilt from scratch because bringing the temperature to more than 100 degrees below zero causes ice and condensation to form. Leaving those systems installed throughout the night, without taking the right precautions, would have meant finding the hardware in a puddle of water the next morning with devastating consequences. Therefore, at the end of the previous benchmarking session, everything was dismantled and dried so it was ready for reassembly the next day.

About an hour and a half after the teams’ arrival, an almost-surreal silence pervaded the room, with those working on the systems absorbed in their thoughts and those gazing concertedly at the monitor thinking about whether it would have been wise to raise a voltage, change a timing, or increase a frequency. The conversations are almost in whispers. There is a need to do something to change this rhythm and so, we decide it’s time for a coffee break.

A few minutes later, normal conversation seems to resume and systems are ready for a new shower of nitrogen.

Between Moments Of Panic And Despair

If, on Saturday, a few small victories were won, Sunday was to be the day of truth, and the teams dedicated themselves to fine-tuning. The hours ahead would reveal the winner.

Memory Extreme keeps us all on eggshells, cursing their motherboard. The randomly-chosen components seem to be affected by an inexplicable evil. The system didn’t start or it’d restart and freeze, even at the lowest frequencies—practically factory settings. How could it be possible that this team had been so unlucky? Fortunately, it turned out that the problem was in the operating system, which destabilized the tests the day before. Restoring the installation images allowed Memory Extreme to get back in the race.

Meanwhile, new numbers were coming from the other teams—all except Extreme Hardware, who, having installed the new motherboard, seemed to be having other problems. Shortly afterward, C1rcu1tburn3r, Principino84, and F.O.G.N.A. had a most unfortunate report: a second motherboard burnt. The regulations provided only one substitute board for each team, and so with this, Extreme Hardware has no more motherboards available to let them continue the event. C1rcu1tburn3r makes an extreme attempt to save the board by trying to replace an electronic component, but it cannot be revived. Unfortunately for Extreme Hardware—a team that started out very well the first day—the challenge ends here.

Next Hardware’s Turn

After Memory Extreme and Extreme Hardware—who between the two teams have now disposed of three motherboards—it is now Next Hardware’s turn to show what it can do. A now-familiar odor of burning spreads around the room, and sadly it comes from the video card. Wasky and Marco74 are in the testing zone, already showing some good scores and are at the most critical point, when the wave of bad luck hits them too, forcing them to stop.

They, too, look to save their video card, but some tests show video artifacts, an unambiguous sign of a damaged card. Next Hardware is not put off entirely, and forgets the 3D tests, focusing instead on those in 2D.

What has happened so far is more than enough to demonstrate just how delicate the practice of extreme overclocking really is. Just one nanosecond, a parameter just over the limit, and a card worth more than $500 goes up in smoke. This lesson teaches us another thing—extreme overclocking is not for everyone. It takes a lot of experience, money, and above all, the patience of a saint, because a moment of distraction can mark the end of the game.

Head To Head: Memory Extreme Versus Hiwa-Zilla-Qballe

With Extreme Hardware out of the game and Next Hardware still recovering from the shock of frying a graphics card (and concentrating on the 2D tests), the war of the titans begins.

GiorgioPrimo and Leghorn make a quick start and begin to mark one score after another. Unfortunately, it is not possible to push the system to the max, and Saturday’s results cannot be replicated. However with a clear strategy in mind, the two teams churn out one test after another, improving minute by minute.

Team Hiwa-Qballe-Zilla, who have already scored and improved on four out of five 2D tests—with good times and no need to change motherboards—have more time to administer additional tests. With the video card modded and installed on their system, it is time for 3D.

GiorgioPrimo and Leghorn are going like bats out of hell. GiorgioPrimo, not wanting to be disturbed, constantly follows the results display and plans strategy. Hiwa, Zilla and Qballe refill the system pots with liquid nitrogen and start on the 3D tests.

Meanwhile, Next Hardware busies themselves with 2D, marking excellent results and improving even further upon repetition.

Two hours from race finish, the tension rises, liquid nitrogen flows freely, and benchmarks pour from the monitor. The tension can be felt in the air—it’s only an hour until we know who will grab the tickets for Paris.

Only One Remaining

Wasky and Marco74 were fighting to improve their times, preferring to test 2D. Hiwa, Zilla and Qballe tangled themselves in AquaMark and 3DMark, while Memory Extreme passed through all the benchmarks, trying to improve their positions.

Unfortunately, less than an hour from the end, team Hiwa-Zilla-Qballe acknowledged some problems in the system. The first diagnosis regarded the video card, which does not seem to work correctly. After several scores in AquaMark, which were improved several times, the system refused to remain stable with any degree of overclocking.

As if that weren’t enough, a deeper analysis revealed that the motherboard was suffering, reducing performance. The team was thus forced to significantly cut frequencies practically to default levels. But even in this condition, the instability of the system made it impossible to complete the tests. Therefore, some precious points were lost, and with time running out, they couldn’t even think about preparing a new motherboard.

Meanwhile, GiorgioPrimo and Leghorn were absorbed in their world—a parallel reality. The room was full of people coming and going, talking among themselves, laughing and joking. But nothing distracted the two members of Memory Extreme, who had shown a tremendous ability to concentrate. Leghorn sat in front of the screen and keyboard while GiorgioPrimo, with liquid nitrogen dewar in hand, was careful to observe the results, fill the CPU and GPU pots and consult with his companion. Test after test, score after score, the times got better and the scores got fatter.

When time ran out, everyone already knew the verdict: Memory Extreme was the Italian winner of the Tom’s Hardware overclocking contest!

Results

Team Memory Extreme, after the first day’s difficulties and those encountered on the morning of the second day, changed strategy. Following all the benchmarks to mark a score, they then improved all the results one by one. It seems that the strategy was successful—Memory Extreme reached the first position in 9 out of 10 specialties, followed by Team Hiwa-Zilla-Qballe. It was third place for NextHardware, while closing the ranks was Extreme Hardware, due to two burned motherboards putting them out of the challenge.

You would think that an overclocking championship is all about setting up the system as quickly as possible to be sure you can get first results quickly. While that is true, we found that the approach is different for all the teams. After only two hours, some were still working on hardware modifications, while another team already commenced to pour liquid nitrogen into its processor cooling device. We provided two motherboards and additional graphics cards to facilitate installation and preparation: While the systems are installed on an old graphics card, other team members can mod the GeForce GTX280 card.

France

The French team had to start 90 minutes later, as Jean-Baptiste Marmot missed his plane yesterday. We'll provide more team to the French team WR in the end to catch up. The guys start with installing the operating system, as they say this takes the most time. Components are also checked on the way, and they started insulating motherboard components for the use with liquid nitrogen.

Germany

Typical German virtues drive the German BenchBros overclocking team, as they first make sure that all components are operational. Then they started working on insulation of the GeForce GTX280 graphics card and the motherboard. BenchBros doesn't use insulating foam, but a soft rubber from Faber Castell in Germany. This is typically used to remove pencil writing, but it is perfectly suitable for insulating motherboard components, as it bends like modeling clay, but has a lower humidity ratio.

Italy

The MemoryExtreme Team Italy starts with setting up Windows, but they check the hardware at the same time. While they do this, they make some modifications as well, as they don't trust the motherboard voltage readings: The Italians check the correct solder points to read the voltages for important system components such as the processor and the memory. Once this is done, they try how fast they can go with the memory. They say it will be a few more hours until they start working with liquid nitrogen.

Taiwan

The Taiwanese team CBB, headed by Coolaler, seems to be incredibly efficient, as it was already handling liquid nitrogen after only two hours of preparation. While they setup Windows on one motherboard, they make modifications and apply insulation to the second board. At the same time, they setup a liquid cooling solution for the graphics card, which also was up and running.

US

The American team IronMods starts with testing all components at stock conditions, as they want to be 100% sure that the hardware works properly. A good example for the necessity of this checkup was the X58 heat sink, which didn't sit firmly on the X58 north bridge. This could have prevented the team from reaching best scores, but they found and fixed the issue. Another team member took care of the voltage modding for the graphics card, although the GTC280 is still running on the standard air cooling solution. The last part of the preparation was cutting out templates of insulating foam.

{{Paris, December 13, 2008: The Tom's Hardware Overdrive championship is definitely heating up. Today and tomorrow, five teams from France, Germany, Italy, Taiwan and the United States will be battling for bragging rights on having the fastest benchmarks and highest CPU frequency. Intel's state of the art Core i7 systems will be the new playing field for our competitors.}}

Our first prize is a $5,000 gift certificate, which the winners can redeem in their home country. To our knowledge, this is the first overclocking event that is based on Intel's Core i7 and we're expecting speeds in excess of 5GHz.

Five Countries

Overdrive began with overclocking trials in the United States on November 8 and 9, continued in Germany on November 21, Taiwan on November 22 and 23 and concluded the trials in France and Italy on November 29 and 30. Up to five teams competed in each country, and the winners were flown to our Paris headquarters for this weekend's grand finals. All the teams brought their overclocking tools needed to modify hardware. Some of them even brought family members to witness the fun with them.

Scores

All the Overdrive trials were based on the same components, which makes it possible to compare the overclocking results and benchmarks reached in our five countries. the US team was the strongest in SuperPI 32M, while Taiwan and Germany dominated the 3DMark benchmarks. France was strong in SuperPI 1M and wPrime, while Italy missed a top ranking.

However, the gap is very little and Italy made the second or third place in various benchmarks. Since we switched to Core i7 and X58 platforms, the playing field is even. Although all the teams already have some experience with the new platforms, the new platform significantly reduces experience bias between the teams..

Sponsors: MSI, FSP/Fortron, Samsung, Kingston, Logitech

Five global key players supported us in this event, providing the necessary hardware and technical support. MSI provided P45 motherboards, FSP supplied Everest 1010 1,000 W power supplies, Samsung made sure the teams got Spinpoint F 1-TB hard drive and Super Writemaster DVD burners. Kingston added specially screened and hand-selected HyperX DDR3 DIMMs, which are capable of reaching speeds way beyond DDR3-2000. And lastly, Logitech provides G11 gaming keyboards and G5 Laser mice.

We posted a article dedicated to the components used for overdrive: The Hardware Behind Overdrive. However, two details need to be updated as our finals are based on Core i7 platforms rather than the Core 2 Duo. We're now using MSI X58 Eclipse boards and Core i7 Extreme processors to provide the best possible platform for the competitors.

At this time, the competition between all teams are extremely close and well rounded. We're extremely excited to see what the teams can do with Intel's new Core i7. More information to come soon! Good luck to all the teams!

Check out our entire Overdrive coverage on our Overclocking micro site!

More Overclocking With LN2

The German regional of Tom’s Hardware’s Overdrive overclocking competition came to a close late last month (see the news post Overclocking Competition: Germany Edges Past USA 8-2 for a breakdown of the scores).

The three members of the team "Benchbros" were declared the winners after a 10 hour competition. The crew, made up of David Schöppe, Manuel Wiesner, and Andreas Bock, produced record results in eight of the ten categories, achieving a clock speed of 5.9 GHz on the Intel Core 2 Duo E8600 and an MSI P45 motherboard. The winning team prepared for the task even before the event. The consensus was that the trio reached the limitations of its P45-based motherboard and processor, though the hand-selected Kingston memory modules running at DDR3-2000 didn’t even come close to their limits.

The MacSafe team chose an unusual strategy—something we didn’t see in the US—cooling the graphics card via copper sheeting, which was affixed to a self-made pot. This picture story chronicles their challenges in building a custom cooling solution for the competition.

Studying Circuit Diagrams

Jürgen Augstein from MacSafe concentrated hard, trying to make the modifications to the MSI P45D3 Platinum motherboard. The necessary layout documentation was provided beforehand.

Working Under Pressure

Division of labor at the MacSafe team: Jürgen Augstein on the left and Dominikus Baur on the right. The duo employed a very creative approach and constructed most of its mods within a short period of time.

More Work, More Pressure

Everything needed to go smoothly. With only 10 hours to mod and benchmark, there wasn’t much room for error.

Tools Of The Trade

Tools everywhere you look: the Dremel came in handy. The creative MacSafe team had to get in a number of time-intensive shopping errands in order to finalize its build.

Creative Engineering

Built-to-order: here the copper sheet for cooling the graphics card is being cut to size. Copper is an excellent heat conductor and transfers the extreme cold from the CPU pot filled with liquid nitrogen to the graphics chip. The risk of this strategy is that humidity in the air may condense on the components, causing a short.

Attention To Details

Manufacturing the copper cooling sheet took up too much of the MacSafe team’s time, so they only had a few minutes left for running the actual benchmarks.

Just Add Sub-Zero Temps

Nitrogen canisters at the ready: time to get going, assuming the CPU and graphics chip cooler are ready for LN2.

Joined At The Hip

This build is certainly inventive. In this shot we see the CPU pot and copper sheeting, used to draw heat from the graphics processor. Again, we didn’t witness anything like this during our local competition. Linking the two components was certainly an interesting thought, though.

Custom Engineering

The coupling of the processor and the graphics chip for cooling required a lot of work. At the end of the day, there was no time left for successful overclocking attempts. Perhaps this would have gone over more successfully had the competitors had more time for testing.

Here in the States, our overclockers took the GTX 280s apart completely in order to cool them with custom copper pots.

Too Much Prep, Not Enough Benchmarking

Although it’s ready for competition, this custom cooling solution involving copper wrapped around the GeForce GTX 280 simply took too long to manufacture.

All Dressed Up, No Time To Go

Here you see the GTX 280 with its copper attachment, ready to dissipate heat to the much colder custom cooling pot.

When Cold Meets Warm...

Condensation halts progress. The MacSafe team was forced to take breaks when condensation got out of hand. Those break were used for additional shopping trips.

A Tad Too Small

Simple and effective solution á la Tom’s Hardware in 2003: copper piping with high-heat conducting properties and insulation to prevent the formation of condensation. However, the mass of the CPU pot was not sufficiently taken into account.

Don't Short It: Condensation Forms

MSI’s P45 board with condensation forming at the CPU socket: here, the MacSafe team had not paid enough attention to the perfect insulation. A waiting period was required in order for the board to dry out.

Setting The Custom Cooling Aside

Felling the need to warm up: the formation of condensation on the copper connection between the CPU pot and the graphics card prompted the team to take yet another break. Had they continued, condensation could have easily led to a short-circuit—an issue the contestants here in the States battled as well.

Consultations between jury and team

Consultations between the Tom’s Hardware Germany jury and the individual teams: here Jürgen Augstein from the MacSafe team is explaining his team’s strategy for constructing an efficient cooling mechanism using liquid nitrogen.

Atom Just Isn’t For Desktops

The time when the processor and platform markets were simply divided into desktop, server, and mobile hardware segments is long gone. Mobile products and related technologies have entered the enterprise and home markets to increase efficiency. Server technology increasingly meets the demand of enthusiasts. And OEMs are adding desktop components to mobile computers to reduce costs. In short, the same platforms and technologies are adopted for multiple different applications rather than specialized market segments.

However, a new generation of hardware has emerged. Geared toward the mobile and low-cost audiences, these new solutions address the needs of shrinking system dimensions, cost pressures, and ultra-mobile system requirements. VIA’s C7 and Nano processors have been around for a while, but it’s Intel’s Atom that is widely being adopted for low-cost solutions and netbook-type devices, along with nettops. However, while Atom certainly is great for mobile devices with limited performance requirements, we found that its system power consumption can be matched by a mainstream Core 2 Duo system–-which also offers a significant performance boost.

Core 2 Doesn’t Do It All

We don’t want to give you the wrong impression with this article, so it’s important to clearly define the purpose and difference between the Atom and Core 2 processors. Core 2 is Intel’s mainstream flagship product for PCs and notebooks. It is a processor that has to be installed in a socket. Models are available with two or four processing cores with large, medium, or small cache capacities and varying system speeds.

Currently, Core 2 processors cover the full spectrum, from maximum performance to maximum mobility and power efficiency for PC systems. However, Core 2 processors are complex and somewhat power-hungry. Although they are efficient, Extreme models require up to 130 W, which means that substantial cooling is required.

What Atom Is All About

Atom was developed to be cost-effective from the beginning (for Intel and its customers), to deliver a certain minimum performance, and to serve as a low-power part. While Atom is Intel’s smallest processor, the chip giant also touts its very low power specs: the desktop single cores are rated at 4 W, while the mobile versions are rated at only 2 W. The Atom 330 dual core is even rated at only 8 W. All Atom processors have to be soldered onto a motherboard by the manufacturer, which means that they are incapable of supporting any future upgrades.

On the other hand, Atom retails for only $20-$70, which is a quarter to half of the cost of an entry-level Core 2 processor. Unfortunately, Atom isn’t really fast, which makes it clear that it should not be used for systems that have to deal with anything that is more demanding than YouTube video playback, checking email, or browsing Web sites.

Atom Vs. Core 2: The Efficiency Shootout

We thought about this comparison when we looked at two G31 motherboards using a Core 2 Duo E7200 processor, as one of the platforms was capable of consuming only 31 W of power at the plug with the use of an efficient power supply unit. After we saw the 28 W idle power level of the Atom system that we used in the comparison between Atom and VIA’s Nano, we became curious about which CPU was really more efficient.

Efficiency is not measured by the lowest power level a CPU can maintain, but it is instead determined by its performance–per-power consumption ratio. We wanted to know if we could match the Atom’s system idle power consumption to that of a Core 2 Duo system and find out the real efficiency champion.

Atom or Core 2?

Intel Atom 230 (1.6 GHz Single Core)

The Atom 230 processor is a 26 mm², 45 nm, single-core device that utilizes the same front side bus (FSB) as the Core 2 processor family. However, it doesn’t employ Socket 775, but has a 22x22 mm 437-ball micro FCBGA package that Intel decided to pair with the 945GC chipset, which is a 945G chipset with support for FSB1066 and DDR2-667. It is thus limited to FSB800 (Atom utilizes FSB533) and DDR2-533 memory.

Atom supports Intel’s Hyper-Threading technology, offering thread-level parallelism so that one physical processor core serves as two logical cores for software that takes advantage of the feature. Indeed, Hyper-Threading has proven valuable as it boosts system responsiveness and offers slight performance increases. As already mentioned above, the Atom 230 is rated at a thermal design power (TDP) of only 4 W despite its 1.6 GHz clock speed. Intel’s most powerful tool for maintaining this low power level despite the Atom’s 512 KB L2 cache and 47 million transistors is the 45 nm high-k silicon manufacturing process, which ensures low leakage currents.

The platform remains the real issue–while the processor is highly power-optimized, the chipset still is based on the same, three-year old 945 design. This is an advantage when it comes to implementation and compatibility, as 945 still includes UltraATA controllers, for example. But it’s a huge disadvantage when it comes to total system power—a 4 W TDP processor combined with a 22.2 W TDP platform (total TDP rated at 29.5 W) doesn’t match up well.

There also is a dual-core Atom 330 processor available in retail channels that is rated at 8 W TDP at the same 1.6 GHz clock speed. We didn’t have a suitable platform available at the time we finalized this review, but we will follow up on Atom dual-core performance. While we expect it to increase the performance-per-watt benchmarks for a number of desktop applications, it still isn’t even capable of beating a current Celeron processor.

Intel Core 2 Duo E7200 (2.53 GHz Dual Core)

Core 2 Duo was launched July 2006 and was manufacturing using Intel’s 65 nm process at the time. Intel then improved the device’s features and scaled its size down as well with a 45 nm process in February 2008. The first models were called E4000 and E6000, with 2 MB and 4 MB L2 cache, respectively, and clock speeds of up to 3.0 GHz. All current models are called E7000 and E8000, have 3 MB and 6 MB L2 cache, respectively, and offer performance-per-watt improvements. The E8600 top model runs at 3.33 GHz and is the best version for performance and overclocking, but we found the E7000 particularly interesting in terms of performance per watt.

Although they’re based on the same high-end 45 nm process, Core 2 processors carry hundreds of millions of transistors and consequently consume a lot more power than the Atom does–but they also deliver several times more performance. The question is: Can we create a Core 2 Duo system that offers the low system idle power of an Atom 230 system? This would result in a desktop PC with exceptionally low power requirements and performance that is well capable of handling more demanding applications.

Comparison Table

Intel Atom: ECS 945GCT-D

Once again, we used the ECS motherboard 945GCT-D, which is based on the 945GC chipset and the Atom 230 processor at 1.6 GHz. As you can see in the photo, which we took after removing all heat sinks, the Atom processor is the smallest of all the components (right side). The large part in the middle is the 945GC Memory Controller Hub (MCH), while the chip to the left, visibly marked with the name “Intel,” is the southbridge, also known as the ICH (I/O Controller Hub).

MiniITX for Net Tops

The MiniITX form factor is conducive for small form factor PC builds. Any kiosk solution, vending machine, information terminal, or system for medical applications could be based on such a platform. The board offers two Serial ATA ports for hard drives or optical drives, one UltraATA channel for up to two legacy storage devices, one x1 PCI Express slot and one 32-bit PCI slot for expansion cards. There is a HD-capable sound system, two fan headers, two DDR2 DIMM sockets and eight USB 2.0 ports. Four of them are ready to be used on the board’s back panel.

Limitations

The concept does have severe limitations, which are mainly important for desktop users. You cannot install any of the recent graphics cards that require x16 PCI Express. This means that you have to use Intel’s GMA950 graphics engine, which is good enough for an operating system’s graphic interface or SD video, but won’t please any user with specific performance requirements such as HD capabilities or 3D performance. However, the processor probably is the most noticeable limitation, as it cannot be swapped out. You would have to replace the entire platform if it were to fail or become too slow for your requirements. Luckily, such a solution costs less than $90.

Core 2 Duo E7200 and Foxconn G31MG-S

G31 and E7200 is the real low-power story–as we found out recently. The combination of an efficient motherboard with integrated graphics (Foxconn uses only a three-phase voltage regulator) and a reasonable Core 2 Duo E7000 processor based on the latest M0 stepping delivered an amazingly low idle power of 31 W. It is important to know that we had to use a low-power power supply unit, as the models rated for several hundred watts usually are rather inefficient for small loads.

Since the performances of different Intel chipsets does not differ very much if the processors are comparable (clock speeds, timing settings, and power saving options), it doesn’t matter that much whether you purchase a high-end X48 motherboard or an entry-level G31 solution like this one. Computing power depends on the processor and the memory installed. Responsiveness increases with memory capacity and fast drives. And 3D performance depends on how much you’re willing to invest in graphics. We ran the G31MG-S with the E7200 with the same core components as the Atom solution from ECS in order to get a fair comparison for office and multimedia use.

Reason Above All

The motherboard on its own isn’t exciting. In fact, it’s rather boring, as it has no features that would make it really special, but it’s still one of the most reasonable options for those seeking an ideal performance-per-watt ratio. The board only has two DDR2 DIMM sockets, which doesn’t allow for a memory upgrade if you installed two DIMMs, but going for the maximum of 2x 2 GB RAM is extremely inexpensive now, and it will make the system last for at least two years. The x16 PCI Express slot allows users to install a fully featured graphics card, which isn’t possible with most Atom solutions. Two 32-bit PCI slots and a x1 PCI Express slot can accommodate add-on cards. HD audio is there, 4+4 USB 2.0 ports are available, four SATA/300 ports should suffice for most systems, and UltraATA/100 is present as well. However, most people will want more features. Please look at the test setup page for information about the other hardware components we used.

The Power Supply Matters!

System efficiency does not only depend on the core components you use, but it is highly dependent on the power supply, since all power supplies have a certain efficiency when they convert 110 or 220/230 V AC power into 3.3 V, 5 V, and 12 V DC power. Depending on how complex the power supply is, and on how it is laid out (one rail, multiple rails), it will demonstrate certain efficiency characteristics.

As all PSUs have a power loss, efficiency is typically rated in percentage terms. The industry standard for measuring and rating efficiency is called 80 Plus, and it splits power supply units into three categories: bronze, silver, and gold. An 80 Plus rating approaches 90% power efficiency at varying loads, taking low, medium, and high loads into account and specifying this as a percentage of the total output.

Here’s the issue: if a 1,000 W power supply unit is rated excellent at low load, this means that “low load” is still a rather high power in the case of our two systems, which idle at less than 30 W. At such a low load, the excellent 80 Plus, 1,000 W power supply unit will show rather pathetic power efficiency just because it wasn’t designed for such a load.

We decided to run the Foxconn G31 system and the Core 2 Duo E7200 with four different power supplies that are rated at 1,000 W, 850 W, 300 W, and 220 W. You will find the results on the following page, and they should convince you to get a power supply that matches your system’s requirements, as the power requirement differences are larger than the difference in power consumption that you would get by exchanging the E7200 for a Core 2 Duo E8600.

1000 W: OCZ EliteXstream 1000

We used a high-end OCZ power supply to test power consumption for such a hardcore power supply unit. The EliteXstream 1000 is an excellent device. Designed by PC Power and Cooling, it is rated 80 Plus and offers an 82% level of efficiency. An active power factor corrector (PFC) and lots of connectors are mandatory for such a high-end product, which is targeted at workstations and high-end gaming. The device comes with an impressive five-year warranty. However, the device is overkill for our low-power systems, as you will see on the next page.

850 W: Coolermaster Real Power Pro RS850 EMBA

Cooler Master’s 850-W RS850-EMBA has been our reference power supply for almost a year. We use it for many of our reviews, as we appreciate its powerful and reliable qualities. It is also backed by a five-year warranty, is 80 Plus compliant, and definitely well-equipped for an enthusiast user’s requirements.

300 W: Etasis EFN-300

We didn’t have a power supply at 400-500 W available at the time we finished this article, but we found a 300 W model by Etasis, the EFN-300. Etasis Electronics is not one of the big, popular brands, and the EFN-300 isn’t special in any way except that it is a no-noise device rated at 300 W output. While it likely will not impress enthusiasts, the power supply unit helped to decrease the system’s power consumption compared to the Cooler Master and OCZ power units.

220 W: Fortron FSP220

If you remember our recent articles on VIA’s Nano vs. the Intel Atom and the comparison of Foxconn and Gigabyte G31 motherboards, then you might also remember the FSP220 from Fortron, which is a compact 220 W power supply.

This one is rated for an even lower maximum output and some of you might wonder why we even bothered to use it. Clearly, this is the power supply that allowed us to reach the lowest overall power consumption numbers, as it is capable of answering the power demands of the Atom and Core 2 systems at a much higher efficiency level than the other power supply units we tried.

Power Consumption With Different PSUs

Having introduced four entirely different power supply units, it’s now time to look at the power consumption results each of them produced. We used the Foxconn G31MG-S and the Core 2 Duo E7200 for this test. Clearly, the OCZ and Cooler Master power supplies don’t do our low-power systems any service, as they lead to an increase in system power when idle by almost 45%:

While the small FSP220 power supply from Fortron reduces the system idle power of the G31 platform to only 29 W (we achieved this by using only one 2 GB DIM module!), any of the other units would have this value increase to 38 W (with the Etasis) and over 40 W (with the Cooler Master or OCZ).

The results are similar when the processor is running at peak power using the FSP220–system peak power is limited to 64 W. The 300 W power supply unit from Etasis increased the result to 72 W; Cooler Master had this number increase to 78 W; and the most powerful OCZ power supply drove peak power all the way up to 78 W, which is almost 22% higher than the ideal result taken with the Fortron 220 W power supply unit.

The difference is smaller at peak power, as the more powerful power supply units operate at more efficient power levels, which helps to close the gap.

Test Setup and Benchmarks

Please note that we only used a single DDR2 DIMM that ran at DDR2-533 speed with the Atom system and at DDR2-800 for the G31 motherboard. This was the only measure we took to bring the Core 2 Duo’s power consumption from 31 W to only 29 W, which was hence closer to the Atom system’s 28 W. This step makes little sense if you take a performance approach, as the memory bandwidth decreases. But it makes sense considering that many desktop Atom systems don’t even have a second DIMM socket to install more than one memory module.

Further power requirement reductions for our Core 2 Duo E7200 and G31 combo can probably be achieved by optimizing CPU voltage, which would involve “under-voltaging” the processor on suitable motherboards.

Benchmarks and Settings

Application Benchmark Results

As expected, Core 2 Duo is many times faster than the Atom 230 for popular applications:

6.1x more performance for the 2.53 GHz Core 2 Duo in Cinebench R10.

Converting WAV audio into Apple’s AAC audio format for iPod utilization takes more than three times longer with the Atom 230 than it does with the Core 2 Duo E7200.

Converting WAV audio into the MP3 format takes almost four times as long with the Atom 230 as it does with the Core 2 Duo E7200.

Compressing files on WinRAR took almost three times more time with the Atom system.

WinZIP isn’t thread-optimized, which is the reason why the Atom 230 appears to be faster. Still, it takes more than twice as much time to complete this workload.

Synthetic Benchmark Results

Everest

Memory performance seems to be much better with the Core 2 Duo as well, mostly because of the faster and larger L2 cache, but also because of the faster DDR2 memory speeds the Core 2 Duo system offers.

Futuremark PCMark05

PCMark05 says that the Core 2 Duo E7200 offers 3.76x more processing power than the Atom 230 does.

The graphics score is closer to each other and is a bit skewed because the G31’s graphics engine is faster than the one on the Atom 230’s 945GC chipset. In addition, memory speed is faster on the G31 solution, which should also have an impact.

Overall, PCMark05 reports a 2.45x performance benefit of the Core 2 Duo E7200 over the Atom 230.

SYSmark 2004 Results

Depending on which benchmark section you look at, the Core 2 Duo E7200 is between 70% and 300% faster than the Atom 230 in SYSmark 2004 SE. We used this benchmark because the hardware resources are limited on the Atom 230 system. SYSmark 2007 Preview crashed several times, which prompted us to switch to the older version.

Power Consumption

We now look at the total system power consumption. We obtained these results using the efficient 220 W Fortron power supply. Using a power supply unit that is rated for a higher output will automatically increase these numbers by a considerable amount.

The idle power is almost identical. Our Core 2 Duo E7200 system on the Foxconn G31 motherboard utilizes 31 W idle power when running on two DDR2-800 memory modules. Since many Atom solutions only support one memory module, we decided that power optimization achieved by removing one DIMM wasn’t a bad idea. Please look at our article Parallel Processing, Part 2, which compares performances of single- and dual-channel memory configurations.

Peak power consumption is higher than just the difference of the Atom processor’s 4 W TDP because processor tasks will automatically involve other system components. A peak power value of 36 W for the Atom 230 machine is an exceptional result, as it allows power requirements to be calculated more precisely than it does for other systems. The Core 2 Duo E7200 test system went up 29 W to 54 W, which represents an 86% difference. However, if you remember the performance difference between the Atom 230 and the Core 2 Duo E7200 on the last pages, it’s obvious that the Core 2 Duo delivers up to several times more performance at a power requirement increase of 86% maximum. The performance increase is proportionally higher than the power requirement increase in any case.

Efficiency and Performance per Watt

We calculated the average power requirements of the two systems during an entire SYSmark 2004 SE run. This reflects intensive use of the systems, using popular mainstream office and content creation applications. Surprisingly, the Core 2 Duo E7200 system required only slightly more power on average: 34 W vs. 30 W for the Atom 230.

We also tracked the total power both systems used to complete the SYSmark 2004 SE run measured in watt-hours. This metric gauges system performance per watt hour. Thanks to its faster performance, the Core 2 Duo E7200 system on the Foxconn G31 motherboard required 40 Wh to complete the entire benchmark, while the Atom machine, which is substantially slower, required 46 Wh and a much longer period of time (see diagram on the bottom of this page) to complete the workload. The Core 2 machine may require more power at peak loads, but less power for the specific workload. Expect the result to be more significant in favor of the Core 2 for processor-intensive workloads.

Finally, we calculated the SYSmark 2004 SE score per watt hour. The Atom 230 reached 2.2 SYSmark 2004 SE points per watt hour, while the Core 2 Duo machine did more than twice that: 5.65 points per watt hour. From an efficiency standpoint, the Core 2 Duo E7200 is vastly superior.

The diagram shows the system power consumption over the entire duration of the benchmark run, and it also shows how long the systems took to complete the workload. The Atom 230 took more than one hour and 30 minutes, while the Core 2 Duo system required one hour and 10 minutes. The Atom system usually didn’t exceed a 36 W maximum power requirement, but it continued to use power while the Core 2 system was finished. The power peaks of the Core 2 system were substantial, but they were exactly what you read: peaks. The 50+ W power consumption applies to only a fraction of the benchmark runtime.

Conclusion

The results of our comparison between Intel’s Atom 230 and a Core 2 Duo E7200 with an efficient G31 platform were interesting, to say the least. We learned several things, which should be taken into consideration before purchasing a low-cost or low-power computer, as these things have a larger impact than the performance difference between the two machines. Let’s sum up the findings:

Beware Of Over-Doing It On Your Power Supply!

Purchasing an oversized power supply unit may look attractive as an investment for the future, but it is an issue if you want your system to be power efficient. Matching your power supply’s output to the real power requirement of the target system will make sure that the power supply operates at a load level at which it is truly efficient. If your power supply is extremely powerful, but operates at extremely low loads, then it will operate at very low efficiency and have the system require much more power than necessary. Selecting an appropriate power supply is the key, as it will help to reduce power waste much more than going for more efficient components!

Efficiency And Performance Per Watt

Atom 230 is extremely power-friendly, it’s tiny, and it’s very affordable. This is important for mobile devices, where requirements are limited. But Atom is not an efficient product if efficiency is defined by performance per watt or power. We found that the Core 2 Duo, in our case the model E7200, can be operated at an equally low idle power as the Atom 230 (less than 30 W total), while still delivering between 50% and up to several times more performance if you need it. The result is a performance-per-watt ratio that is substantially better for the Core 2 Duo.

Atom or Core 2?

Let’s get back to the key issue: does Atom make any sense for desktop-like PCs? It does, but only if you have to be very conservative on cost, and only if your requirements are clearly defined. Atom-based nettops or budget PC solutions are very affordable, but they don’t perform really well and you hardly have any upgrade possibilities. If you can afford spending $100 more for the motherboard and a decent processor, you’ll get a solution that offers similar idle power, but much better performance and efficiency, along with upgrade options.

Overdrive Overclocking Championship 2008: The Components

Our 2008 overclocking event, which we named Overdrive, began last weekend in Los Angeles, California. Four more countries with a Tom’s Hardware presence—France, Germany, Italy and Taiwan—will be hosting trials with some of the finest professional overclocking teams to determine local winners. We will fly the best overclockers to Paris, France, where they will compete for the Tom’s Hardware Overdrive world championship.

We wanted to introduce you to the hardware component that we’re using for the Overdrive Overclocking Championship. Note that most of what we’re using is being provided by our sponsors.

All Components Are Provided To The Competitors

In an effort to provide the same conditions at all event locations worldwide, we decided that we wanted global sponsors for as many key components as possible. And we succeeded: MSI is on board to provide overclocking motherboards based on the P45 and X58 chipsets; Fortron provides powerful Everest 1,000 W power supplies; Samsung sponsors us with Spinpoint F 1 TB Serial ATA hard drives; Kingston provides high-end DDR3-2000+ memory modules; and Logitech contributes its gaming-class keyboards and mice.

Core 2 Duo First, Core i7 Extreme In The Finals

Since Core i7 availability could not be guaranteed at the time we began our championship, we had to rely on Core 2 Duo processors and P45 platforms for the worldwide trials. The winning teams, which will battle each other in the first-ever Core i7 (Nehalem/Bloomfield) overclocking competitions, will receive Core i7 Extreme processors.

Let’s look at all of the components in greater detail.

CPUs: Core 2 Duo E8600 and Core i7 Extreme 965

We would have loved to use Intel’s latest Core i7 processors and X58 motherboards right from the start in order to provide all registered overclockers the opportunity to play with Intel’s new processor. However, since processor availability could not be guaranteed in all countries in sufficient quantities, we decided to stick with Core 2 Duo for the trials. This provides a proven and mature platform that most overclockers should be very familiar with.

Core 2 Duo E8600: 5+ GHz

We selected the fastest Core 2 Duo E8600 processor for our overclocking ventures, as this is the model that allows us to set the highest possible clock multiplier. As the processor clock speed is composed by multiplying the system base clock with the multiplier, a higher multiplier introduces more selection choosing settings that’ll yield faster overclocks. Intel’s Extreme Edition processors typically are the best choice for extreme overclocking, as they can be set to almost any multiplier. However, Intel stopped releasing dual core Extreme processors in favor of quad-core models.

Looking at the overclocking capabilities of many P45 platforms, such as the P45D3 Platinum provided by MSI, we can assume minimum system clock speeds of 500 MHz. This results in minimum overclocking speeds of 5 GHz when multiplying the system clock speed of 500 MHz with the E8600’s multiplier of x10. The further the overclockers go with the motherboards, and the better their skills in overclocking with liquid nitrogen, the higher the clock speed they will be able to reach—there are reports on the Internet of 6 GHz clock speeds achieved on a Core 2 Duo.

The Core 2 Duo E8600 is based on Intel’s 45 nm manufacturing process, and runs at a default core clock speed of 3.33 GHz using Intel’s 333 MHz front side bus in quad-pumped mode (FSB1333). We found that most of these processors reach 4 GHz with little overclocking effort; Core 2 Duo is still the best performing dual core processor, and also reaches the highest performance per watt ratings.

Also see:

• Wolfdale Shrinks Transistors, Grows Core 2

Core i7 Extreme 965

Intel’s next-generation processor is named the Core i7. It officially launched on November 17, but reviews were posted starting on November 3. While Core i7 keeps many architecture details of the Core 2 family, it also introduces a plethora of changes:

• Four cores per processor die rather than dual cores that are paired to create a Core 2 Quad with four cores.
• 8 MB shared L3 cache in addition to 256 KB L2 cache per core.
• Integrated tri-channel DDR3 memory controller. Memory speed depends on available memory multipliers, which differ among Core i7 models.
• Serial QuickPath Interconnect (QPI) replaces the front side bus. The interface can also be used to connect multiple Core i7 processors, which means that motherboards with two sockets can be created more easily.
• Reintroduction of Hyper-Threading, which provides two virtual processing cores per physical processing core.

Core i7 requires new motherboards for Socket LGA1366, due to the higher pin count required by the memory controller. All initial Core i7 models—the 920, 940 and 965 Extreme Edition—are rated at a thermal design power of 130 W. Core i7 is a new processor design that is based on proven manufacturing technology, but it is still difficult to predict its initial overclocking margins. The engineering samples that we received for the Core i7 review could be overclocked up to 3.8 GHz on air cooling. We’ll see how far the first Core i7 silicon can go during our Overdrive finals in France on December 13 and 14.

Also see:

• Core i7: Blazing Fast. X58 Chipset with ICH10
• Core i7: 4-Way CrossFire, 3-way SLI, Paradise?

Platforms: P45 and X58

MSI P45D3 Platinum

MSI is the official platform sponsor for the Overdrive Overclocking Championship and contributes two different platforms: a P45 motherboard for our trials, where we have the teams overclock Core 2 Duo processors, and X58 boards for the Core i7 overclocking in the grand finals. The P45 board is MSI’s P45D3 Platinum, which is a DDR3 model that comes with all top-notch features, such as electrolytic capacitors, the CircuPipe 2 heat pipe solution for component cooling, twin-phase north bridge and memory voltage supply, and a plethora of BIOS overclocking features. Although other vendors are often considered ahead of the curve when it comes to hardcore overclocking, MSI has improved a lot in recent years, and created a powerful basis that will serve our events well.

Check out our complete gallery of MSI P45D3 photos.

Also See:

• 11-Way P45 Motherboard Shootout
• MSI Overclocking Challenge: France Finals
• Intel P45 Chipset: Eagle Lake Has Landed

MSI X58 Eclipse

The local winners of our US, France, Germany, Italy, and Taiwan events will meet in Paris, France, where they will compete on Core i7 systems using MSI’s X58 Eclipse motherboard. This is MSI’s Socket LGA1366 flagship; it uses a flexible six-phase voltage regulator for the CPU, and is ready for AMD CrossFire and Nvidia SLI using the latest drivers from both vendors.

The boards have six memory sockets for DDR3 DIMMs, fully utilizing the triple-channel capabilities of Core i7. Our overclockers may run the systems with three or with only two DIMMs—that’s up to them. All that matters is maximum performance.

While our first sample was equipped with traditional chipset coolers, the retail version utilizes a solid heat pipe that cools the north bridge and south bridge, and a second copper heat sink for the voltage regulators. That’s the version you’ll see in the picture to the right.

Check out our complete gallery of MSI Eclipse X58 photos.

Also see:

• MSI’s X58 Motherboard to Have Quad SLI
• Sneak Peek: Four Core i7 X58 Motherboards

RAM: DDR3-2000+ by Kingston

Robust memory is the next critical item for successful overclocking. Kingston has been pretty active recently, announcing new high density heat spreaders for the new HyperX T1 series, which is available in both dual and tri-channel memory kits. DDR2 kits reach 800 and 1066 speeds, while the DDR3 kits go all the way up to DDR3-2000 speeds.

Hand-Selected Memory

When we discussed the sponsorship, we wanted to be sure that our overclocking teams would have access to the fastest memory possible, to avoid any memory bottlenecks and assist in reaching maximum performance levels. To assist us, Kingston went into its labs and hand-picked DDR3 HyperX DIMMs that are capable of running well above DDR3-2000 speeds at increased voltage.

All overclocking teams receive a total of four DDR3 DIMMs, which allows them to run three DIMMs in a tri-channel configuration on X58 and Core i7 at our finals, or different dual channel combinations on the P45D3 Platinum motherboards at the trials in the US, France, Germany, Italy and Taiwan. The HyperX DIMMs can be operated at voltages of 1.8V to 1.9 V when running at low memory speeds at JEDEC timings. At our Overdrive events, however, we expect the teams to push the memory to ridiculous DDR3 speeds to support maximum performance. This will require the voltage to be adjusted to 2.3V to 2.5 V. Although we don’t recommend doing this for your everyday PC, it is an important step in reaching faster clock speeds.

Also see:

• Kingston Launches HyperX Tri-Channel DDR3 Kits
• High-End DDR3 Memory on the Hook
• Mainstream-Ready? DDR3-1600 Shootout
• DDR3-1333 Speed and Latency Shootout
• Vista Workshop: More RAM, More Speed?

Hard Drives: Samsung Spinpoint F

This is the first time we’ve worked with Samsung on an event, but the storage and memory specialist wanted to be sure they supported the Overdrive Overclocking Championship as much as possible. We received Samsung Spinpoint F 1 TB 3.5” SATA/300 hard drives, which provide more than enough capacity for the overclocking systems.

Technically, 20 GB to 30 GB of capacity would be sufficient for the event, but recent price drops make the terabyte hard drives more attractive than ever. Samsung’s Spinpoint F is still one of the fastest 3.5” desktop hard drives when it comes to throughput. Although newer drives like WD’s RAID Edition 3 provide quicker access time and better I/O performance, the 7,200 RPM Spinpoint F remains one of the most balanced hard drive choices for enthusiasts.

Also see:

• Samsung Ships Three Platter Terabyte Drive
• Three Samsung Terabyte Drives: Mainstream, Business or Eco?

Super Writemaster SH-S222/223 Optical Drives

Samsung also provided us with SH-S222/223 DVD burners for our event. Although the optical drive doesn’t directly impact reaching high performance on overclocked systems, our contestants still need optical drives to install Windows Vista, and to deploy drivers and benchmarks to the test systems. The SH-S222 family utilizes the UltraATA interface and provides up to 22x DVD recording speed on DVD+R and DVD-R media, and up to 16x DVD recording speed on DVD+R double layer recordables. The 223 family is similar, but utilizes the SATA/150 interface.

Power Supply: Fortron Everest 1010

While it is very important to select a power supply that actually matches your power requirements—or even exceeds them a bit—overclocking is all about maximum performance. When overclocked and overvoltaged, processors, graphics chips, and memory all face increased resistance and increased power consumption, which has to be provided reliably. An overclocking record that fails due to insufficient power would really be a disappointment.

FSP Everest 1010

Fortron is not yet as popular as many other power supply brands, but it is a real power supply manufacturer—most of the brands you may think of as supply makers don’t actually manufacture the units themselves, rather acting as vendors. Premium PSU companies have their own designers and build-to-order their units with one of the big manufacturing companies, but only some—like Fortron—control the entire product design and manufacturing process. The Overdrive sponsorship by Fortron includes Everest 1010 modular power supplies, which provide a nominal power output of up to 1,000 W (1,010 W, to be exact).

The product is compliant with the ATX12V 2.2 and EPS 12V 2.91 standards, is CrossFire and SLI ready, and comes with an active PFC to reach its maximum energy efficiency of 85%. A 120 mm fan takes care of ventilation for the four channel 12 V rails. The Everest 1010’s modular layout allows users to attach two floppy connectors, six SATA power cables, six Molex connectors, two PCI Express graphics auxiliary power plugs, plus the main 20+4 pin power connectors.

Check out our complete gallery of Fortron Everest 1010 photos.

Also see:

• Silent PSUs: Fortron vs. Silverstone
• Power Supply Roundup, Part II

Logitech G11 and G5 Keyboards and Mice

Logitech G11 Gaming Keyboard

Logitech is one of the most popular providers of input devices. The firm wanted to be part of our Overdrive Overclocking Championship as well, and provided G11 Gaming Keyboards for our contest. The G11 is an gaming-class keyboard that comes with a blue character backlight to facilitate keyboard control in dark environments. You can set three different brightness levels, and it offers additional media control keys to manage audio and video features.

Gamers will probably love the additional so-called “G keys:” Logitech added 18 programmable keys on the left side of the keyboard. Custom macros can be assigned to each of the keys, and you can assign up to three different functions per key. Logitech says that this can be done on the fly, even when you’re in the middle of a game.

The keyboard clearly benefits from Logitech’s many years of experience, as it has a smooth and nice design, and cord channels on the bottom to route cables. One of the few downsides of the product is its dimensions, as it is considerably wider and deeper than a regular keyboard that lacks the additional features.

Check out our complete gallery of Logitech G11 Keyboard photos.

Logitech G5 Laser Mouse

Finally, we also need a mouse to operate the overclocking systems, and since Logitech is an ideal partner, the firm also provided G5 Laser Mice for our events. This isn’t the company’s top product, but it is one of the most precise and customizable wired mice you can get. It is only suitable for the right hand despite various customization options, though—if you need a mouse that works with the left hand, you should look for one of the V series models.

The G5 Laser Mouse offers very high precision, and allows setting the sensitivity to 2,000, 800 or 400 dpi without any drivers. Logitech says that its image processor manages up to 6.4 megapixels per second. We also liked the weight cartridge: since different people like to have a lighter or heavier mouse, it is possible to customize the mouse weight by adding up to 36 g of weight into the weight cartridge.

Check out our complete gallery of Logitech G5 Mouse photos.

Operating System & Conclusion

Windows Vista Ultimate

Although Windows XP is still considered the one and only operating system by many overclockers, we had to go with Windows Vista Ultimate. After talking to Microsoft during our planning stage, we found that Microsoft does not want to promote Windows XP any more, despite continuing support for several years. Instead, we received copies of Windows Vista Ultimate. Sooner or later Windows XP will disappear, and at some point it will be necessary to make a switch.

All of our benchmark software is fully compatible with Windows Vista, although we had to get a patch for Aquamark 3 (Direcpll_V4.10.1.92). You might be able to wait for Windows 7, but since the similarities to Vista will be significant, making the step and giving Vista a try doesn’t hurt. Ultimate is the most complete edition of Windows Vista, and also the ultimate in cost, as it retails for $319.95. The Ultimate edition comes with Media Center, which is part of the Home Premium version, adds BitLocker drive encryption (not available in other editions), and it has all the features of the Business editions as well.

Hosting The Games

It’s been incredibly busy around our Culver City office for the past month or so, while preparing for our first-ever Overdrive overclocking competition. For more information on the contest itself, please check out Overclocking Goes Int’l.: Overdrive In The USA. If you’d like to see some of the pictures we were posting semi-live as the event transpired, you can check those albums out here (Tom’s OC Competition: Mid-day Pics ) and here (Tom’s OC Competition: LN2 Craziness ). We also posted a video of some of the event’s highlights. Check out VIDEO: Tom’s Overclocking Competition.

But the purpose of this piece is to give you a play-by-play of how things went down, to introduce you to our winners and, in the conclusion, present you with a video interview of the guys who’ll represent the United States in our international finals next month. We’re used to seeing world records being broken by fiercely competitive teams from Greece, Taiwan, and other countries hosting serious overclocking talent. Show your support for Team USA, though—they’re going to be working with high-end Core i7 equipment. The potential to set brand new world records will be there, and we hope to take home the competition’s grand prize.

The list of contestants who flew and drove in to compete looked a little something like this:

All 12 of our overclockers proved to be not only seasoned veterans of hardware tweaking, but also true gentlemen and outright good sports. We saw plenty of help and advice flowing between the four teams, even as the group kept its collective eye on our score board, which was kept updated in real-time. A sincere thank you to all 12 for sharing the weekend with us.

Day 1: An Early Start, Unexpected Heat, And Government Intervention

We started our first day off bright and early, setting up the outdoor venue at 6:45 AM. At 8:00 AM, the hardware was handed out and each team set to work prepping the gear. If you’ve never overclocked competitively, you’d be amazed at the work that goes into readying a complete platform for two days worth of abuse. Each team approached the process differently, which was surprising, given that they were all ultimately aiming for the same bull’s eye.

The gentlemen from Team 5G, for instance, meticulously put together their platform, first coating the MSI P45-based motherboard with liquid electric tape to protect every surface against the damaging effects of water condensation. The group disassembled its GeForce GTX 280 and covered it completely in insulation while mounting a custom copper pot for LN2. The preparation process proved to be quite in-depth—so much so that by the end of the first day, Team 5G only had one score on our board. Nevertheless, the prep work proved invaluable, as even the worst condensation didn’t faze the board, processor, or graphics card. As a result of the upfront effort, 5G didn’t run into problems for the duration of the contest.

Day One: Continued

In sharp contrast, Team Ironmods rushed right into tuning its hardware and had an entire board’s worth of baseline scores posted before the end of the first day of competition. The group did do some prep work, laying down foam, covering sensitive areas with petroleum jelly, and packing on the paper towels. But they did just enough to keep their hardware safe, dedicating valuable time to benching instead.

Team OCForums took a similar approach and ended up trading blows with Ironmods for most of the competition. The trio was quickly up and running, generating benchmarks and taking screenshots that’d ultimately pave the way for a nail-biting finish on the contest’s second day.

An unfortunate string of hardware troubles prevented 411 Overkill Extreme from hitting clocks much higher than 4 GHz, which kept them from generating baseline numbers that they knew wouldn’t be competitive. The group’s first motherboard just didn’t have the headroom they needed. So, after fighting it for most of the first day, a second board was provided, which turned out to be missing a pin inside of the processor socket. The challenge would have been significant enough had a third board performed as expected, but even then, frequencies above 4 GHz just weren’t in the cards.

Before we were able to make too much progress with actual benchmarking, we received a visit from the friendly Culver City Fire Department to check in on the event (more detail on that in this video). Once they signed off on our setup, the game was on. We wheeled out the dewars full of liquid nitrogen and the teams started cooling their systems, looking for the cold bugs that’d lock up each of their configurations. For reference, there is such a thing as getting these components too cold—go past a certain point and everything just locks up.

Lunch Break

Knowing that several of our guests had never partaken in the joy that is In and Out, our Tom’s Guide managing editor Rachel Rosmarin and fearless CEO Antoine Boulin made a dash to the west coast chain for Double Doubles and fries—mostly Animal Style, of course.

Lunch didn’t stop the four teams from battling on. In between bites they continued tuning and tweaking, searching for the right combination of settings that’d allow each benchmark to complete successfully. One interesting piece of information that the teams shared was that because each benchmark taxes a different part of each component, overclocking competitively didn’t mean finding the one combination of settings that’d do the trick. Rather, for each test, the teams needed to start over and find the very edge before backing down in search of a speed stable enough to benchmark. Suddenly, the task at hand sounded all the more time-sensitive given just two days of testing.

Wrapping Up Day One

We were originally scheduled to wrap up around 5:00 PM, leaving plenty of time for the contestants to get back to their hotel rooms, find food, and relax before the next day of competition. It quickly became clear, though, that our group of 12 would rather tune their hardware throughout the night than succumb to sleep. “Fine,” we said. “Let the games continue until the first overclocker drops.”

A stack of 10 pizzas came and went. We made another run for Red Bull, and a guest on the scene obliged the overclocker’s requests for Coronas and made a beer run. Supposedly, that’s the stuff that makes higher frequencies possible. At around 10:00 PM, the group had had enough, and we officially called the first day of competition. Two teams had posted a full score board’s worth of results, and two teams shared a single score between them. Though it might have seemed easy to predict a winner at that point, the second day was full of surprises that saw the scales tip three different ways.

Day One Ends, Day Two Begins

We picked everyone up from the hotel at 8:30 AM and resumed the competition spot on at 9:00 AM. Because the event took place outside, we had had to clean up the night before, chain all of our LN2 dewars together in the parking lot, and deploy all of the gear first thing in the morning.

With half of a score board’s worth of results in the bag, we started day two with even more strategy. Team 5G was now firing on all cylinders and the results started flying from the team’s 5 GHz+ Core 2 Duo and nitrogen-cooled GeForce GTX 280. In fact, 5G even held a few first-place results for a time.

But because Ironmods had established solid baselines and was able to spend its second day fine tuning, every top score seemed to fall within hours of being posted. The strategy was simple, but it worked: post scores for everything on Day 1, focusing on processor benchmarks and air-cooled GPU tests. Shift full attention to the volt-modded GTX 280 on day two and really crank through improved 3D numbers.

The Ironmods guys were getting it from both sides, since the OCForums crew had great luck with its motherboard/processor combination. Instead of cold-bugging at -80 or -90 degrees Celsius, like the other guys, its platform was hitting -115 and -120 degrees before locking up. The result was strong WPrime, Super PI, and pifast numbers. Like Ironmods, OCForums saved its GPU testing for Day 2. Very quickly, the team started improving on the results generated the day before.

As Ironmods started in on its LN2-cooled 3D scores, the competition really got heated—or should we say super-cooled? Using liquid nitrogen and a custom-manufactured copper pot, the guys were able to get their graphics card down to -186 degrees Celsius, facilitating some great clocks. The only frustration was that they couldn’t apply extra voltage. No matter what they tried, turning up the juice on our GTX 280s would cause a brown-out, attributed to insufficient power on the PSU’s +12V rail. We’ll work with our guys to solve this in time for the finals next month.

By mid-day, and after a rush on McDonald’s, Team 5G was fully competitive on the board, OCForums had a solid second place position, and Ironmods was having trouble keeping its platform running. The weekend’s stress was seemingly too much. Admitting the motherboard was fried and grabbing a second to replace it would have cost a 5% penalty, so the group held onto its existing scores, believing them to be ample.

As we approached 5:00 PM, the official end of the competition, 411 Overkill had long since thrown in the towel and 5G was ready to call it a day, but OCForums knew that if it could beat just two of Ironmods’ CPU-based results, it’d clinch the trip to France. The group spent the final five minutes edging closer to Ironmods’ SuperPI 1M score with everyone crowded around its bench.

And The Winner Is…

With two days of intense competition behind us (in addition to 25 In-and-Out Double-Doubles animal style, 25 Big Macs, fries from both esteemed establishments, 10 pizzas, two sit-down dinners, and cases of Red Bull), we were ready to tally up the scores from each of our tests. Not surprisingly to anyone who had been doing the math in their head, the overwhelming victor turned out to be Team Ironmods.

We had a great time hosting the competition, learning about the ins and outs of working with (and drinking) LN2—seriously, don’t try it at home. We have a better idea of how hardware can excel in an air-cooled environment, then crash and burn in an extreme situation. And we now more thoroughly appreciate why someone would want to take a perfectly-functional $500 GPU or $1,000 CPU and tweak the heck out of it for the sake of competition. We’ve seen overclocking at its very limits—and we can’t wait to send off Team USA to the finals in Paris next month!

Check out Chris Angelini’s LN2-filled interview with the winners. Overdrive Overclocking Competition U.S. Winners.

Scores, Screenshots, And Points

Each team’s scores are are as follows:

Winning screenshots—preserved for the sake of completeness:

An Introduction To Our Championship

As you no doubt have already seen, we were quite busy last weekend hosting the first leg of an international overclocking competition that’ll see the overclocking elite from five different countries battle it out for the chance to meet in Paris, France. But before hitting you with the results of this last weekend’s festivities, we wanted to present a little introduction to the contest, the folks who’ll be participating around the world, what sort of hardware they’ll be tweaking, and the prize for which they’re fighting.

The idea for doing an overclocking event sprang up at the beginning of this year—probably even before other organizers decided they wanted to do something similar. Gigabyte and MSI have been holding their own overclocking events and other vendors partnered with media outlets to get events started in recent months. Great, we say. More enthusiasts interested in maximizing their hardware means a larger audience to keep Intel, AMD, Nvidia, all of the motherboard manufacturers, and memory vendors on top of their game, continuously innovating.

However, we wanted to do something truly unique. We were hoping to have our participants be the first to do battle on the new X58 platform and Core i7 Extreme 965 processors from Intel. We decided that we wanted to do this in five countries, and to fly the winning teams to our brand new Tom’s Hardware and Best of Media headquarters in Paris, France. And here we are today: The Overdrive Overclocking Championship is already in full swing.

The premise sounds easy for a bunch of self-proclaimed hardware enthusiasts—find overclockers, give them components plus liquid nitrogen, and watch them do their thing. But the international coordination proved difficult indeed. The Core i7 processor is still not officially available, even though you can already read about its overclocking and gaming performance.

Unfortunately, Intel couldn’t make samples ready in time, so we adjusted the preliminaries to include Core 2 Duo E8600s instead—fine, since most of these guys are already season veterans with the current Core 2-based platforms. When the finals happen next month, we’re hoping to see some world records fall as the globe’s most talented overclockers tweak the world’s fastest desktop architecture.

Full Throttle And In Good Hands

It took several months until we had all the components for the event lined up. Obviously, we had to find sponsors were willing and able to provide the product quantities to execute such an event in five countries plus the grand finals. Then our localized teams in the United States, in France, in Germany, in Italy and in Taiwan went out and searched for the best overclocking teams.

To make the event even more attractive, we not only decided to provide the first opportunity of overclocking Core i7, but we also wanted this to be free-of-charge for the participants. Travel, accommodation, food, and drinks (lots of drinks, as the US event demonstrated) are taken care of.

Who Will Win?

This isn’t only about overclocking Core i7. It’s a battle between some of the finest overclocking teams from all over the world. The teams have to bring all their custom gear and they have to utilize all of their experience to win the trials in the US, in France, in Germany, in Italy and in Taiwan on the popular LGA775-based Core 2 Duo platform. The winners will then have to push themselves even further on a brand new architecture to become the Overdrive World Champions.

In our initial photographic coverage of the event here in the United States, we saw readers wondering “what’s the point of using LN2 on an open bench for one day’s worth of testing? That’s not representative of usable overclocking” We’ll talk about the idea there next.

What is Overclocking?

Most of our readers are very familiar with the term overclocking. The word defines itself, really. Integrated circuits are clocked above the sanctioned specifications in an effort to generate more performance. Many PC components can be overclocked. First and foremost, there is the processor. But you can also tweak the graphics card, the motherboard chipset, and system memory. Typically, clock speeds of some of these components are related, meaning that you may not be able to overclock one component without overclocking the other.

How to Start

The first stop for overclockers, after having purchased components that are suitable for overclocking, is the motherboard BIOS. For more information, please look into our Overclocking Guide, Part 1: Risk, Choices And Benefits. We’ll caution you right now—if you’ve never overclocked before, do lots of research before taking the plunge. This is a piece of advice we heard over and over from our guests. There are a plethora of online resources: sites like ocforums.com and XtremeSystems are two of the most revered. Of course, you can’t forget our own Overclocking Forums, either.

Where to Go?

It isn’t difficult to reach the overclocking limits of many mainstream components, which means that the overclocked component will become unstable, leaving you with a flaky hardware configuration. Early on in the game, you probably haven’t done any damage to your hardware and in 99% of all cases you can reset your system, go back to the BIOS and restore the original settings.

The taste of free performance can be addicting though, and if you want to milk even more speed from your hardware then you have to start accepting a bit more risk. Here’s where it helps to have a more intimate knowledge of your hardware (and hopefully, this is an area where our analysis can help improve your odds). If you’re familiar with integrated circuits, you know that they all need a certain voltage to operate and they show a certain power loss, which increases with clock speed and voltage. Increasing the operating voltage almost always gives you better clock speed margins, meaning that components may run at faster clock speeds reliably when increasing the voltage.

The downside is elevated heat dissipation, which will invariably stop you in your tracks once you start serious overclocking attempts. At this point you will want more efficient cooling solutions for your components.

Why a Contest?

Here’s where we answer that question of “why overclock to an extreme beyond daily use?” Let’s say you enjoy playing basketball or tennis or soccer. But you’d be no match against a Kobe Bryant or a Roger Federer or a Cristisno Ronaldo. Those guys play at the most competitive degree possible, taking their respective sports to the next level. Well, if you want to take overclocking to the next level, it involves more than $50 copper heatsinks, ball-bearing fans, and five years of reliability testing. To really show off the talent of the experienced overclocker, you have to get much more extreme in technique and equipment.

Enter the realm of soldering irons, volt-modding, cold bugging, liquid nitrogen, and components that only last a day before bowing under the extraneous conditions of high voltage and extreme cold. When you’re overclocking at the bleeding edge, you design your own custom pots to maximize heat transfer according to the processor you’re tuning. You show up to competitions with a bevy of multimeters and thermocouples. You handle liquid nitrogen with bare hands and flip-flops—not because you don’t know how dangerous this stuff can be, but because you’re prepared for a full 18 hours of pouring, so comfort is key.

Yeah, it’s hard to appreciate the prep work and extreme duress that goes into elite overclocking by looking at a couple albums full of pictures, but spend two days around these guys and their lust for faster benchmark speeds becomes contagious and you can suddenly see why they spend their own money buying hardware doomed to a life of flogging under custom cooling gear.

Thus, an overclocking contest makes good sense. This stuff has become an art and a science of its own. It’s amazing to watch a hardware overclocker take the Vcore on a P45 board and E8600 to 1.9V (stock is up to 1.3625) and then complain that he doesn’t have enough voltage headroom and would really like to have 2.1V and 2.2V settings available. Other overclockers custom-build their cooling devices to achieve a better cooling effect. There are dozens of tools and multiple manufacturer utilities to assist overclocking, and utilizing very unconventional cooling techniques such as compressor cooling, dry ice, and liquid nitrogen help to further push the limits away.

We decided to gather some of these hardcore overclockers, because they all collected tremendous experience when it comes to operating mainstream hardware. We will provide an insight as to how they work, and we’ll provide them the chance to battle each other using Intel’s latest Core i7 processors at our finals in December.

Of course, the event wouldn’t be complete without a prize. So, in addition to giving our overclockers the hardware the so mercilessly beat down, they’re also in the running for a $5,000 grand prize, which will go to the winner of our finals next month.

The Championship

We will hold the contest at the following dates and locations:

In most cases the participants will arrive one day early, so they have two full days to prepare and to overclock the living daylights out of our components.

Due to limited resources, we cannot make these events open to the public, but we welcome all media and industry people who would like to attend. If you’d like to attend one of the international events, please email pschmid [at] bestofmedia.com. We are happy to accommodate all requests in the order they arrive. We had a number of friends and family attend our US trials and everyone seemed to have a very good time.

Participants

Although the full list of participants has not been fixed for all countries, here are the teams that were selected for three of the competing countries:

France

• Westyle & Jmax Hardware
• Marmott & Florian76
• Thor941 & CtrlFix
• boblemagnifique & redratamd
• **Drazendead & Joker_V3

Italy

• Memory Extreme Team Italy
• Shark OC Team
• Sfera OC Team
• NextHardware Overclocking Team
• Xtremehardware-OC-Team

USA

• Team IRONMODS
• Team 5G
• OCForums
• 411 Overkill Extreme Overclocking

Germany and Taiwan

Well be following up with more detailed information about our German and Taiwanese competitors soon.

Sponsors and Components

We found a number of premium hardware manufacturers that were willing to officially sponsor our Overdrive event:

• MSI is the main partner, providing P45 and X58 high-end motherboards.
• FSP/Fortron contributed Everest 1010 high-end power supplies.
• Samsung provided Spinpoint F 1-TB hard drives and optical drives.

Apart from these, we have two more sponsors:

• Kingston provides DDR3-2000+ HyperX DIMMs for all countries.
• Logitech participates with keyboards and mice.

We’ll publish an article on the components shortly.

Software, Benchmarks and Scoring

Operating System

The software question was a hard one to answer. Although Windows XP is the de-facto standard for overclocking systems, we could not get Microsoft to support us with licenses for this event. The reason is obvious: Though Windows XP will still be supported for several years, Microsoft does not want to promote it any longer. Hence the operating system choice had to be Windows Vista Ultimate. After a lot of back and forth (and some surprise discoveries that the 64-bit edition was faster in many overclocked benchmarks) we went with the more tried and true x32 environment.

Benchmarks

We decided to put emphasis both on CPU overclocking as well as on graphics processor overclocking. All devices may be cooled by any means considered necessary, and we expect the teams to utilize our liquid nitrogen for both components.

CPU Benchmarks

There are three individual benchmarks, which we run at various settings:

• SuperPI 1.5 1M
• SuperPI 1.5 32M
• PiFast 4.1
• wPrime 1.55 32M
• wPrime 1.55 1024M

Graphics Benchmarks

• Aquamark V3
• 3DMark01 SE V.330
• 3DMark03 V3.6
• 3DMark05 V1.3
• 3Dmark06 V1.1

All of the graphics tests were expected to be run at their default settings, though the contestants were allowed to pick whichever driver settings they wanted in order to achieve the best possible numbers.

Scoring

The team reaching the best benchmark results in each test wins. For instance, reaching the best result in 3DMark06 will result in a score of 100 points. The second place receives 60 points, others receive a smaller amount.

Should two teams reach the same score in a particular benchmark, then they will both receive the higher score (e.g. 100 points for 1st place, leaving the benchmark without a 2nd place).

We also decided to introduce a 5% score penalty should a team fry one of the components. There will be sufficient spare parts, but we cannot provide multiple spares.

Surprisingly, we didn’t lose any hardware in the US trial—likely a result of not being able to push voltages high enough on either the P45 motherboards or the GeForce GTX 280 graphics cards. We’re expecting the finals to heat up considerably—pardon the pun—with access to more aggressive settings.

Software Tweaks

We decided to allow tweaks on Windows Vista, the drivers and system components to maximize performance. However, it is not allowed to apply and modification to the benchmarks.

And so, with fully configured hard drives and copper pots ready for liquid nitrogen, all of our contestants prepare to do battle for the chance to duke it out in France. Stay tuned for a more in-depth breakdown of how the first event here in the States transpired.

Our Second Round Of Mainstream PSUs

In the second part of our mainstream power supply test, we compare seven additional units with a maximum power of 700 watts. To enable an easier comparison, the units from the first part have been included in the charts for this piece as well.

The new candidates in the comparison are as follows :

And of course, the power supplies already tested in the previous part :

On to the first model, Enermax’s Modu 82+...

Enermax Modu 82+ 525 Watts

Enermax is jumping onto the energy-savings train, and has entered its 525 watt Modu82+ model into the race. The marketing name is designed to suggest greater efficiency to the customer: the reference to the 80Plus standard is surely not coincidental, and the Modu82+ does have this certification. But there is no 82Plus standard.

As the name suggests, this is a power supply with modular connection cables. The ATX connections and the CPU power are attached directly to the power supply. The sockets for PCIe as well as Molex and SATA are different, unlike the previous models we’ve seen, and cannot be confused. The connection cables are not compatible with the company’s older products.

The cooling of the power supply is handled by a 120 mm fan, with speed automatically regulated based on temperature. There is no manual setting, so you’ll have to trust that Enermax’s balance between noise and heat is acceptable The fan follow-up control used to protect the components, which you may have been familiar with from Enermax’s previous series is no longer included with the Modu82+.

Enermax Modu 82+ 525 Watts (Continued)

Enermax makes a sufficient number of SATA and Molex connections available, but the gaps between the connections are very small. This may lead to trouble with installing drives that require more clearance.

Enermax is the only manufacturer to have added a cable with a speed indicator for the fan. This enables the user to keep an eye on the power supply fan using monitoring software—a welcome addition for enthusiasts.

As we’ve come to expect with an Enermax modular power supply, a bag is included for the connection cables. Four additional cable ties with hook and eye fastenings ensure a nice, neat system build.

Voltage

Efficiency

The efficiency at medium load is surprisingly good, at almost 87%, but this drops considerably at maximum and low loads. In both of these cases, it only reaches 84%. However, the "Modu 82+" name is justified even here, as the efficiency is constantly above that figure. For what it’s worth, that’s good enough to qualify the supply for the 80Plus Bronze certification, if Enermax pursued it.

This model has some difficulty in our Low Power Test with a low load of 35 watts. Here it barely reaches 72%. The standby energy expenditure lies at 600 mW which is very low. With load, the power supply consumes 4.35 watts, which is average.

The 525 watt version of the Enermax Modu 82+ is available for a price of around $140.

Enermax Modu 82+

The Enermax Modu 82+ is a solid, modular power supply with high efficiency in the partial load sector.

• Positive
• Negative

• High efficiency under partial loadsLow standby consumption
• Confusing “82+” specification in the product name

Hiper Type R MK II 680 Watt

Hiper sent us the MK2, the second generation of its Type R power supplies to take part in our test. Unlike the earlier models, the one is no longer fitted with modular connection cables; the attractive metal screw connectors are a thing of the past. But it is still sold in the traditional manner, in a practical plastic box. With its reflective surfaces and perforated sheets, the power supply makes a good visual impression.

A highlight here is the USB ports on the power supply. These effectively make the power supply a USB hub as well, so it is able to supply a full 500 mA of current to any USB port. For the USB ports on the motherboard, this power is limited to a single root hub. The USB data cables, on the other hand, are connected through to the motherboard. An additional USB port is available as a power supply-only (without the data function) and can supply up to 1 A current, making it ideal for charging a cell phone or connecting USB mug warmers or other passive "USB toys."

Technically, the power supply is very impressive, with all voltages constantly within tolerances. Hiper has clearly learned from its past mistakes.

Hiper Type R MK II 680 Watt (Continued)

The power supply can only really be considered "partially modular," as the manufacturer has only supplied short adapter cables, called Xtenders. This means that it is possible to enable neat cable management, without the classic modular cables.

The connection cable, on the other hand, is modular. The manufacturer has included a modular IEC cable with German and English style power plugs to save weight.

As a nice bonus, the power supply packaging can be used as a box for small parts in the workshop or the garage.

Voltage

Efficiency

The Hiper achieves a good efficiency level at half load of 87%, while under full load it is a respectable 85%. If the load is reduced, the level of efficiency also drops. At low load, it is a slim 83.5%, while at a load of 250 watts it is still 84%. The efficiency becomes poor when it comes to our Low Power Test: 65% is not as good as we’d like to see here.

In standby mode, the energy consumption is relatively high. Without load, the power supply draws 1.63 watts from the main power source, and with a load of 2.50 watts, the power supply draws some 5.93 watts. This high consumption is probably a result of the integrated USB hub.

The 680 W Type R MK II watts can be purchased for around $150.

Hiper Type R MK II

Hiper brings new innovations to the power supply market; its integrated USB interfaces offer interesting capabilities.

• Positive
• Negative

• Integrated USB hubAdditional USB power supplyHigh efficiency under full and partial loads
• Low efficiency in standby

Tagan SuperRock TG680-U33II

The Tagan SuperRock TG680-U33II has a maximum performance of 680 watts. Conspicuous on this model is the LED on the rear of the power supply, which shows the operating status: green means that the unit is operating, while orange tells you that it is in standby mode. The power supply is fitted with a 120 mm fan, which is located on the underside and regulated automatically.

Tagan SuperRock TG680-U33II (Continued)

The manufacturer has chosen not to fit the power supply with modular connection cables. Cables and connectors are available in suitable quantities, though: 6x SATA, 6x Molex, 1x PCIe, 1x PCIe 6/8, and 1x P4/P8.

The cables have a plastic braided coating along their entire length, which makes them very stiff at the connectors.

One positive is the different colors used for PCIe and P4/P8 connectors, which prevents confusion.

Voltage

Efficiency

The Tagan power supply has a very high level of efficiency, not dropping below 85%. The highest value of 88% is achieved at medium load. Only when the load is very low—our fictitious Low Power System with a 35 watt loss—does the efficiency drop below 65%.

The Tagan unit does not do so well in standby mode, though. Without load, the power supply draws 1.58 watts from the power source, and if the power supply is loaded with 500 mA on the standby rail, it takes 4.70 watts. Here, the efficiency is barely over 50%.

Unfortunately, the SuperRock appears to be one of those products that is enjoying plenty of popularity overseas, but isn’t yet available to North American customers.

Tagan SuperRock TG680-U33II

The Tagan SuperRock power supply is a quiet unit featuring modular connection cables.

• Positive
• Negative

• High efficiency under load
• Non-modular designLow efficiency in standby mode

Fortron/FSP Everest 80Plus 600

Fortron has given its new 600 W Everest power supply the name "80 Plus", probably to ensure that it is found when customers search for this term. It lives up to this name, and has the appropriate certification (as do many of its competitors). As is usually the case for Fortron retail products, the power supply is blue in color.

The Everest is modular; only the ATX connector, one PCIe line and the P4 and P8 connectors are permanently fixed to the power supply. Aside from these, there are three ports for SATA and Molex cables, as well as two for additional PCIe cables. For the latter, the cable coatings and the ports are both also blue in color.

Fortron advertises its Everest power supply with a special function: at a load of up to 20% (120 watts) the power supply functions silently. This can be seen as soon as the unit is switched on. The fan starts up for a brief moment and then stops again immediately. Interestingly, the operating instructions don’t make any mention of this feature, which could cause the more easily spooked user to assume a fault. After all, even when under full load, the fan only starts up after a time.

Fortron/FSP Everest 80Plus 600 (Continued)

The power supply may be completely without noise at low loads, but when it reaches the switch-on threshold, the fan is clocked at such a low speed that it makes a creaking noise. This is very disturbing in the range between 110 and 250 watts. The temperature in the power supply increases, the fan starts to rotate with a creak and cools the temperature down, it stops, and then starts up again less than two minutes later. If the user isn’t constantly working in either an extremely low performance or an extremely high performance range, he or she will soon be rather annoyed with the power supply turning the fan off and on.

Our measurements also found that fan-less operation at a full 20% load is not possible. The threshold in the Tom’s Hardware lab with an ambient temperature of about 25 °C lies at about 100 watts.

An interesting detail is the blue lit power switch: this becomes brighter when the unit is switched on than it is when in standby. If the power supply is switched off completely, it is not lit at all.

Voltage

Efficiency

The efficiency of the power supply is very good with a partial load. It achieves a level of 88% at 50% load and almost 86% at 20% load. The power supply also does well in the 35 watt test, where it still reaches 74.4%. Under full load, the power supply reaches 85.6% which is still a good level of efficiency.

The power supply doesn’t achieve particularly special results in standby mode. A power consumption of 4.41 watts results in a level of efficiency of just 56.6%. Without load, consumption is pretty low at 0.61 watts.

Incidentally, the difference in energy consumption with the fan on or off is just 0.5 watts.

Fortron FSP Everest 80Plus 600W

Fortron/FSP’s power supply with partially passive operation was a good idea. The implementation is not successful, though, as the power supply is louder than a traditional unit.

• Positive
• Negative

• Silent at very low loadsHigh efficiency
• Passive operation is not possible at 20% loadPower supply is much too loud when switching the fan on

PC Power & Cooling Silencer 610W

PC Power & Cooling takes quiet operation seriously, or at least one would assume so based on the name "Silencer" given to its 610 watt power supply. As soon as it is switched on, though, we had to take issue a bit with that word. Although the fan doesn’t turn all that quickly, it hums rather disturbingly. On the plus side, at higher loads, the noise from the power supply does not increase very much.

This power supply is one of the few units to employ a classical construction with a 80 mm fan and airflow from front to back through the housing. This means it can be used in server housings and small home theater PCs.

Other details also remind us of a server power supply. The black and relatively heavy power supply has no frills and the connection cables are not modular. We know from experience that the folks at PC Power and Cooling consider modularity a liability, since each connected cable is just another possible failure point as the supply ages.

A supplied test protocol from the Taiwanese company "Fast Auto Electronic" informs the purchaser of the tests carried out, including a test with maximum load on the individual voltage rails. The FA-4200ATE test device was used for the tests.

PC Power & Cooling is one of the few manufacturers to use a single rail voltage supply on the 12 volt rail, which can be loaded with 49 Amperes.

PC Power & Cooling Silencer 610W (Continued)

The SATA and Molex connectors are available in suitable quantities. The cables with the Molex connectors are of varying lengths, which is very practical when dealing with large tower housings.

Two blue caps can be slipped onto the redundant Molex connectors to prevent short-circuiting due to contact with electronic components or the housing. These caps are not available for SATA or other connectors though, given the much safer design of those interfaces.

Voltage

Efficiency

As far as efficiency is concerned, the power supply from PC Power & Cooling is not able to produce the best results in our roundup. The maximum efficiency reached by the unit is at a 50 percent load, where it hits 86%. If the load is increased or decreased, the efficiency drops considerably, though; in the low load range (20%) it is just 83%, while under full load, the unit achieves 84%.

In standby, though, the efficiency of the power supply achieves top marks: under a standby load of 0.5 A, the unit draws 3.73 watts, which corresponds to a level of efficiency of 67%. Without load, the power consumption is less than one watt. Here, the unit draws 0.72 watts from the power source.

Fortunately, the Silencer is widely available in North America and is available online for just less than $100, making it a reasonable value.

PC Power & Cooling Silencer 610W

The Silencer 610W is a power supply with a classical design. Its construction and large size means that it is more suited to use in servers.

• Positive
• Negative

• Good standby efficiencyClassical fan design for servers and HTPCs
• Humming fan

Silver Power SP-SS500

The 500 W SP-SS500 from Silver Power is one of the few units to not have cable management. The connection cables are coated along their full length, which means that they are somewhat stiff at the drive connectors, with the exception of the SATA cables.

The power supply is relatively quiet at low and medium loads, but when put under high load, the 120 mm fan rotates quickly and the power supply becomes rather loud. The voltages are very stable, and do not change even under high load.

Silver Power SP-SS500 (Continued)

Peripheral connections are available in sufficient quality (6x SATA and 6x Molex), but the gaps between the individual connectors are very small, which may cause problems when it comes to connecting numerous devices. The SATA connection cables and the Molex cables are all of differing lengths. Along with a purely PCIe connection employing 6 pins, there is also an divisible connector with either 8 or 6 pins.

Voltage

Efficiency

The efficiency of the Silver Power supply is refreshingly high at 87% for medium loads, and remains at 86% even at low loads. Only under full load does the efficiency sink to 84%.

In our 35 watt test, the power supply achieves 75% efficiency, which is still a comparatively high level.

The low load in standby is worthy of note here; the Silver Power supply only uses 0.43 watts and thus has the lowest power consumption of the entire test field. Also, with load on standby of 500 mA, the Silver Power power supply is the most efficient, with a power consumption of 3.70 watts. This puts it at the top of the field (even if only by a small margin).

Yet again, however, we find this supply available overseas without much availability in the US.

Silver Power SP-SS500

Efficient power supply that is too loud under high load.

• Positive
• Negative

• High efficiency
• Very loud under high load

Corsair HX520W

The Corsair HX520W is something of a veteran among power supplies ; the product has been on the market for quite some time now. This unit is a modular power supply with an output level of 520 watts.

The power supply has a 120 mm fan which is very quiet when running, even under high loads. Conventional cables permanently affixed to the power supply are used for both the CPU supply and the ATX connections.

The labeling of the power supply on the rear is very informative, allowing you to determine the characteristics of the supply without needing to open up the case. Other manufacturers should take a page from Corsair here.

Corsair HX520W (Continued)

Flat, flexible cables are used as the modular cables, similar to the Cooler Master Silent Pro series; they have the advantage of being very easy to install in the housing. It is just as easy to form tight rings at the graphics card, as it is to form small loops at closely fitted drives, and unneeded cables can be stored in the supplied cable bag. The connections for the floppy drive are implemented via an adapter connection.

Voltage

Efficiency

The maximum level of efficiency for this model is achieved at half of the maximum load, where we measured 86%. Under full load, the value drops to about 84.5%, while at low loads (20%), the power supply achieves a comparatively low efficiency of 82.4%—the lowest in our test field. Here, we need to take into account the fact that the design of this power supply unit is older than that of the other models.

In testing with our 35 watt PC, the efficiency sunk to 67.9%, but in comparison with the competition, that was not a bad result.

The standby consumption of 0.70 watts with no load is respectable, and even with load on the standby cable, the unit still does alright. A power consumption of 3.93 watts with a load of 2.5 watts is good when compared to the competition.

The Corsair HX520W can be purchased from retailers at a price of around $120. And because it comes from Corsair, local availability is plentiful.

Corsair HX520W

The "veteran" amongst the power supplies can still keep up with the competition.

• Positive
• Negative

• High efficiency at full loadQuietGood modular cables
• Low efficiency under low loads

Charts: Efficiency Specs

Charts: Efficiency At 250 Watts, 35 Watts, And Standby

Conclusion: Recommendations for Hiper and Corsair

Our editorial team’s recommendation is the Type R MKII 680 from Hiper. It is a solid power supply which has an interesting, unique feature—USB interfaces.

The voltages provided by this model are stable and constantly within specifications. In addition, the unit has a commendable level of efficiency in normal operation. It only drops very low in standby mode due to the integration of the USB interfaces.

The best buy recommendation is the older HX520W unit from Corsair. For right around $120, you get a solid power supply with modular, flat, flexible cables that are easy to install.

Despite the older design, the efficiency of the power supply can still impress even today, particularly under partial load and during standby.

The Fortron and the Everest 600 power supplies came up with a good idea, enabling partially passive operation. But our high expectations for the feature were not met. Their fans start up at a lower load than is specified and becomes very loud in the transition range.

Mainstream PSUs Up To 700W

It’s been a while since our last large power supply test. Nowadays, almost all manufacturers label their power supplies with the 80 PLUS logo (for more information on 80 PLUS and what the branding actually means, check out the organization’s site. Energy efficiency has become essential to both vendors and users.

But other characteristics also play a more critical role now. Today’s power supplies include more flexibility than ever before. In addition to the familiar fan control, power consumption display, and cable management system, you’ll sometimes find USB ports and chic accessory bags. But what features—and what fundamental specifications—do you really need ?

Our German lab recently tested 10 current mainstream power supplies with power delivery of up to 700 watts. In part one (this one) we introduce the first five products. Look for part two shortly, when we’ll discuss the next five.

In part two, we test the following products :

First Up: Cooler Master Silent Pro 500 Watt

The Silent Pro series from Cooler Master consists of three models ranging from 500 watts to 700 watts. For our test, Cooler Master sent its 500 watt offering.

Unlike competing PSUs, the Silent Pro power supplies includes its own form of cable management—you’ll immediately notice the flat black cables, which are much easier to tuck out of the way than the thinner, rounded bundles typical of high-end PSUs.

At first glance, the cables appear unexciting, but they offer tremendous advantages: to begin, the cords are extremely flexible and allow you to connect devices without putting tension on the cords, which would then also put pressure on the ports. This innovation eliminates the common concern about SATA power connectors when attaching hard drives using power supplies with stiff cords.

Cooler Master Silent Pro 500 Watt (Continued)

The cable management connector sockets are labeled, and they’re shaped so they can’t be plugged in incorrectly. The five-pin sockets are designed so that you can attach modular 4-pin Molex and SATA cables as needed. The 6-pin sockets are for auxiliary PCI Express graphics cables.

The power supply is cooled by a 140 mm fan on the underside. The warm air is exhausted through a breadboard on the back. The fan is controlled automatically by the power supply according to operating temperature—no need for manual adjustment.

The maximum power is, as advertised, 500 watts. The maximum combined load on the 5 V and 3.3 V rails is 145 watts.

This one seems to be more widely available in Europe—none of our favorite online vendors seem to be stocking it yet .

Voltage

Efficiency

The power supplies we’ve seen from Cooler Master have relatively high efficiency. However, the Silent Pro is a bit disappointing. Instead of remaining at the top, the Cooler Master unit ends up somewhere in the middle of our pack; it gets close to 87 percent efficiency but doesn’t quite make it. Therefore, the efficiency at low loads of 85.3 percent doesn’t drop as much as it might if our 50% load numbers were higher. Under maximum load, the PSU reaches almost 84 percent.

Very surprising, however, is the standby power consumption: a mere 0.35 watts—the best among the products we tested. This is the draw taken from the connection socket when there is no load. When the load increases to 2.5 watts, the power consumption remains good, at 4.12 watts.

Cooler Master Silent Pro 500

The power supply offers acceptable performance for the price. The flexible cables make wiring a much simpler proposition, which we like.

• Positives
• Negatives

• Flat, flexible wires Lots of connectivityLow standby energy consumption
• Relatively low efficiencyCords aren’t color coded

Thermaltake Toughpower QFan 650W

Thermaltake promises robust power, hence the name “Toughpower.” We tested the 650 watt version of the QFan series.

The white fan keeping the power supply cool is fairly large at 140 mm, and it spins somewhat slowly due to its size. This is good news—lower RPMs translate to less noise output. The fan intake in the housing is designed to improve air supply even as acoustics remain manageable.

The power supply provides its own form of cable management. The SATA and 4-pin Molex connectors, as well as the PCIe cords, can be plugged into the power supply’s available 6-pin receptacles. They’re pinned differently, making it impossible to connect them incorrectly.

Thermaltake Toughpower QFan 650W (Continued)

At the three load levels (low, medium and high), the voltages are within the allowed deviations.

However, the situation changes dramatically at low loads. If the current on the 12-volt lines drops below 1 amp, the voltage on the 5 volt line decreases drastically. At a load of 0.42 amps, we measure only 4.03 volts (minimum value according to specification is 4.75 volts). At even lower loads the voltage drops to 3.8 volts. This is clearly a malfunction. The manufacturer has not yet responded to this concern.

Due to these issues, the Low-Power PC Test could not be performed.

The power supply can be purchased online for right around $150.

Voltage

Efficiency

Efficiency-wise, Thermaltakes’ power supply falls into the middle of the pack. Its most promising value is at medium load—a respectable 86 percent. At full and low loads, it’s about 84 percent.

In standby mode the device consumes 1.4 watts. At a load of 2.5 watts it uses 4.90 watts. We think that these results could be better.

Thermaltake Toughpower QFan 650

In normal operation, the power supply shows good results. However, at low load the device is outside the specifications and therefore doesn’t pass.

• Positives
• Negatives

• Good Ventilation Concept
• 5 volt line at low load is outside the specifications

Silverstone SST-ST70F Strider 700W

Silverstone sent its 700-watt SST-ST70F Strider to our test course. Although the power supply arrived in fairly unassuming packaging, the unit itself features good cable management system and plenty of cables for attaching peripherals.

The power supply comes with all its lines plugged in by default. The SST-ST70F is the only modular power supply in the test field where you can literally remove the lines—including the processor power supply and the ATX cable.

The removable nature of the ATX cable doesn’t make sense because you can’t run a power supply without its ATX line, but it lets you get the stiff ATX cable out of the way when you want to remove other components from the small device.

Silverstone SST-ST70F Strider (Continued)

The design of the cables themselves warrants our recognition. A plastic mesh coats the long part of the cable up to the first plug. The rest of the cable is held together with a plastic wire strip. This simplifies installation and eliminates concern about breaking a connector on the hard drive during installation.

In addition to six SATA and six Molex plugs, you’ll find two floppy connectors on the included cables. The power supply also comes with four PCIe plugs as well as 4- and 8-pin processor connectors—all on separate lines. The ATX plug is seperate.

This power supply is available online for right around $139.

Voltage

Efficiency

Silverstone, the winner in many of our previous efficiency tests, seems to be resting on its laurels. This time around, it lands in last place. Let’s look at the numbers:

Under full load, the Strider’s efficiency is 82.2 percent. At half-load it’s a bit better—Silverstone ends up in the midfield with 86.5 percent. The SST-ST70F truly looks good only at low load, where the power supply outperforms its competitors with 86.7 percent.

This trend continues as the load falls. In our low-power comparison at 35 watts, the Silverstone power supply performs best in the test field with more than 78 percent. But if you’re a low-power user, you’re probably not in the market for a 700 watt power supply, so it’s a bit of a catch 22.

The result for standby is average. At zero load the Strider consumes 1.17 watts; at a load of 2.5 watts it needs 4.57 watts.

Silverstone SST-ST70F

The power supply from Silverstone combines high performance with good cable management, long cables, and many plugs.

• Positives
• Negatives

• Many peripheral plugsLong cablesGood efficiency at low load
• Low efficiency at full load

Antec Signature 650W

The chic packaging for Antec’s Signature power supply is certainly eye-catching. The power supply and its accessories are elegantly organized. This is one of the few power supplies that ia still cooled via conventional means (with the fan on the back)—that allows for use on flat server or HTPC housings.

An included test report from Antec gives us some additional detail about the ATX specification.

Antec Signature (Continued)

The power supply is partially modular. The ATX, processor, SATA, and Molex plugs are connected directly to the body of the supply. Extension cables are available for the SATA and Molex plugs.

Antec is the only manufacturer to provide more cables than can be plugged in simultaneously. As a result, the user can choose whether to use more SATA or more Molex cables. The additional cables are labeled with the letter O (Optional) in the cable list.

The voltages are always within the allowed specifications. However, under full load, the voltages on the 3.3 volt and 5 volt lines drop slightly.

Naturally, all of Antec’s niceties come at a price. You can expect to pay just over $200 for this 650 W power supply online.

Voltages

Efficiency

Antec’s power supply shows remarkable efficiency under full load as well as half load. It’s able to demonstrate 87.2 percent and 88 percent, respectively. Under low load, efficiency drops slightly to 83.3 percent. Just bear in mind that under full load, this is the best device in our test field.

The power supply is frugal in standby, too. It uses 690 milliwatts at zero load. At a load of 2.5 watts on the standby cable it consumes 4 watts, an efficiency of 62 percent. Antec’s voltage converter does relatively well with this result, too.

Antec Signature 650 Watt

The Antec power supply is the classic size and is thus well-suited to use in 19-inch and HTPC housings.

• Positives
• Negatives

• Fan design is ideal for HTPCs.Low energy need in standbyHigher efficiencyChoice options with Molex and SATA
• Higher price

Dragon Force DF-530GT

The DF-530GT is the least expensive power supply we tested (roughly 40 euro), though again, this is one of those units readily available in Europe and Asia, but not so much in North America. Clearly, you won’t get any special features for such a low price—this power supply has a straightforward design with no cable management.

The connection options are merely sufficient for today’s systems. For graphics cards, there is a 6- and an 8-pin plug PCI Express auxiliary power plug, which can be broken off into a second 6-pin connector. For the processor power cord, there’s only a 4-pin plug—you’ll have to use an adapter for boards that require an 8-pin connector. The ATX plug has 24 pins and is not separable.

Dragon Force DF-530GT (Continued)

The Dragon Force includes only three SATA plugs, three Molex plugs and one floppy connector. Needless to say, it’s a little light in the connectivity department.

The 120 mm fan has three settings: low-speed is silent, high-speed is high-performance, and standard adjusts speed automatically depending on the temperature.

The power supply delivers the needed power without a problem. The voltages are always within the allowed deviation and very stable—even under full load they hardly differ from the nominal values.

Voltages

Efficiency

The Dragon Force’s efficiency is very high relative to its price. At mid-load, the power supply has an excellent efficiency of almost 88 percent. This drops to 84 percent at maximum load, but even under low load the DF-530GT gets a good result of 85.5 percent.

The Dragon Force also does well in the Low-Power PC Test. Its efficiency drops to 76 percent at a 35-watt load, but that’s still extremely high compared with competitors.

In standby mode without load, we measured 1.14 watts out of the socket. Here, some competitors consume significantly less. At standby load it consumes about 4.86 watts, an efficiency of 51 percent. At this level, the power supply is in the middle of the pack.

Dragon Force DF-530GT

Straightforward power supply without special features for an inexpensive price.

• Positives
• Negatives

• Very low price
• No cable managementShort cables with few sockets

Charts: Efficiency Specifications

Charts: Efficiency @ 250 Watts, 35 Watts, And Standby

Conclusion: Dragon Force and Silverstone Win

The clear “performance for your money” winner in our power supply test is the Dragon Force. The DF-530GT is a solid device that delivers its performance without any issues, running on stable voltages that never deviate much from the nominal values. But you don’t get everything for this low price, of course. There’s no cable management, no accessories, and few cables are included. More problematic for North American customers is availability. Unfortunately, this unit will likely only be an option for our overseas readers.

Thus, our overall recommendation for this round (and Best of Tom’s award) goes to Silverstone’s Strider SST-ST70F. Priced at about $139, this power supply is unbeatable given its performance. It also comes with many long connection cables, which simplify configuration in component-laden PCs and eliminate the need for add-on adapters. Although the Silverstone’s efficiency drops under full load, it remains very high under half-load and low-load conditions.

Combining Performance and Efficiency

Combining minimal power requirements with decent desktop performance seemed to be a nearly insurmountable task for many years, but it now seems to be possible. There are processors and chipsets on the market that will allow you to run a decent desktop system at a very low idle power level, while providing amped-up CPU performance when it’s needed. We’ re taking a look at two motherboards, which their vendors Foxconn and Gigabyte aim at power-conscious customers, and pairing them to one of the most affordable mainstream processors on the market: Intel’s Core 2 Duo E7200.

Efficient Computing Requires Reasonable Components

When you want to create a truly efficient computer, you must bid farewell to some of the components that you may have once taken for granted. First of all, it is necessary to avoid hardware that increases idle power without doing anything important—this includes all add-on cards and drives that you may not really need, as well as most high-end or enthusiast products, and discrete graphics cards in particular. Once this decision has been made, you can go ahead and select your components according to your efficiency goals.

Upper mainstream and high-end graphics cards are typically some of the most power-hungry, as most are optimized for 3D features and performance— often at the cost of high power consumption. Despite recent optimizations in Nvidia’s lineup, for instance, most high-end cards still don’t offer sufficient power saving options, resulting in idle power requirements of graphics cards that match or even exceed the power requirements of an entire entry-level PC. While a high power requirement is not an issue as long as you get something in return, components in idle mode don’t provide any benefit, so we demand minimum idle power during times of low or no activity. In the case of our test systems, this meant we bid farewell to gaming in order to stay with the integrated graphics unit. Luckily, there is always the option of upgrading to a discrete graphics card down the road should it be something you’re missing.

Components: Core 2 Duo E7200 and G31

Once you accept the fact that powerful graphics, high-end components, and numerous disk drives cannot be part of an efficient low-power computer, you can choose the right foundation for your machine. We went with a mainstream Core 2 Duo E7200processor manufactured using 45 nm technology, which runs on a 1,066 MHz FSB with a 2.53 GHz core speed. Unlike the E8000-series, the E7200 has only 3 MB rather than 6 MB L2 cache. We’re also arming this setup with 4 GB of RAM and a nimble 7,200 RPM hard drive.

We decided to look at two motherboards using Intel’s G31 chipset: Foxconn’s G31MG-S and Gigabyte’s GA-EG31MF-S2. Both are reasonable platforms for our high-efficiency PC project.

Performance And Efficiency Basics

Performance Basics

There are a few things you need to know in order to assemble a desktop system that is fast enough for everyday work. The following recommendations apply to mainstream PC users who don’t want to spend a fortune as they work in Windows, modern applications, and with multimedia data; our recommendations will not be suitable for power users and enthusiasts.

• Be sure to get a reasonable dual core processor, which should be a Core 2 Duo E7000- or E8000-class chip (45 nm). AMD has competitive products with its Phenom X3, but these products require more power. Four cores are clearly more expensive, require more power, and are only faster if applications can take advantage of multiple cores. A decent dual core will provide high system responsiveness and sufficient performance for your daily routine, and you can always upgrade to a quad core if needed.

Our 2008 CPU Charts provide benchmark comparisons of today’s most relevant desktop processors.

• A discrete graphics card is necessary if you want substantial 3D graphics performance for games or other 3D applications, or if you want to attach multiple displays. You also need one if you insist on specific acceleration features, such as HD video playback acceleration by AMD/ATI or Nvidia, or application acceleration as emerging with Nvidia’s Cuda technology. If neither is applicable, integrated graphics are probably sufficient.

• Get at least 2 GB of RAM; even 4 GB memory kits are highly affordable today. Memory timings are secondary, but we recommend going for a brand you trust. If you insert two memory modules into two different memory channels you’ll ensure that the memory runs in dual-channel mode, which offers the best performance.

• Get a fast hard drive and insist on the latest model generation for performance reasons. Although prices have dropped considerably, there are still noticeable difference among similar products.

Our 3.5” Hard Drive Charts show the benchmark results of most of the latest desktop hard drives.

The question of how much performance you really need depends on what you want to do with your PC. We recommend always going for the fastest hard drive possible, as this is still the slowest core component inside your PC. Flash-based solid state drives have the potential to be many times faster, but you’ll also spend many times the amount you typically need for a hard drive. Currently, you get the best bang for the buck at capacity points of 500 GB, 640 GB or 750 GB with the Samsung Spinpoint F, Seagate’s Barracuda 7200.11, and the Western Digital Caviar Blue.

Differences between a 2.5 GHz processor and an identical 3 GHz model are only really noticeable if you require substantial performance for applications. As long as you don’t have such requirements, you should go for the model that provides the best cost per gigahertz within a product line. Search for the best processor stepping to make sure that you get the latest revision; for Core 2 processors this is currently the M0 stepping, which we used for this review—it requires less idle power than older versions.

Efficiency Basics

It is important to use a power supply that is efficient in its power specifications and to match the power supply to the expected power requirements to make sure the PSU can operate within a high efficiency range.

When selecting components for an efficient PC, it is important to start with the one component that has the most influence on system power consumption: the power supply unit. Though most of the branded power supplies comply with 80 Plus recommendations, which means that they reach 80% or more energy efficiency at popular load levels, you will still receive disappointing efficiency if you get a PSU that does not match your power requirements. A 1,000 W PSU running at only 40 W power requirements will not be efficient at all, which is why we decided to go for a suitable PSU: Fortron’s FSP220 at a maximum output of 220 W. It will not allow us to install a fast graphics card, but it is sufficient to run the Core 2 Duo on an integrated motherboard with a few drives. It would even suffice for a Core 2 Quad processor.

Once you’re sure that the power supply is appropriately efficient, you should pick your system components according to the efficiency principle. Going for low-cost hardware doesn’t mean getting efficient parts, and going for expensive high-end components typically doesn’t gain you any efficiency benefits either. There are a few basics to stick to: more components, such as additional storage controllers on a motherboard, increase your overall power consumption, as do multi-phase voltage regulators. Low clock speeds (FSB speed, memory speed) as well as low supply voltages help in optimizing power consumption. As a result, high-end or high-speed components should always be avoided if they are not necessary for your applications.

Selecting a few components with a focus on efficiency doesn’t help much. You have to pick all components with regards to efficiency if you really want to bring power consumption down. The results of this effort are impressive, as you’ll see in the benchmark section.

Intel G31 Express Chipset

The diversity in Intel’s chipset portfolio can be difficult to understand, as some features overlap from one model to the next. Picking a higher model number will not necessarily give you a more advanced chipset, either—the G35 offers the more advanced graphics unit, while it does not support DDR3 memory nor the newer ICH9 southbridge like the G33. Intel’s G31, however, is an entry-level chipset indeed and it typically requires the least power.

The G31 is classified as a mainstream desktop chipset, which Intel targets at “essential computing.” This means that the product clearly isn’t suitable for high-end systems and it does not support top notch features. Instead, it was designed to provide a great bang for the buck for average users. Hence it is aimed at hosting a Core 2 processor or Pentium Dual Core or Celeron equivalent based on the Core 2 micro-architecture.

Limited Feature Set

The G31 chipset is limited to 4 GB of memory, while the G33 and G35 support up to 8 GB. The lower-end model also only supports DDR2-800 dual channel memory (though this isn’t a disadvantage compared to DDR3) and it is bundled with the ICH7 southbridge instead of the ICH8, ICH9, or ICH10. This results in only four SATA/300 ports, but it also provides the platform with two UltraATA/100 channels, while the newer chipsets support either one legacy ATA channel or none at all. The G31 with ICH7 sports eight USB 2.0 ports, HD audio, and conventional PCI slots, but only a 100 Mbit network controller. Should you need a quicker Ethernet connection, please look for a motherboard that comes with a PCIe network controller for Gigabit speeds. This is the case for both boards we looked at. Finally, while supporting one x16 PCI Express slot for upgrades, G31 is not PCI Express 2.0-compliant.

Although initially specified for FSB1066 speeds, all the current G31 motherboards we looked at will also support FSB1333 processors. Both boards in this piece support the entire Core 2 processor portfolio, including the Core 2 Quads. Believe it or not, even the three-phase voltage regulator on the Foxconn G31 motherboard is sufficient to operate a Core 2 Quad Q9550 easily, should you decide to ditch a dual-core in favor of a quad-core processor.

GMA3100 Graphics

Integrated graphics solutions will not deliver performance or features anywhere close to those of decent 3D graphics cards such as the Radeon HD4850/4870 or the GeForce 260/280 by Nvidia. Even mainstream graphics cards based on GeForce 9600GT processors vastly outperform integrated solutions. However, we decided to stay with integrated graphics for the sake of maximizing power efficiency.

The G31 chipset includes Intel’s GMA3100 integrated graphics unit, which is based on the GMA3000 core. This is the graphics core that Intel first deployed on the 965 chipset family (called GMA X3000), and although the G965 supports pixel shader model 3.0, the G31, G33, Q35 are limited to SM 2.0, which means DirectX 9.0c capabilities. This is sufficient for all Aero features on Windows Vista, though. Motherboards with the G31 chipset typically include one analog D-SUB15 display output and sometimes a digital DVI output. Since the GMA3100 isn’t suitable for home theater PCs (HTPCs), boards will typically not have any HDMI outputs; don’t expect two digital outputs on these boards, either.

Core 2 Duo E7200

Intel’s Core 2 Duo processor was launched in late summer of 2006, and it arrived with a bang. All of a sudden, AMD’s Athlon 64 X2 processors were outclassed by a product that delivered better performance while simultaneously requiring less energy. Launched at a 2.66 GHz clock speed (2.93 GHz for the unaffordable Extreme edition), it held its superior standing for more than two years.

The first update was a bus speed bump that took the processors from FSB1066 to FSB1333 in the summer of 2007 with the launch of the P35 platform and slightly modified processors. The second modification was FSB1600 speed in the high end and shrinking the process from 65 nm to 45 nm. All of the Core 2 Duo E7000, E8000 processors and Core 2 Quad Q8000 and Q9000 are based on the 45 nm Wolfdale cores. Yorkfield, which is the core name for the quad core, utilizes two Wolfdale die.

Core 2 Duo E7000

Over time, Wolfdale received some improvements, the latest being the introduction of the M0 stepping, which continues to reduce processor power in idle states when compared to the L steppings. We looked around for a reasonable processor that utilized the latest stepping and we found the Core 2 Duo E7200, which utilizes only 3 MB L2 cache instead of 6 MB cache. Undoubtedly, the reduced L2 cache capacity represents another power saving opportunity, of which we wanted to take advantage.

The E7000-series currently consists of only two models: the E7300 at 2.66 GHz and the E7200 at 2.53 GHz. Both are based on the Wolfdale core, but again have only 3 MB L2 cache and a reduced front side bus speed of FSB1066—this is in contrast to the FSB1333 used for all Core 2 Duo E8000 models (which have 6 MB L2 cache). The E7000 series does not support virtualization technology (VT) nor Trusted Execution technology (TXT)—this isn’t an issue for our purposes, since both are insignificant for mainstream desktop users. A smaller L2 cache capacity together with relaxed clock speeds makes this processor an excellent choice for a high efficiency computer as you’ll see in the benchmark section. Despite being rated at the default TDP of 65 W for desktop processors, the Core 2 Duo E7200 stays pretty far away from that limit.

Foxconn G31MG-S

Foxconn’s G31 motherboard is called the G31MG-S. The second “G” stands for its “Green” series, which consists of three features that Foxconn refers to as 3G technology. GPS means Green Power Saving; GoD stands for Green on Demand; and GSM represents the Green System Mode. Green Power Saving refers to combining the motherboard with an efficient power supply to maximize power efficiency. Although we didn’t use a Foxconn PSU, we’re sure that our choice fits very well. Green on Demand is very much the same as implemented by other motherboard makers under different brands: voltage regulator phases are switched off when they’re not used, to increase efficiency per voltage regulator. Finally, Green System Mode refers to utilizing all power saving features that are available.

Motherboard Details

The board supports a FSB1600 setting, which we found to be absolutely reliable. However, this setting is not officially supported by the chipset. Foxconn uses only solid capacitors and a three-phase voltage regulator that we believe makes the most sense for a high efficiency system. The chipset components are passively cooled and there are only two memory sockets. While this isn’t upgrade-friendly, it ensures that you don’t install more memory modules than necessary. In addition, using two 2 GB DDR2-800 DIMMs will already take you to the G31 chipset’s 4 GB memory maximum anyway.

Upgrade Options

As with most other motherboards utilizing integrated graphics, the G31MG-S comes with an x16 PCI Express slot, which will host a discrete graphics card should you need serious 3D performance. An additional x1 PCIe slot and two 32-bit PCI slots accept other add-on cards. Foxconn implemented one of the two possible UltraATA/100 channels, which should still be sufficient. All four SATA/300 ports are available, but RAID is not supported.

The board clearly was designed for low-cost desktop applications, as it only features an analog D-SUB15 display output. We believe that at least a DVI port is mandatory, as the analog output decreases visual quality on high-resolution displays. All other interfaces are very much standard.

Power and Performance

The performance of the Foxconn G31MG-S was very much in line with what we saw from Gigabyte’s board, so the benchmark differences are not worth mentioning. Idle power, however, was amazingly low on this board. While Gigabyte’s solution required a minimum of 38 W at all times, the Foxconn G31MG-S runs with as little as 31 W on idle—this represents 18% less idle power needed for Foxconn. Power required to finish the SYSmark 2007 Preview was smaller as well, hence this board won the battle for performance per watt. The only stain was the slightly higher peak power of 61 W as opposed to 60 W with the Gigabyte board.

Voltage power doesn’t seem to change between idle and max power: in both cases, CPU-Z reported 1.215 V. Foxconn still showed much better power efficiency with its product.

Gigabyte GA-EG31MF-S2 (G31 Chipset)

The second board in this little roundup comes from Gigabyte and is called the GA-EG31MF-S2. Unlike the Foxconn product, it offers four DDR2 memory slots, although you are still limited to the maximum supported memory capacity of 4 GB for the G31 chipset. It does not come with all solid capacitors, which doesn’t have to be a disadvantage. Again we only found a D-SUB15 analog output for displays, which we think is not the best choice—any modern graphics solution should at least have a digital DVI output.

Features

The board comes with the same basic features as its competitor, but Gigabyte adds some more components, such as a FireWire/400 controller. A second port can be activated once you purchase an adapter cable and attach it to the additional FireWire header on the board. The same applies to four more USB 2.0 ports; four are ready to be used. Chipset components are kept cool using heat sinks with no moving parts, which is good for keeping both system noise and power consumption low. You get one UltraATA/100 channel and a floppy port as well as two 32-bit PCI slots. Like Foxconn, Gigabyte provides one x16 and one x1 PCI Express slot for upgrades.

Power and Performance

All mainstream and better Gigabyte boards are equipped with voltage regulators that Gigabyte states are fully VRM 11.1-compliant. The layout on the GA-EG31MF-S2 is based on four phases and solid capacitors, which provides better voltage stability at high current levels. This is necessary for power-hungry processors such as 3+ GHz quad core models. However, we found that the Gigabyte board required considerably more idle power than the Foxconn board, despite similar components and despite the fact that Gigabyte’s Dynamic Energy Saver (DES) technology was activated. DES allows the system to switch off voltage regulator phases when they’re not required.

Test Setup

We used a Fortron FSP220 power supply (on the right), which is sufficient for powering any mainstream PC without discrete graphics. Thanks to this power supply, the power efficiency went up considerably compared to our reference PSU, the Cooler Master RS850-EMBA (the PSU on the left).

Benchmarks and Settings

Synthetic Benchmark Results

Windows Vista Experience Index

Application Benchmark Results

Audio/Video Benchmark Results

Audio Benchmarks

Video Benchmarks

SYSmark2007 Results

Power And Efficiency Results

System Idle Power (w/ All Power Saving Options)

Peak Power Requirement in SYSmark 2007 Preview

Average Power Requirements During SYSmark 2007 Preview

Total Power Used for a SYSmark 2007 Run

SYSmark 2007 Score per Watt-Hour

Power Diagram for SYSmark 2007

You can clearly see that the Foxconn system’s idle power is a few watts below the minimum power requirement of the Gigabyte system, both using exactly the same hardware. Peak power, however, is similar; this is because of Gigabyte’s four-phase voltage regulator, which is more efficient at high loads.

Conclusion

The G31 chipset does not have a single feature to make it really special. It does not support DirectX 10, it does not provide 3D performance that is anywhere near useful for gaming, and it is limited to 4 GB of memory. But none of this is necessary for a basic workhorse PC for everyday use. The G31 is cheap, it supports all Core 2 processors including the quad-core models, and it can be upgraded with any high end graphics card, making it almost as suitable for gamers as a high end chipset. It’s the motherboard makers who turned G31 from a lame mainstream duck into a chipset option for efficient platforms.

Foxconn Shows Better Efficiency

Both motherboards do a good job of providing a reliable platform for office and multimedia desktops, and they artfully do the dual-duty of combining low idle power with decent processing power for mainstream applications. The Core 2 Duo E7200 in the current M0 stepping we used helped to maximize power savings when idle, while delivering sufficient performance for everyday work. The Gigabyte GA-EG31MF-S2 required between 38 and 60 W; Foxconn was more conservative with motherboard features, but managed to be even more efficient in idle at an amazing 31 W minimum power requirement. Apart from the missing DVI display port there was little to criticize about the two boards. They are both priced well below $75, and both also ran flawlessly with a fast Core 2 Quad CPU if necessary.

31 W Idle Power Requirement

Let us underline that the 31 W idle power figure isn’t far from the figures achieved by a low-power system using an Intel Atom or a VIA Nano processor, where systems typically require between 20 and 35 W, depending on the power supply, chipset memory and hard drive used. The Core 2 Duo E7200, however, provides serious performance when you need it, while the Atom and Nano cannot even beat the performance of a three-year-old PC.

Smart Component Choice Beats Low-Power PCs

The idea of getting an Atom system or one based on VIA’s C7 or Nano processors for your home server or multi-purpose hallway or living room PC sounds very reasonable, as these were designed to consume little power. But these products will fail you as soon as you need more performance. Atom, Nano and others are great solutions for workloads that are fully predictable and controllable, but your everyday work is not. You may want to use a new application, which means the dedicated low power machines will fall flat on their figurative faces. Going for a reasonably-configured high-efficiency PC will cost a bit more, but it will not increase the idle power by much, while providing you with performance reserves that are many times above the level of the low-power options we’ve been testing for the past few weeks here at Tom’s Hardware.

Passive Power Supplies In The Lab

If you want a computer to truly be silent, you need to use passive cooling components. Whereas passive cooling components are more frequently available in the graphics card and CPU sector, they are something of a niche market when it comes to power supplies. This is not really a surprise though, given that purely passive units are considerably more expensive than those with fans, and they generally provide relatively poorer performance as well.

Our test compares the Zen 400 from Fortron (FSP) with a power output of 400 watts, and the Nightjar 450 from Silverstone which provides a maximum of 450 watts. Both models can be purchased for less than $200.

In order to ensure proper comparison of the units—and in anticipation of future testing with other units having different power output levels—we tested the efficiency at 100%, 50% and 20% load. In the future, we will also be testing the efficiency at an output of 250 watts, as well as an extremely low output of just 35 watts. The latter is of particular importance for low power systems, to ensure that users are able to find the right kind of power supply for these as well.

Fortron/FSP Zen 400

The Zen 400 from Fortron, with its output of 400 watts, stands out immediately thanks to its blue color. Compared to its 300 watt predecessor, the first thing you notice is a large aluminum cooling element on the underside of the unit, which is reflected in its considerably higher weight.

There is a single 6-pin connection for graphics cards. The ATX connection corresponds to the ATX12V 2.2 standard; it has 24 pins and can be split to accommodate older motherboards.

The unit’s connections are no longer up-to-date. It only has four SATA connections, yet these are becoming more and more important. Oddly, there are still six of the 4-pin Molex connections used by older drives. There are also two floppy connections on the power supply, though these can be used for fan controllers. The +12V CPU connection is only available as 4-pin version, which can lead to problems with some boards that require an 8-pin connection.

The length of the connection cables also leaves a lot to be desired: only 16” (40 cm) to the ATX connection. This may be too short for larger cases, and as a result, it may not be possible to route the cables as neatly as one would like.

Fortron/FSP Zen 400 (Continued)

On the rear of the power supply you’ll find the IEC socket and a blue illuminated main switch. The latter is of particular importance, as a user operating a passive power supply cannot hear whether or not it is switched on. There are no other indicators showing the operational status of the unit.

The power supply is a full range unit that can be operated on a voltage ranging between 110 V and 240 V. There is no need to manually switch between the voltage ranges.

The output of the power supply is 400 watts, and 560 watt peaks are possible for short periods of time. The combined performance of the 3.3 V and 5 V rails is a maximum of 130 watts.

During use, the power supply makes its presence known by a high-pitched, high-frequency noise that occurs most often when under heavy use. Up to a load of approximately 250 watts this noise can hardly be heard, but the higher the load gets, the louder the noise. This is a very annoying noise that detracts from the "silent PC" concept.

With no additional ventilation, constant operation under full load at 400 watts is not possible. The power supply switches itself off due to overheating after approximately two hours. A level of 300 watts, however, can be maintained permanently. A minimal movement of air in the computer case—such as a slowly rotating 120 mm fan—enables the device to be operated permanently at high load.

The test with the D201GLY2 board from Intel was not a problem—this kind of mini ITX board often causes problems with power supplies.

Voltages

The output voltages ranged within the permitted values for the entire duration of the test period. In particular, the important 3.3 V, 5 V and 12 V voltage rails only deviated by a few millivolts from the nominal values. Only under full load were there slightly increased breaks in voltages, which were again well within the specification.

Efficiency

Efficiency plays an ever-increasingly important role in modern computers. Particularly with HTPCs designed to run constantly in living rooms, the high level of efficiency of the power supply ensures significantly lower electricity costs at the end of the year.

Overall, the Fortron power supply performed well. In the three load stages of the specification, the unit was constantly at over 84%. The best performance was obtained at below half load, where the efficiency increased to over 88%.

In our fictional system with a 250 watt load, the efficiency remained at a very good 87.7%, but in our low power system it only managed 76.3%.

In standby mode, with no load at all, the power supply only used 440 mW. With a standby load of just 2.5 W (500 mA at 5 V standby voltage) the Fortron converter drew 3.77 W from the socket, which is an efficiency of 65.5%.

The Zen 400 (PPA4000300) from Fortron (FSP) is available for a price of around $150.

Silverstone Nightjar 450

The Nightjar 450 from Silverstone, which is also listed in shops under the name ST45NF, has a maximum nominal performance rating of 450 watts (a higher peak load is not specified by the manufacturer). The basic construction of the power supply, with a large aluminum cooling element, corresponds to that of the Fortron unit.

When it comes to connections, Silverstone has invested in the future somewhat. As a result, there are six SATA connections on the unit, along with six 4-pin Molex connections. The white-grey color combination reminds you of days gone by, but when hunting around in a tower for the right connections, you realize that the different SATA connection colors are practical.

The CPU connection is split, and can thus be used to suit both four- and eight-pin requirements. Silverstone has also employed this method for the PCIe connections: two six-pin connections are available, one of which can be extended to an eight-pin connection for modern graphics cards. The ATX connection is split in this way, and offers options with both 24-pin and 20-pin connections.

Silverstone has also not scrimped on the cable lengths; the ATX connection has a cable over 21” (55 cm) in length.

Silverstone Nightjar 450 (Continued)

On the rear of the power supply there is an IEC socket and an non-illuminated main switch, as well as two LEDs that indicate the operational status of the power supply.

Whereas the top LED shows the temperatures—at over 55°C the usually green LED switches to yellow—the lower LED tells you whether the unit is in standby (orange), switched on (green) or cannot be switched on due to a fault (red). The latter occurs, for example, when the power supply has overheated and deactivated itself.

Unlike the Fortron Zen, the Silverstone unit is silent—even under the highest load—and does not make any whirring noises.

If the unit is placed under a load of 450 watts without ventilation, it switches itself off automatically after about two hours due to overheating, but at an output of 330 watts, we were able to operate it constantly. As with the Fortron unit, a very low flow of air is sufficient to enable constant operation even under a full load of 450 watts.

The Silverstone Nightjar was also able to operate our D201GLY2 board from Intel with no trouble at all.

Voltages

The Silverstone Nightjar makes no mistakes when it comes to voltages either. All values were always within the permitted ranges set out by the ATX12V specification.

Efficiency

Under full load, the efficiency of 87.3% is the same as the Fortron unit, while at medium and lower loads, it is a mere 1% worse than its competitor.

The direct comparison using our fictional 250 watt system showed marginal differences, with the 87% output of the Silverstone Nightjar is again on a par. The biggest differences could be seen with our low power system. Here, the Silverstone converter performed at 71.5%, almost 5% below the Zen 400 unit.

Even in standby mode, the Silverstone unit couldn’t do any better. Without load, the power supply uses 2.57 watts, probably due to the LEDs and their controller. Even with load on the standby cable, the result is pretty unimpressive. The power supply draws almost 6 watts of power in order to make 2.5 watts of output available. The result is a marginal efficiency of 42.7%.

The Silverstone power supply is a fair bit more expensive, priced online around $200.

Charts

Results: Our Choice--Silverstone

Our results favor Silverstone’s Nightjar. At medium and high outputs, the efficiency is very high, and the unit cannot be heard even under full load. Also advantageous are connection cables at sensible lengths, which also enable use in larger cases, and flexible connection options with the separable 8-pin connections for CPU and graphics card.

The less expensive power supply from Fortron only gains points for operation with low power systems, which work with outputs in the 50 watt to 150 watt range—at higher output ranges the power supply starts to emit a bothersome high-pitched noise. On the other hand, the constantly high efficiency and low standby power consumption are notable.

There are so many power supply units on display at Computex. Most manufacturers focus on showcasing the big stuff, meaning 1,000-W PSUs and up. Most companies offer units that provide up to 1,200 W, and some are already talking about 1,500 W. Those ones, however, don’t fit into the compact standard ATX dimensions of 163x150x85 mm. FPS’s new and modular 1,200-W model Everest Pro 1200/1250 is slightly larger: 180x150x85 mm. It has all the features you’d expect from a high-end PSU: Active PFC, a 135-mm fan, ATX12V 2.2 compliancy and EPS12V 2.92 compliancy, a modular cabling design, multiple 12-V rails and support for both Crossfire and SLI multi-graphics setup. However, FSP’s product presentation was somewhat misleading, as the firm put the 80plus Gold logo right above this new PSU. In fact, it isn’t certified for Gold (no PSU has officially reached the Gold standard yet), but it was already submitted to 80plus.org. One last word on the packaging, which we aren’t sure it truly adds value: All Everest Pro 1200 PSUs are shipped in a small aluminum case. At least it looks really nice.

We found two more products on the FPS booth. The second discovery is the firm’s new notebook power supply, called the NB Plus. This is a compact and portable power supply, which can be used universally, for almost all notebook devices that require no more than 90 W maximum power input. The NB Plus can be set to output between 15 V and 21 V in 1-V increments. It will remember the last setting, so you don’t have to configure it every time you switch it on. A little LED display on the front will tell you the current output voltage.

As a nice addition, the NB Plus also comes with a USB port, which can be used to charge or operate USB devices without you having to switch the notebook on. Hence you could use this to charge a cell phone or an MP3 player. The NB Plus has a height of 16 mm and is smaller than high performance notebook power supply units.