System Builder Marathon, December 2011: The Articles
Here are links to each of the four articles in this month’s System Builder Marathon (we’ll update them as each story is published). And remember, these systems are all being given away at the end of the marathon.
To enter the giveaway, please fill out this SurveyGizmo form, and be sure to read the complete rules before entering!
Day 1: The $2400 Performance PC
Day 2: The $1200 Enthusiast PC
Day 3: The $600 Gaming PC
Day 4: Performance And Value, Dissected
Introduction
Although prices are up on just about everything (except memory), the real reason we increased the budget this quarter was that we grew weary of making sacrifices. Surely the most expensive PC had the most room for compromise, but in reality, nobody is going to give up big performance potential on a $2000 machine when an extra couple hundred bucks could help even more. For $300, we were able to fix some of the problems that plagued past builds.
The other $100 covered the recent hard drive price increase attributable to the horrible flooding in Thailand.

But what were those problems, specifically? And why did they cost so much to fix? Why the heck would we worry about hard drive prices in a machine powered by SSDs?
| December 2011 $2400 PC Components | ||
|---|---|---|
| Motherboard | MSI Z68A-GD55 (G3): LGA 1155, Intel Z68 Express | $170 |
| Graphics | PNY XLR8 VCGGTX580XPB-LC-CPU GeForce GTX 580 PNY XLR8 VCGGTX580XPB-LC GeForce GTX 580 (for SLI) | $1030 |
| Processor | Intel Core i7-2600K: 3.4-3.8 GHz, 8 MB Shared L3 Cache | $320 |
| Memory | Corsair CMZ8GX3M2A1600C8R: DDR3-1600 C8, 4 GB x2 (8 GB) | $75 |
| System Drive | Crucial M4 CT128M4SSD2: 128 GB, SATA 6Gb/s SSD | $210 |
| Storage Drive | WD Caviar Green WD15EARS: 1.5 TB, 7200 RPM Hard Drive | $180 |
| Optical | LG WH12LS38 BD-RE: 12x BD-R, 16x DVD±R | $70 |
| Case | Fractal Design Arc Midi FD-CA-ARC-BL | $110 |
| Power | Mushkin Joule MKNPSJL1000: 1000 W, ATX12V v2.31, 80 PLUS Gold | $160 |
| Heat Sink | PNY XLR8 / Asetek Integrated Liquid Cooling | |
| Total Cost: | $2325 | |
The first problem suffered by our previous build, an overheated voltage regulator that severely limited overclocking, gets addressed by swapping out the motherboard. The second problem, a pair of graphics cards with built-in overclocking limits, get exchanged for a pair known to overclock more readily. And the third issue (one we didn’t worry all that much about), a lack of TRIM support in our RAID 0 SSD array, is eliminated by picking a single drive with the combined capacity of the previous pair. A real-world user still needs more space for movies, music, and other user files, but hard drive shortages force us to pay $100 more for 500 GB less compared to the last time around. It's a bad time to have to buy a hard drive, for sure.
Motherboard: MSI Z68A-GD55 (G3)
Overclocking is the biggest performance upgrade any builder can add to a PC, even though it comes at some risk. Our previous $2000 machine used a motherboard that had been awarded for its low-cost overclocking capability, yet our retail sample overheated when it was pushed past 1.35 V. In the famous words of a not-so-famous politician, “Oops.”

Read Customer Reviews of MSI's Z68A-GD55 (G3)
We were hoping to come back with the runner up from one of our previous motherboard round-ups to see if it would out-overclock the former winner in a real-world test. While MSI’s Z68A-GD55 was no longer available in its original form, the G3 version was priced only $10 higher.
Both versions include heat sinks on all CPU voltage regulator phases, and the extra $10 gets buyers a set of third-generation PCIe switches to keep their next-generation processors and graphics cards working at peak performance. That last part assumes that, between now and launch day, Intel won’t make its Ivy Bridge-based processors incompatible with existing platforms.
Processor: Intel Core i7-2600K
We ordered our System Builder Marathon machines prior to Intel’s Sandy Bridge-E launch, secure in the knowledge that at least the flagship would cost more than our $2400 budget could bear. Chris Angelini’s recent Core i7-3930K review did little to change that opinion, since such a small upgrade to our desktop-oriented benchmarks would have taken hundreds of dollars away from other components.

Read Customer Reviews of Intel's Core i7-2600K
That left Intel’s Core i7-2600K and Core i5-2500K as the only choices for top performance (Ed.: assuming the -2700K wasn't worth the increase either, right?). We’re counting on this processor’s legendary overclocking capability to push its value, even though the added benefit Hyper-Threading alone is somewhat questionable.
RAM: Corsair CMZ8GX3M2A1600C8R
Also known as Vengeance DDR3-1600 CAS 8, this is the same memory that won our recent 8 GB value-oriented round-up.

Read Customer Reviews of Corsair's CMZ8GX3M2A1600C8R Memory Kit
Differences between AMD and Intel memory controllers mean that we won’t see the exact results from that round-up, but we’re still certain this memory will give us excellent timings at low cost.
Graphics Cards: PNY Liquid Cooled GTX 580s in SLI
We were terribly disappointed by the GPU overclocking capabilities of our previous $2000 build, but should have expected what we found, given then we used cards from a company famous for binning its top cards to sell as more expensive models.
Those cards were purchased for their exceptional warranty, rather than their overclocking headroom. This time, however, we put performance first.

Read Customer Reviews of PNY's Liquid-Cooled GeForce GTX 580s
With proven overclocking capability, PNY’s liquid cooled GeForce GTX 580s came to mind when we were putting together a system specifically designed to highlight overclocked value. Its XLR8 VCGGTX580XPB-LC includes a single-thickness radiator and single fan (Antec H2O 620 equivalent) with a custom GPU-mounted pump, while its VCGGTX580XPB-LC-CPU uses the double-thick two-fan radiator and CPU-mounted water pump found in Antec’s H2O 920. Yes, that's a graphics card that can cool the CPU!
Case: Fractal Design Arc Midi
Pleasing results from our recent MicroATX Gaming Build round-up compelled us to take a closer look at one of the contender's bigger brother, Fractal Design’s Arc Midi.
Read Customer Reviews of Fractal Design's Arc Midi
Designed to hold a double-120 mm-fan radiator on top, our pair of single-fan radiators fit using alternative placement.
Power: Mushkin Joule 1000 W Gold
We wanted something a little bigger and more-efficient than our previously-chosen SeaSonic 850W Silver unit and found an excellent price in Mushkin’s MKNPSJL1000.

Read Customer Reviews of Mushkin's Joule 1000 W
Though Mushkin’s original Joule power supply earned a somewhat-mixed reputation, support for the later Gold version is nearly unanimous among its reviewers. Distrust for its predecessor might have driven down the price of the improved product, but we have enough faith in the newer 80 PLUS Gold-rated part to treat its low cost as a value bonus.

System Drive: Crucial m4 128GB SSD
Fellow editor Andrew Ku finally talked me into ditching my striped pair of 64 GB SSDs in favor of a single, faster drive. Crucial’s 128 GB m4 topped the list of somewhat-affordable models based on its excellent performance reputation.

Read Customer Reviews of Crucial's m4 128 GB
Storage Drive: Western Digital Caviar Green 1.5 TB
Pining for the days when a 2 TB drive cost only $80, we spent $180 dollars to get a big drop in capacity as an unfortunate result of the recent production issues affecting the mechanical drive industry.

Read Customer Reviews of Western Digital's Caviar Green 1.5 TB
Anyone who questions the money spent on added storage is probably missing the point that this is not a pure gaming system, but instead a general performance machine that plays games well.
Optical Drive: LG WH12LS38 BD-RE
For a mere $70, LG’s basic BD-RE drive allows users to back up 50 GB of data at 12x speed, though single-layer media still provides the best cost-to-capacity ratio.

Read Customer Reviews of LG's WH12LS38 BD-RE Drive
While this drive could also be used to view Blu-ray movies, it doesn’t include the software needed to do so. We’re usually pleasantly surprised to find free software with our bare drives, and would have probably spent an extra $10 on a different model if we knew we weren’t going to get it with this one.
Fractal’s Arc Midi might be designed for a dual fan top-mounted radiator, but that doesn’t mean two single-fan units can fit. Though it might have been possible to position these side-by-side within the empty case, tanks on the top and bottom would have prevented motherboard and side panel installation. Generally, we like our PCs to include motherboards.

Instead, one radiator needs to be installed on the top panel and the other on the rear panel. Both are fitted with their fans blowing out of the case as exhaust, pulling heat away from memory and voltage regulator components.

The snap-off front panel comes with a single intake fan. Since our radiators have their own fans, the original rear-panel exhaust fan is snapped into the spare front-panel mount as an extra intake.

The Arc Midi’s sliding trays each hold a 3.5” drive on vibration-dampening grommets using shoulder screws, or a 2.5” drive without grommets using standard screws. Since the 2.5” drives most often found in full-sized systems are SSDs, the lack of noise dampening for this form factor is acceptable.

The top-mounted radiator is placed in the case's forward mounting position to make room for the thicker radiator on the rear panel. Had the thicker radiator been mounted on top, it would have prevented memory installation. We also like our PCs to include memory.

The finished system is particularly attractive to users who don’t like the distraction of gaudy lights.
The motherboard of our previous $2000 build became unstable after prolonged heavy use when overclocked. Relatively low CPU temperatures tend to indicate a voltage regulator that has been pushed beyond its optimal thermal range. The previous build was able to maintain stable voltage up to the point where it got hot, crossing its thermal threshold at around 1.35 V with our Core i7-2600K CPU maxed out.
The problem was that our CPU needed a little more voltage to get a decent overclock, and we couldn't give it anything more without asking too much from the platform.

MSI’s Z68A-GD55 lost our motherboard comparison partly because its voltage regulator was always saggy. And yet, the same board had superior voltage regulator cooling. The practical impact of switching to this board was that we needed to set 1.40 V to approach our target 1.38 V, but we got there without encountering heat issues. Mission accomplished.

Our memory is known to overclock past DDR3-1866. Unfortunately, its default timings wouldn’t permit that setting on MSI's Z68A-GD55. The board didn’t automatically loosen timings when we specified higher data rates in the same way as many competing products, and manually choosing higher latencies still wouldn't permit DDR3-1866 operation. To get a slight boost in memory bandwidth, we chose 45 x 102.2 MHz, rather than 46 x 100 MHz, to reach a CPU clock of 4.6 GHz.

With our memory only 2.2% above stock, we were instead forced to optimize latencies for added performance in the system’s overclocked state.

Even at the board’s 1.40 V setting, voltage dropped to 1.368 V when the overclocked CPU was loaded with eight threads of Prime95.

A combination of higher-grade chips and better cooling allowed our GeForce GTX 580s to overclock eagerly, even though we didn’t touch their voltage levels.

Take this warning to heart: PNY’s liquid-cooled cards should only be used with custom fan speeds, even when they're not overclocked. That’s because the memory and voltage regulator are still fan-cooled, and the fan is controlled by GPU temperature. Our tests indicate that PNY doesn't alter the air-cooled-card’s fan profile, and the lower GPU temperatures enabled by liquid cooling result in the fan slowing down to a point that the card can overheat.

An ideal solution would be to move the thermal sensor that controls the fan to the card’s voltage regulator. However, the reference GeForce GTX 580s that PNY uses aren't designed to allow this. Instead, we used MSI Afterburner to set a custom fan profile keyed to the moderate temperatures these liquid-cooled GPUs experience under stress. Afterburner must be running for its fan controls to work, so we set the “Start with Windows” and “Start minimized” options under its Properties menu.

| Test Hardware Configurations | ||
|---|---|---|
| Q4 $2400 PC | Q3 $2000 PC | |
| Motherboard (Overclock) | MSI Z68A-GD55 (G3) LGA 1155, Intel Z68 Express O/C at 102 MHz BCLK | Gigbyte Z68XP-UD3 LGA 1155, Intel Z68 Express O/C at 102 MHz BCLK |
| Processor (Overclock) | Intel Core i7-2600K 3.40 GHz, Four Physical Cores O/C to to 4.60 GHz, 1.37 V | Intel Core i7-2600K 3.40 GHz, Four Physical Cores O/C to to 4.48 GHz, 1.36 V |
| Memory (Overclock) | 8 GB Corsair DDR3-1600 CAS 8-8-8-24, O/C at 1.60 V to DDR3-1636 CL 7-8-7-12 | 8 GB G.Skill DDR3-1866 CAS 9-10-9-28, O/C at 1.60 V to DDR3-1901 CL 9-10-9-16 |
| Graphics (Overclock) | 2 x PNY GeForce GTX 580: 857 MHz GPU, GDDR5-4212 O/C to 951 MHz GDDR5-4312 | 2 x Evga 015-P3-1580-AR: 772 MHz GPU, GDDR5-4008 No O/C |
| Optical | LG WH12LS38 BD-RE: 12x BD-R, 16x DVD±R | Lite-On iHAS224-06 24x/8x DVD±R/RW |
| Case | Fractal Design Arc Midi | Antec Three Hundred Illusion |
| CPU Cooler | PNY XLR8 Integrated Liquid Cooling | Cooler Master Hyper 212 Plus |
| Hard Drive | Crucial M4 128 GB, SATA 6Gb/s SSD | Adata S511 120 GB, SATA 6Gb/s SSD |
| Power | Mushkin Joule MKNPSJL1000: 1000 W, ATX12V v2.31, 80 PLUS Gold | Seasonic SS-850HT: 850W, ATX12V v2.31, 80 PLUS Silver |
| Software | ||
| OS | Microsoft Windows 7 Ultimate x64 | |
| Graphics | Nvidia GeForce 285.62 | Nvidia GeForce 280.26 |
| Chipset | Intel INF 9.2.0.1030 | |
| Benchmark Configuration | |
|---|---|
| 3D Games | |
| Crysis | Patch 1.2.1, DirectX 10, 64-bit executable, benchmark tool Test Set 1: High Quality, No AA Test Set 2: Very High Quality, 8x AA |
| F1 2010 | V1.01, Run with -benchmark example_benchmark.xml Test Set 1: High Quality Preset, No AA Test Set 2: Ultra Quality Preset, 8x AA |
| Just Cause 2 | Version 1.0.0.2, Built-In Benchmark "Concrete Jungle" Test Set 1: Medium Details, No AA, 8x AF Test Set 2: Highest Details, 8x AA, 16x AF |
| Metro 2033 | Full Game, Built-In Benchmark, "Frontline" Scene Test Set 1: DX11, High, AAA, 4x AF, No PhysX, No DoF Test Set 2: DX11, Very High, 4x AA, 16x AF, No PhysX, DoF On |
| Audio/Video Encoding | |
| iTunes | Version 9.0.3.15 x64: Audio CD (Terminator II SE), 53 minutes, default AAC format |
| Lame MP3 | Version 3.98.3: Audio CD "Terminator II SE", 53 min, convert WAV to MP3 audio format, Command: -b 160 --nores (160 Kb/s) |
| HandBrake CLI | Version 0.94: "Big Buck Bunny" (720x480, 23.972 FPS)Five Minutes, Audio: Dolby Digital, 48000 Hz, Six-Channel, English, to Video: AVC Audio: AC3 Audio2: AAC (High Profile) |
| MainConcept Reference | Version: 2.0.0.1555: MPEG-2 to H.264, MainConcept H.264/AVC Codec, 28 sec HDTV 1920x1080 (MPEG-2), Audio: MPEG-2 (44.1 kHz, Two-Channel, 16-Bit, 224 Kb/s), Codec: H.264 Pro, Mode: PAL 50i (25 FPS), Profile: H.264 BD HDMV |
| Productivity | |
| Adobe Photoshop CS5 | Version 12.0 x64: Filter 15.7 MB TIF Image: Radial Blur, Shape Blur, Median, Polar Coordinates |
| Autodesk 3ds Max 2010 | Version 12.0 x64: Space Flyby Mentalray, 248 Frames, 1440x1080 |
| WinZip | Version 14.0 Pro: THG-Workload (464 MB) to ZIP, command line switches "-a -ez -p -r" |
| WinRAR | Version 4.0 Beta 4: THG-Workload (464 MB) to RAR, command line switches "winrar a -r -m3" |
| 7-Zip | Version 9.2: THG-Workload (464 MB) to .7z, command line switches "a -t7z -r -m0=LZMA2 -mx=5" |
| ABBYY FineReader | Version 10.0.102.82: Read PDF save to Doc, Source: Political Economy (J. Broadhurst 1842) 111 Pages |
| Synthetic Benchmarks and Settings | |
| 3DMark 11 | Version: 1.0.1.0, Benchmark Only |
| PCMark 7 | Version: 1.0.4 x64, System, Productivity, Hard Disk Drive benchmarks |
| SiSoftware Sandra 2011 | Version 2011.1.17.15, CPU Test = CPU Arithmetic / MultiMedia, Memory Test = Bandwidth Benchmark |
Factory-overclocked graphics processors put the new system ahead of our previous build in 3DMark at stock settings. Liquid cooling on performance-binned GPUs give the same system an enormous lead when both are overclocked.
Moreover, superior overclocking keeps the new system in the lead in PCMark, though it falters at stock settings. The previous system’s motherboard had a small overclock when it was set up at its default clocks, which will probably show up in our CPU tests.


Crucial's m4 beats the previous build's S511s in PCMark’s storage suite, though the new system’s hard drive scores actually go down when overclocked.

Sandra Arithmetic tells a sad tale about the new system’s motherboard at its default settings. Since both builds sport the same CPU, a performance difference this large can only be attributed to differences in processor multiplier.
Of course, we're also assuming that anyone who actually uses this machine is going to overclock it to similar effect. So, the victory there is reassuring.


Finding similar losses in Sandra Multimedia, we checked to see how well Intel Turbo Boost technology was working on MSI’s board. Its top 38x almost never showed itself, and only then we saw it under moderate single-core loads. Meanwhile, it stuck at 34x in most threaded programs.

Disappointed by the so-called high-end memory from our previous build, we tried running overclockable lower-cost parts in the new PC, only to be foiled by an MSI firmware that wasn’t overclock-friendly to DRAM. Chalk it up to round two of disappointment, resulting in lower memory bandwidth numbers.
Somewhat poorer CPU performance hands the new system a performance deficit in our Crysis High Quality test charts, at least before it’s overclocked.
The platform's overclocked performance lead increases at higher resolution, where the graphics cards are responsible for more of the outcome.


However, our new system’s factory-overclocked graphics cards don’t help its poor baseline performance until we reach 2560x1600, even at Crysis’ Very High quality settings. We're happy to report that Crysis will be retired starting next quarter, hopefully putting an end to hard-to-decipher results from this always-demanding, but somewhat-dicey test.

F1 2010 appears to favor memory performance using its High Quality preset, with the new system’s poor memory overclock hampering frame rates to some degree.

With a larger portion of the load shifted onto the graphics cards, F1 2010's Ultra Quality setting reflects our Crysis performance findings.
Just Cause 2 appears to be CPU-bottlenecked at medium details, and our new motherboard appears to be seriously hindering our CPU at baseline settings.
Fortunately, that same motherboard also offers a superior overclock, and that's where we'd expect a majority of enthusiasts to run their own tweaked configurations.


Even at higher details with more of the load shifted to GPU performance, the $2400 PC’s lower CPU performance holds it back. Once again, overclocking superiority turns it into a different class of machine.


Metro 2033 finally presents us with a gaming scenario where most settings benefit from the new build’s higher GPU performance, even prior to overclocking.
Apple iTunes transcoding performance is almost perfectly pinned to CPU frequency, and the new system has a noticeable deficit at default frequencies.
It wins when overclocked, though, which is what counts the most in our book.


Lame continues to paint the picture of a motherboard that doesn’t allow its CPU to shine in the new system at default settings, since both machines have the same CPU. Fortunately, the MSI board in our new configuration still overclocks better.


The MSI board’s baseline performance looks even worse in video transcoding, though its higher overclock compensates.
Adobe Photoshop again shows our new system's poor baseline and excellent overclocked performance.
All the way through ABBY FineReader, our new system needs its overclocking advantage to offset poor baseline results. When we're judging value, that's what matters most.




Liquid cooling pumps are somewhat of a power hog, so we expected our new system to consume a little more at idle compared to the previous build.
Overclocking requires much more power, so its big consumption numbers are par for its superior overclock.

Liquid cooling usually drops temperatures, though sharing a small radiator with the GPU doesn’t help our CPU temperatures compared to the previous machine.

Efficiency compares work to energy, so we first compile a performance chart to gauge work per unit of time in comparison to the previous system’s baseline. After this, we compare the performance ranking to the power ranking.

The current overclocked PC is around 9% faster in games than its tweaked predecessor, but it also consumes more than 9% additional power. Keeping the former system’s standard speed as our baseline, we calculate our efficiency chart around it. Since the baseline is 100% (and yet nothing is 100% efficient), we then subtract 100% from the efficiency chart results to show only how much more or less efficient each system is by comparison.

First calculating overall performance at 10% for storage performance and 30% each for Gaming, Encoding, and Productivity scores, our power-per-performance chart shows that the previous build was more efficient in spite of its lower performance. Overclocking boosted its efficiency by increasing performance by more than power consumption.
A big boost in CPU voltage required by the new system increased its power consumption by a greater percentage than performance, resulting in less efficiency.
Our new PC was more expensive and it performed better. But was it worth the increase in price? Using the same performance formula from our efficiency calculations (10% storage, 30% each Games/Encoding/Productivity), we compared the overall performance of each system to its build cost using the former non-overclocked build as our baseline.

Most of the extra money in this month’s budget was spent to promote improved overclocking, and the new system did overclock better. Its performance increase was far smaller than its price increase, however. Furthermore, if you're not an overclocker, then that money would go to waste on unrealized potential.
At the same time, most of that extra money was spent in an effort to boost graphics performance, so perhaps a chart comparing each system’s highest-resolution gaming performance to its price will be the saving grace for our latest build?

And there’s what we wanted to see. The only problem with our new build is that it doesn't achieve similar benefits at lower game settings or in other applications. If you're spending big money on a pair of GeForce GTX 580s, though, we'd recommend using a display (or displays) befitting such a powerful combination.
The reason we put such a large percentage of our budget into graphics is that it accounts for 30% of our overall performance score, and there are few other places where big gains can be achieved. We are after all using the second-best LGA 1155 CPU at a fairly high overclock, and even fewer of our programs could benefit from the hundreds of dollars it would cost to upgrade to a Core i7-3930K. In this instance, the only buyers who can justify the added expensive of liquid-cooled graphics cards are the ones who value gaming supremacy above every other task they plan to throw at their new PC.

