Sign in with
Sign up | Sign in
System Builder Marathon, June 2012: System Value Compared
By ,
1. A Close Competition, Complements Of Tough Decisions

System Builder Marathon, June 2012: The Articles

Here are links to each of the four articles in this quarter’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 $2000 Performance PC
Day 2: The $1000 Enthusiast PC
Day 3: The $500 Gaming PC
Day 4: Performance And Value, Dissected

Introduction

In the PC world, the point of diminishing returns is where you throw increasing amounts of money at hardware for smaller performance increases. This happens for a number of reasons. To begin, top-binned components are some of the most scarce, so supply is limited, making them more expensive. Also, bottlenecks start surfacing more prominently at the bleeding edge, preventing the same great scaling we often see in the mid-range space. Unfortunately, there’s also a floor at which certain parts cannot be manufactured any less expensively while still retaining their core capabilities. When we start comparing prices to performance, we always end up finding a point in the middle that we call the best value.

With a budget the comes closest to that point of diminishing returns, builder Paul Henningsen’s low-cost machines usually end up at or near the top of our performance per dollar charts. At the same time, when we drill into the numbers his machines generate, an argument can be made that they're sometimes inadequate for the most discerning power users, despite their value proposition. Don Woligroski's efforts, on the other hand, involve spending more money, often result in a less impressive finish when we compare performance to cost, but are better able to satisfy more of our audience. That makes the $1000+ machine easier for us to recommend.

This month’s austerity measures undercut Paul’s build completely, forcing him to choose between creating a capable graphics-oriented platform over a better-balanced combination of parts. No doubt, that was a hard choice to make, but given his gaming focus, he clearly did the best he could with a $500 ceiling.

And with that compromise in mind, we're set up for one of the rare occasions where a machine other than the lowest-priced build could end up topping the value charts.

Q2 2012 $2000 PC Components
 $500 Gaming PC$1000 Enthusiast PC$2000 Performance PC
ProcessorIntel Celeron G530: 2.4 GHz, LGA 1155, 2 MB CacheIntel Core i5-2400: 3.1-3.4 GHz, LGA 1155, 6 MB CacheIntel Core i7-3770K: 3.5-3.9GHz, LGA 1155, 8 MB Cache
GraphicsECS NGT560TI-1GPI-F1 GeForce GTX 560 TiSapphire 11197-01-40G: Radeon HD 7970 (Factory O/C)
Asus GTX680-DC2T-2GD5: GeForce GTX 680 (Factory O/C)
MotherboardGigabyte GA-H61MA-D3V: LGA 1155, Intel H61 ExpressGigabyte P67X-UD3-B3: LGA 1155, Intel P67 ExpressASRock Z77 Extreme6: LGA 1155, Intel Z77 Express
MemoryPareema MD313C80809L2: DDR3-1333 C9, 2 GB x 2 (4 GB)Patriot AP38G1608U2K: DDR3-1600 C8, 2 GB x 2 (4 GB)G.Skill F3-1600C8D-8GAB: DDR3-1600 C8, 4 GB x 2 (8 GB)
System DriveWestern Digital WD3200AAKX: 320 GB, 7200 RPM HDDSeagate Barracuda ST3750525AS: 750 GB, 7200 RPM HDDMushkin MKNSSDCR120GB-MX: 120 GB, SATA 6Gb/s SSD
Storage DriveUses System DriveUses System DriveSeagate Barracuda Green ST2000DL003: 2 TB, 5900 RPM HDD
OpticalLG GH22NS90B: 22x DVD±R, 8x DVD±R DLLG GH22NS90B: 22x DVD±R, 8x DVD±R DLLite-On iHAS124-04: 24x DVD±R, 12x DVD±R DL
CaseRosewill R101-P-BKLogisys Optimus IIAntec Nine Hundred w/USB 3.0
PowerAntec VP-450: 450 W, ATX V2.3Corsair CX600 V2: 600 W, ATX12V V2.3, 80 PLUS-CertifiedSeasonic X750 Gold SS-750KM: ATX12V V2.3, 80 PLUS Gold
CPU CoolerIntel Boxed CoolerIntel Boxed CoolerZalman CNPS12X
Total Price$500 $1034 $1741


But a trio of factors combine to make a compelling case for our high-end build. First, I scored one of Nvidia’s hard-to-find GeForce GTX 680s for just a few dollars more than the Radeon HD 7970 used in Don's $1000 build. Second, the processor I picked accommodates overclocking, which is something the Tom's Hardware audience favors for adding value, in spite of the expensive cooling apparatus often required. And third, I decided to forgo most of the parts that haven’t contributed to the overall performance of previous builds, leaving the savings off of the balance sheet altogether. Could this be our first three-way performance-per-dollar tie?

2. Test Settings And Benchmarks

Test Hardware Configurations
 $500 Gaming PC$1000 Enthusiast PC$2000 Performance PC
Processor
(Overclock)
Intel Celeron G530: 2.4 GHz, Two Physical Cores, No O/CIntel Core i5-2400: 3.1 GHz, Four Physical Cores, O/C to 3.8 GHz, +0.1 VIntel Core i7-3770K: 3.50 GHz, Four Physical Cores, O/C to 4.60 GHz, 1.25 V
Graphics
(Overclock)
ECS NGT560TI-1GPI-F1: 823 MHz GPU, GDDR5-4008, O/C to 891 MHz GDDR5-4410Sapphire 11197-01-40G: 925 MHz GPU, GDDR5-5500, O/C to 1125 MHz GDDR5-6400 Asus GTX680-DC2T-2GD5: 1201 MHz GPU,  GDDR5-6008, O/C to 1262 MHz GDDR5-7204
Memory
(Overclock)
4 GB Pareema DDR3-1333 CAS 9-9-9-24, O/C at 1.60V to DDR3-1066 CL 6-6-6-144 GB Patriot DDR3-1600 CL 8-9-8-24, No O/C8 GB G.Skill DDR3-1600 CAS 8-8-8-24, O/C at 1.60 V to DDR3-2000 CL 8-10-9-16
Motherboard
(Overclock)
Gigabyte GA-H61MA-D3V: LGA 1155, Intel H61 Express, Stock BCLKGigabyte P67X-UD3-B3: LGA 1155, Intel P67 Express, Stock BCLKASRock Z77 Extreme6: LGA 1155, Intel Z77 Express, Stock BCLK
OpticalLG GH22NS90B 22x DVD±RLG GH22NS90B 22x DVD±RLite-On iHAS124-04 24x DVD±R
CaseRosewill R101-P-BKLogisys Optimus IIAntec Nine Hundred
CPU CoolerIntel Boxed CoolerIntel Boxed CoolerZalman CNPS12X
Hard DriveWestern Digital WD3200AAKX: 320 GB, 7200 RPM HDDSeagate Barracuda ST3750525AS: 750 GB, 7200 RPM HDDMushkin Chronos Deluxe MX: 120 GB, SATA 6Gb/s SSD
PowerAntec VP-450: 450 W, ATX V2.3Corsair CX600 V2: 600 W, ATX12V V2.3, 80 PLUS-CertifiedSeaSonic SS-750KM: 750 W, ATX12V, 80 PLUS Gold 
Software
OSMicrosoft Windows 7 Ultimate x64
GraphicsNvidia GeForce 296.10AMD Catalyst 12.4Nvidia GeForce 301.42
ChipsetIntel INF 9.2.3.1020


The inability to overclock his processor didn’t completely stop our $500 PC builder in his search for optimizing performance. He chose tighter memory timings and GPU overclocking to slightly boost his scores. Similarly, the $1000 machine builder, Don Woligroski, used his motherboard’s ability to increase his semi-locked CPU's Turbo Boost multipliers by 4x across the board, while also achieving a respectable graphics overclock.

The value competition will still be tough, however, as the 4.60 GHz CPU clock and incredible 1262 MHz GPU / GDDR5-7204 graphics clocks of the sub-$2000 machine is ready to take all comers on the performance side of the price-versus-performance equation.

Benchmark Configuration
3D Games
Battlefield 3Campaign Mode, "Going Hunting" 90-Seconds Fraps
Test Set 1: Medium Quality Defaults (No AA, 4x AF)
Test Set 2: Ultra Quality Defaults (4x AA, 16x AF)
DiRT 3V1.01, Run with -benchmark example_benchmark.xml
Test Set 1: High Quality Preset, No AA
Test Set 2: Ultra Quality Preset, 8x AA
Elder Scrolls V: SkyrimUpdate 1.5.26, Celedon Aethirborn Level 6, 25 Seconds Fraps
Test Set 1: DX11, High Details No AA, 8x AF, FXAA enabled
Test Set 2: DX11, Ultra Details, 8x AA, 16x AF, FXAA enabled
StarCraft IICustom map "Tom's Hardware Guide V2", 60 seconds Fraps
Test Set 1: High Details, High Quality
Test Set 2: Ultra Details, Extreme Quality
Audio/Video Encoding
iTunesVersion 10.4.1.10 x64: Audio CD (Terminator II SE), 53 minutes, default AAC format 
Lame MP3Version 3.98.3: Audio CD "Terminator II SE", 53 min, convert WAV to MP3 audio format, Command: -b 160 --nores (160 Kb/s)
HandBrake CLIVersion 0.95: "Big Buck Bunny" (720x480, 23.972 FPS) 5 Minutes, Audio: Dolby Digital, 48 000 Hz, Six-Channel, English, to Video: AVC Audio: AC3 Audio2: AAC (High Profile)
MainConcept ReferenceVersion: 2.2.0.5440: 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 CS5Version 12.1 x64: Filter 15.7 MB TIF Image: Radial Blur, Shape Blur, Median, Polar Coordinates
Autodesk 3ds Max 2012Version 14.0 x64: Space Flyby Mentalray, 248 Frames, 1440x1080
WinZipVersion 15.5 Pro: THG-Workload (650 MB) to ZIP, command line switches "-a -ez -p -r"
WinRARVersion 4.1: THG-Workload (650 MB) to RAR, command line switches "winrar a -r -m3"
7-ZipVersion 9.22: THG-Workload (650 MB) to .7z, command line switches "a -t7z -r -m0=LZMA2 -mx=5"
ABBYY FineReaderVersion 10.0.102.82: Read PDF save to Doc, Source: Political Economy (J. Broadhurst 1842) 111 Pages
Synthetic Benchmarks and Settings
3DMark 11Version: 1.0.1.0, Benchmark Only
PCMark 7Version: 1.0.4 x64, System, Productivity, Hard Disk Drive benchmarks
SiSoftware Sandra 2011Version 2011.10.17.80, CPU Test = CPU Arithmetic / MultiMedia, Memory Test = Bandwidth Benchmark
3. Benchmark Results: 3DMark And PCMark

The sub-$2000 PC offers up less than twice the performance of the $1000 build at less than twice the cost in 3DMark 11. It also scales to four times the performance of the $500 PC at less than four times the cost. Budget-builder Paul Henningsen should be very happy that these scores aren’t used in our value analysis, while I should be bummed that exceptional numbers won't be counted toward my overall finish.

PCMark is heavily weighted toward system drive performance, and our recent budget cuts prevented the use of an SSD in any machine except the most expensive build. Yet, this tests overall scores are also excluded from our value analysis, which focuses primarily on real-world performance tests and not synthetics.

These three charted profiles from PCMark’s storage benchmark are the closest we can get to real-world experience tests for hard drive performance, since most of our application starts are unbearably difficult to time accurately. These count for only 10% of the overall performance in our final analysis, and are the only synthetics in that analysis. Of course, the SSD wins.

4. Benchmark Results: SiSoftware Sandra

Sandra’s benchmark modules are particularly useful in the search for a system’s specific weaknesses, even though they don’t represent real-world performance.

These results don’t count towards our value analysis, but they do help us quantify extreme weakness in the $500 machine’s Celeron G530 CPU, and almost-excessive power in the big build’s overclocked Core i7-3770K.

For nearly two-thirds of the cost of the sub-$2000 PC’s memory, the $1000 build suffers half the capacity and less bandwidth. This could be the other side of that diminishing returns discussion.

5. Benchmark Results: Battlefield 3

Anyone who thinks that Battlefield 3's single-player campaign might be slightly CPU-bound needs only look at our Medium quality results to find differently. At the most CPU-dependent settings, the frequency-deficient $1000 build nearly catches the sub-$2000 PC.

Cranking up the details to the game’s Ultra preset gives the highly-overclocked $2000 PC very little breathing room, with the $1000 build hot on its tail. Moreover, the $500 machine proves adequate at 1920x1080, if only barely.

6. Benchmark Result: DiRT 3

We read suggestions as recently as last week that DiRT 3 is a GPU-bound game. However, our High quality preset results show how CPU-bottlenecked the game can be. It’s not until we crank the game up to Ultra details that our GPUs begin to struggle.

The sub-$2000 machine excels at 2560x1600, while the $500 PC merely survives at its 1920x1080 target. Surprisingly, the $1000 PC gains ground against the higher-cost build as details are increased, almost catching it at our highest test setting.

7. Benchmark Results: The Elder Scrolls V: Skyrim

The $2000 PC begins our Skyirm-based benchmark set with a big lead at its lowest setting, only to watch as the $1000 build catches up and finally passes it at Ultra details and 2560x1600.

Even though the $500 PC is CPU-bottlenecked throughout the entire Skyrim benchmark set, it still manages to survive to its Ultra detail 1920x1080 target.

8. Benchmark Results: StarCraft II

StarCraft II presents somewhat of an analysis problem for me, since it appears somewhat CPU-constrained at all but our lowest test resolution. Most games run into GPU-limits higher resolution.

Overclocking keeps the sub-$2000 PC on top throughout these tests, though the magnitude of that overclock may cause our readers to question whether some of its components were really worth the added cost.

As an example of why it doesn't always make sense to spend more, the $500 PC puts out 62.4 FPS at our highest StarCraft II detail levels and its target 1920x1080 resolution.

9. Benchmark Results: Audio And Video Encoding

The priciest machine’s superb overclock won it little favor in most gaming scenarios, but it shines brightly in both iTunes and Lame. These single-threaded applications would have been equally satisfied with an overclocked dual-core chip, if Intel hadn’t locked that option out of its newest processor models. We instead have to buy K-series SKUs, all of which are four- or six-core processors.

HandBrake and MainConcept demonstrate the true benefits of a multi-core processor, as the $500 PC’s lowly dual-core chip takes a beating. Both programs also respond eagerly to overclocking, giving the $2000 machine a substantial victory.

10. Benchmark Results: Productivity

Adobe Photoshop favors the $1000 PC by utilizing its four cores without showing much benefit to the more expensive machines Hyper-Threaded CPU. A four-bin overclock (400 MHz) allows the cheaper machine to nearly perform on par with the highest-end build's baseline. The $500 PC’s dual-core CPU, unfortunately, holds it back once again.

3ds Max generates slightly more predictable results, reflecting differences in both clock frequency and core count as we'd expect from a threaded app.

WinRAR and WinZip are the least-optimized for threading of our three compression-oriented programs, while 7-Zip appears best able to utilize all available resources.

Though the difference between the overclocked $2000 build’s workload time and the $1000-PC’s stock clock rate is almost 100%. The only CPU that appears truly inadequate is that of the $500 build. We can’t complain much, though, since the builder himself called this a $290 base system with a $210 graphics card added-in.

11. Energy And Efficiency

Perhaps the biggest surprise from our tests is how little power the sub-$2000 build consumes in comparison to its predecessor. The big question is how it will compare to the less expensive PCs, with their less potent hardware?

Our three builds line up on the power chart in the expected order, with none of the big jumps that would have occurred had we added SLI or CrossFire to any of them.

Big storage scores count for only 10% of the $2000 build’s roughly 50% average performance gain over its $1000 rival, and that’s certainly much larger than the power difference. Could all of the articles we’ve read on low-power computing be wrong? Could performance be the true path to greater efficiency?

Assuming that you’re replacing several slow systems with one larger unit, a single machine could give you greater efficiency. The overclocked $2000 PC performs nearly 2.5 times the work of the $500 baseline system, while consuming only 72% more power.

12. Crowning A Value Winner

The lowest-cost build usually wins our performance-per-dollar comparison, though we typically have to flag it with a caveat that most performance seekers probably won't be satisfied with its speed.

This time could be different, though, since a boost in the little machine's GPU performance came at a serious cost: its CPU budget. Might the $1000 system, or even (cough) my $2000 build win this race?

For the fist time in recent memory, the mid-priced build tops our value charts. The only configuration to fall behind is the $2000 at its stock settings; overclocking is required to justify its added expense. Fortunately, I picked an overclockable CPU, which responded well to tuning.

The $1000 PC performed so well and presented so much value that I'm almost ready to give its parts list a mass recommendation. The one shortfall holding me back is that its case was simply garbage. Fortunately,  a suitable replacement shouldn’t be very difficult to find near its $42 price.

The $1000 PC’s other shortcoming is one that, if fixed, could have boosted its value score. The unlocked Core i5-2500K would have added only $30 to the system, and a more adequate cooler another $20 or so. CPU overclocking is a time-proven way to bolster performance at minimal cost, so a roughly 5% budget increase might have returned a two to threefold increase in benchmark performance.

As for the $500 build's second-place value finish, even its builder admitted that the machine would be woefully inadequate for anything beyond the scope of gaming. While it’s hard for many of us to recommend a machine exclusively for that purpose, some folks do nothing but game and surf the Web on their PCs. For them, the entry-level system's limitations aren't debilitating at all.

Finally, there’s the matter of gaming at high resolutions. This is typically where the most expensive configuration shines. But the $1000 build’s Radeon HD 7970 performed amazingly well in spite of the system’s low memory capacity and moderate CPU frequency:

This chart almost shows that a high-end gaming-oriented PC is overkill (at least at the resolutions we're using for testing). We paid much more for a higher-end PC with lots of memory, but without a corresponding return on our investment. There’s also the expense of an SSD, which improves game loading times, but hasn't had much of an effect in-game since we stopped testing the original Crysis.

And so, the almost-excellent $1000 build almost gets our recommendation. With a couple of little alterations (and a slightly higher cost) we end up asking: would anyone like a $1100 Enthusiast PC?