System Builder Marathon, Q3 2014: 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 Budget Gaming PC
Day 2: Our Mainstream Enthusiast System
Day 3: The $1600 High-End Build
Day 4: Performance And Value, Dissected

Our latest round of System Builder Marathon machines saw Paul and Don chasing bigger overclocks while I simply tried to fix mine. Purchased just before Intel launched Devil's Canyon, my machine last quarter was stuck with a mere 4.2 GHz CPU overclock that required a massive 1.28 V to reach. Lacking the Haswell update's cooling advantage, my -4770K appeared to be nothing more than a reject, cast off from Intel’s binning process as the company began stockpiling anything resembling a good die for its next new model. Or maybe it was just bad luck-of-the-draw.
Paul switched his $500 PC to Intel’s low-cost overclocking CPU, the Pentium G3258, after noting a new way to use cheap boards with unlocked CPUs.
Don took advantage of a long-standing $75 discount on Zotac’s factory-overclocked GeForce GTX 770, putting any savings on his $1000 PC towards a larger CPU cooler.
Meanwhile, I avoided the binning tomfoolery altogether by ordering its flagship Haswell-based Core i7-4790K, using a recent motherboard price drop to offset the CPU upcharge.
But wait, didn’t we call these $600, $1300, and $1600 builds? In theory, we’re supposed to have $500, $1000, and $1500 to cover mandatory hardware, and negotiations with all three builders yielded $100, $200, and $300 for stuff that wouldn’t be needed to make the system operational (the platform).
That leads to $600, $1200, and $1800 budgets including the operating system, case, optical drive, and accessories. Paul can’t fit anything more than the OS into his $100, so he refers to his platform budget at $450. Don overspends, so he just changed the name of his $1200 machine to $1300. And though more money was available to me, I’m still trying to fit all of my hardware into a 3x multiple of Paul’s total hardware budget.
| Q3 2014 System Builder Marathon Components | |||
|---|---|---|---|
| Q3 $600 PC | Q2 $1300 PC | Q3 $1600 PC | |
| Processor | Intel Pentium G3258 3.2 GHz, no Turbo Boost Two Cores, 3 MB L3 | Intel Core i5-4690K: 3.5 GHz-3.9 GHz Four Cores, 6 MB L3 | Intel Core i7-4790K: 4.0GHz-4.4GHz Four Cores, 8 MB L3 |
| Graphics | Sapphire Dual-X 100365L 2 GB Radeon R9 270 | Zotac AMP! Superclocked ZT-70303-10P 3 GB GeForce GTX 770 | PowerColor PCS+ AXR9 290X 4 GB Radeon R9 290X |
| Motherboard | MSI H81M-P33: LGA 1150, Intel H81 | ASRock Z97 Killer: LGA1150, Intel Z97 | MSI Z97 Gaming 5: LGA 1150, Intel Z97 |
| Memory | Team Dark TDBD38G1600HC9DC01 DDR3-1600 C9, 8 GB | G.Skill Trident F3-2400C10D-8GTD DDR3-2400 C10, 8 GB | G.Skill Ripjaws X F3-14900CL8D-8GBXM DDR3-1600 C8, 8 GB |
| System Drive | WD Blue WD10EZEX: 1 TB, SATA 6Gb/s HDD | ASP920SS3-128GM-C 128 GB SATA 6Gb/s SSD | Plextor M6S PX-256M6S: 256 GB SATA 6Gb/s SSD |
| Power | Antec VP-450: 450 W Non-Modular No Efficiency Rating | IN WIN GreenMe 650 650 W Non-Modular 80 PLUS Bronze | EVGA Supernova 750 B2: 750 W Semi-Modular 80 PLUS Bronze |
| CPU Cooler | Intel Boxed CPU Cooler | Noctua NH-D14 | Phanteks PH-TC14PE |
| Platform | $455 | $946 | $1,397 |
| Storage Drive | Uses System Drive | WD Blue WD10EZEX: 1 TB, SATA 6Gb/s HDD | WD Blue WD10EZEX: 1 TB, SATA 6Gb/s HDD |
| Optical | LG GH24NSB0B: 24x DVD±R, 48x CD-R | Asus DRW-24B1ST: 24x DVD±R, 48x CD-R | LG GH24NSB0B: 24x DVD±R, 48x CD-R |
| Case | Rosewill Challenger | Cooler Master HAF XM RC-922XM-KKN1 | Enermax Ostrog GT ECA3280A-BR |
| Total HW | $523 | $1146 | $1535 |
| OS | Windows 8.1 x64 OEM | ||
| Total Price | $623 | $1246 | $1635 |
Though I blamed a lackluster CPU sample for my previous overclocking woes, I wanted to remove all doubt from your minds concerning this quarter’s build. Choosing a cheaper case allowed me to spend more on CPU cooling. Don had the same idea, but chose to add the cost of a similar cooler on top of his budget. He did keep the total cost below $1250 though, so I still would have probably called it a slightly over-budget $1200 PC rather than pretend its budget was higher. Then again, I’m treating my $1800 budget as if it were $1600…

Anyone familiar with my motherboard reviews or even my Q3 overclocking comments knows why I keep Haswell cores under 1.30 V. Longevity is why I gave up at 4.60 GHz rather than shoot for the moon at 4.7 GHz at 1.31 volts. Some enthusiasts, on the other hand, have been lucky enough to keep Haswell-based CPUs alive for extended periods at 1.35 volts, and anyone willing to saw-off a few memory fins probably isn’t worried about something so trivial.
With full specs including base height (for DRAM clearance) plus a full page of installation notes in our 2011 NH-D14 review, informed readers need not face these challenges.
| Test Hardware Configurations | |||
|---|---|---|---|
| Q3 $600 Gaming PC | Q2 $1300 Enthusiast PC | Q3 $1600 Performance PC | |
| Processor (Overclock) | Intel Pentium G3258: 3.20 GHz, Two Physical Cores O/C to 4.10 GHz, 1.24 V | Intel Core i7-4690K: 3.5 - 3.90 GHz, Four Physical Cores O/C to 4.60 GHz, 1.335 V | Intel Core i7-4790K: 4.0 - 4.40 GHz, Four Physical Cores O/C to 4.60 GHz, 1.25 V |
| Graphics (Overclock) | Sapphire R9 270: 945 MHz GPU, GDDR5-5600 O/C to 1050 MHz, GDDR5-6000 | Zotac GTX 770: 1202 MHz GPU, GDDR5-7200 O/C to 1250 MHz, GDDR5-8000 | PowerColor 290X: 1050 MHz GPU, GDDR5-5400 O/C to 1082 MHz, GDDR5-5600 |
| Memory (Overclock) | 8 GB Team DDR3-1600 CAS 9-9-9-24, O/C to DDR3-1333 CL 7-7-7-21, 1.55 V | 8 GB G.Skill DDR3-2400 CAS 10-12-12-31, XMP Defaults (1.65 V) | 8 GB G.Skill DDR3-1866 CAS 8-9-9-24, O/C to DDR3-2400 CL 10-12-12-28, 1.6 V |
| Motherboard (Overclock) | MSI H81M-P33: LGA 1150, Intel H81 Express Stock 100 MHz BCLK | ASRock Z97 Killer: LGA 1150, Intel Z97 Express Stock 100 MHz BCLK | MSI Z97 Gaming 5: LGA 1150, Intel Z97 Express Stock 100 MHz BCLK |
| Case | Rosewill Challenger | Cooler Master HAF XM | Enermax Ostrog GT |
| CPU Cooler | Intel Boxed CPU Cooler | Noctua NH-D14 | Phanteks PH-TC14PE 140 mm |
| Hard Drive | WD Blue WD10EZEX 1 TB SATA 6Gb/s HDD | Adata Premier Pro SP920 128 GB SATA 6Gb/s SSD | Plextor M6S PX-256M6S 256 GB SATA 6Gb/s SSD |
| Power | Antec VP-450: 450 W Non-Modular, No Efficiency Rating | In Win GreenMe 650: 650 W Non-Modular, 80 PLUS Bronze | EVGA SuperNova 750 B2: 750 W, 80 PLUS Bronze |
| Software | |||
| OS | Microsoft Windows 8 Pro x64 | ||
| Graphics | AMD Catalyst 14.4 | Nvidia GeForce 344.11 | AMD Catalyst 14.4 |
| Chipset | Intel INF 9.4.0.1017 | Intel INF 9.4.0.1026 | Intel INF 9.4.0.1026 |
| Benchmark Configuration | |
|---|---|
| 3D Games | |
| Battlefield 4 | Version 1.0.0.1, DirectX 11, 100-Sec. Fraps "Tashgar" Test Set 1: High Quality Preset, No AA, 4x AF, SSAO Test Set 2: Ultra Quality Preset, 4x MSAA, 16x AF, HBAO |
| Grid 2 | Steam Version, In-Game Test Test Set 1: High Quality Preset, No AA Test Set 2: Ultra Quality Preset, 8x AA |
| Metro: Last Light | Steam version, Built-In Benchmark, "Frontline" Scene Test Set 1: DX11, Med Quality, 4x AF, Low Blur, No SSAA, No Tesselation, No PhysX Test Set 2: DX11, High Quality, 16x AF, Normal Blur, SSAA, Tesselation Normal, No PhysX |
| Far Cry 3 | V. 1.04, DirectX 11, 50-sec. Fraps "Amanaki Outpost" Test Set 1: High Quality, No AA, Standard ATC., SSAO Test Set 2: Ultra Quality, 4x MSAA, Enhanced ATC, HDAO |
| Adobe Creative Suite | |
| Adobe After Effects CC | Version 12.0.0.404: Create Video that includes three streams, 210 frames, Render Multiple Frames Simultaneously |
| Adobe Photoshop CC | Version 14.0 x64: Filter 15.7 MB TIF Image: Radial Blur, Shape Blur, Median, Polar Coordinates |
| Adobe Premiere Pro CC | Version 7.0.0 (342), 6.61 GB MXF Project to H.264 to H.264 Blu-ray, Output 1920x1080, Maximum Quality |
| Audio/Video Encoding | |
| iTunes | Version 11.0.4.4 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: 0.99: Video from Canon EOS 7D (1920x1080, 25 FPS) 1 Minutes 22 Seconds Audio: PCM-S16, 48,000 Hz, Two-Channel, to Video: AVC1 Audio: AAC (High Profile) |
| TotalCode Studio 2.5 | Version: 2.5.0.10677: MPEG-2 to H.264, MainConcept H.264/AVC Codec, 28 sec HDTV 1920x1080 (MPEG-2), Audio: MPEG-2 (44.1 kHz, 2 Channel, 16-Bit, 224 Kb/s), Codec: H.264 Pro, Mode: PAL 50i (25 FPS), Profile: H.264 BD HDMV |
| Productivity | |
| ABBYY FineReader | Version 10.0.102.95: Read PDF save to Doc, Source: Political Economy (J. Broadhurst 1842) 111 Pages |
| Adobe Acrobat 11 | Version 11.0.0.379: Print PDF from 115 Page PowerPoint, 128-bit RC4 Encryption |
| Autodesk 3ds Max 2013 | Version 15.0 x64: Space Flyby Mentalray, 248 Frames, 1440x1080 |
| Blender | Version: 2.68A, Cycles Engine, Syntax blender -b thg.blend -f 1, 1920x1080, 8x Anti-Aliasing, Render THG.blend frame 1 |
| Visual Studio 2010 | Version 10.0, Compile Google Chrome, Scripted |
| File Compression | |
| WinZip | Version 18.0 Pro: THG-Workload (1.3 GB) to ZIP, command line switches "-a -ez -p -r" |
| WinRAR | Version 5.0: THG-Workload (1.3 GB) to RAR, command line switches "winrar a -r -m3" |
| 7-Zip | Version 9.30 alpha (64-bit): THG-Workload (1.3 GB) to .7z, command line switches "a -t7z -r -m0=LZMA2 -mx=5" |
| Synthetic Benchmarks and Settings | |
| 3DMark Professional | Version: 1.2.250.0 (64-bit), Fire Strike Benchmark |
| PCMark 8 | Version: 1.0.0 x64, Full Test |
| SiSoftware Sandra | Version 2014.02.20.10, CPU Test = CPU Arithmetic / Multimedia / Cryptography, Memory Bandwidth Benchmarks |
3DMark shows progressive CPU and GPU scaling on all three of our machines, with the system that costs nearly 3x as much producing a little over twice the performance of Paul’s $600 PC baseline.
PCMark scaling is a little more rounded, as the $600 PC’s mechanical hard drive falls into a lower performance class compared to the SSDs of $1300 and $1600 systems.
Even though Don’s $1300 machine overclocked as well as my $1600 PC, Sandra's Arithmetic module reports that his overclock performance is less than my baseline. Since both Haswell-based processors have the same number of physical cores, I can only credit Intel's Hyper-Threading technology and a little extra last-level cache for allowing my CPU to outperform by such a wide margin.
Sandra's Cryptography routine shows closer performance scaling between the two quad-core PCs, while Paul’s little dual-core chip suffers from Intel’s disabling of various high-end features (including much of its cache and AES-NI).
Tired of my good-natured jabs over his previous memory benchmarks, Don Woligroski used DDR3-2400 to prove his machine’s metal in Sandra's Memory Bandwidth test, achieving 28 GB/s. Conversely, my lower-profile DDR3-1866 modules reach only 30 GB/s when manually configured to the same frequency and latency.
Everyone wins in Battlefied 4 at Medium Quality. Paul lacks the performance potential to test at 5760x1080, but the target of his $600 PC is only 1920x1080. Reaching 60 FPS at 4800x900 without overclocking makes Paul the smart shopper.
Mr. Henningsen's $600 system drops out of the triple-wide race at Battlefield's Ultra quality preset, but still hits its 1920x1080 target. Don’s $1300 PC might be playable at 5760x1080, but that’s more of a sure thing as I approach 40 FPS with my overclocked $1600 machine.
To say that the Grid 2 High Quality results from Don’s $1300 PC are incredible would be an understatement: I’d love to toss my drive on his machine and take another run at it.
We do know that this benchmark becomes CPU- and DRAM-bound at lower resolutions, and find it tough to believe that Nvidia’s architecture has that much less impact on these components. It's implausible, but still possible. I’ll let him have those numbers and keep my fingers crossed that a card I have coming my way will confirm them.
Those DRAM and CPU bottlenecks go away at higher settings, and that’s where we see reasonable performance scaling. Paul’s little $600 PC pulls through with 39.2 FPS at 4800x900 (and without overclocking), even though it only needed to reach 1920x1080 to be rated a success.
Arma 3 plays smoothly enough at standard settings that any discussion of FPS is usually academic. It’s still nice to see that the cheap PC is powerful enough to play at 4800x900, while the bigger machines have frames per second to spare at 5760x1080.
This game remains playable on the $600 machine at Ultra quality, but only when overclocked and only up to 1920x1080. Even Mr. Hacksaw’s mighty $1300 machine coughs and sputters at triple-screen resolutions. At this point, our most taxing settings are an exclusive club with a $1600 minimum order.
I’ve always been curious about Don’s low-resolution performance scaling in Far Cry 3, since it’s something neither Paul nor I have been able to replicate (even when I’m using the same hardware). Maybe it’s not that we’re doing something wrong—maybe it’s that he’s benchmarking differently.
The $600 machine survived all of its Far Cry 3 tests at our lower test settings, but becomes unplayable at our higher settings. According to Paul, smooth gameplay is possible by either dropping from 1920x1080 to 1600x900 or from Ultra to Very High quality.
The $1300 machine makes a valiant effort to reach 5760x1080 at Ultra quality, but even the $1600 PC struggles. Thankfully, the drop between average and minimum is small in this title, with a 22 FPS minimum recorded for the $1600 machine’s baseline test.
Paul’s $600 PC enjoys substantial performance-to-price advantages in single-threaded applications like iTunes and LAME MP3 encoding. The expensive machines win, but only because they’re clocked higher.
I could probably take a break and return from it while the $600 PC finishes its video encoding tests. My $1600 machine looks moderately faster than Don’s $1300 competitor, but it’s also only moderately more expensive.
Paul’s $600 PC performs half as well as my $1600 PC in After Effects, again proving itself to be a value leader. Don’s marginally-cheaper $1300 machine is also a little bit slower.
Photoshop’s OpenCL-based filters prefer Nvidia architecture, handing a rare lead to the $1300 PC’s mid-price graphics solution.
Premiere Pro punishes the $600 PC’s two CPU cores, while giving little benefit to the $1600 build’s added CPU features.
Single-threaded PDF printing is more closely scaled to CPU clock rate.
Both 3ds Max and Blender favor my $1600 machine’s extra cache and/or logical cores, the latter feature intended to fill idle resources in the execution pipeline, minimizing the amount of time each core is left without work. The dual-core Pentium CPU in Paul’s $600 machine has half the core count and half the cache compared to Don’s more mainstream $1300 solution.
ABBYY FineReader and Visual Studio continue the “you get what you pay for” trend in CPU-oriented performance.
WinRAR, 7-Zip, and WinZip’s less taxing CPU-switched compression routine all appear to love the extra cache and logical cores on my Core i7-4790K. WinZip’s OpenCL-based workload, on the other hand, scales the three machines in order of graphics power, while its –EZ switch sets up a strange phenomenon in Don’s $1300 build that hammers performance.
None of our machines need all the power their PSUs can output. Note also that the chart reflects global (input) limits, and that power supplies are rated in output. Five-hundred watts of input power provides only 425 W at 85% efficiency, so the 750 W unit in my $1600 machine has 325 W to spare. Plenty, then, for the giveaway winner’s eventual CrossFire upgrade.
Running huge coolers that differ primarily in brand name, the $1600 and $1300 machines have nearly identical idle temperatures. The $1600 machine’s higher load temp is due to my use of “silent” fan mode, which assists in its modest 35 dB(A) maximum noise level.
Using the slowest system’s power and performance numbers as a baseline, I calculated how much more or less the other configurations used (energy) and provided (performance) as a percentage. Then, I subtracted the baseline to zero-out our efficiency score.
We can see that the $1300 PC was 78% faster than the $600 machine, yet needed 82% more power. The $1600 PC was over twice as fast as the $600 machine, but needed even more power.
All but one of the results are closely matched in performance-per-watt, and that one standout is the overclocked $1300 machine with its enormous 1.34 V core setting. I only hope that machine lives long enough to make it into the giveaway winner’s hands.
Readers who know the System Builder Marathon know that the cheapest machine always has the best performance-per-dollar, partly because it’s cheap, but mostly because its builder knows which cheap parts are required to reach a certain performance level. But does it really need to always win? Don and I had other plans.
Assuming you’re re-using your case and operating system, building with $946 of the $1300 machine’s parts still gets you a marginal value loss compared to the $600 PC. Yet, if you look back at how slow that machine was in certain benchmarks, it might be worth your time to spend the extra money (if you can).
If you already have an extra Windows license, you might prefer to build with all of our hardware, and only the hardware. Paul’s $600 build retains its value leadership, but my $1600 PC begins to gain on Don’s $1300 machine.
Adding nothing to a system’s performance, a new operating system hits Paul’s value score hard. My $1600 machine starts out stronger than Don’s, but aggressive overclocking pushes his beast into second place.
In fact, games are the only place where I can really find enough of a performance advantage to offset my machine’s extra price. Paul’s $600 machine can't compete there, and if I were to put zeros in places where Don’s system couldn’t play smoothly at 5760x1080, my lead would be even larger. That sounds like a terribly specific place to find value, but Don invented this parameter for our SBMs after all.
In total, Paul’s machine has the best value in numbers, but a look at some of our benchmarks shows that it’s going to be too slow for many of our readers. Make sure you look carefully before you take a leap at that one.
Readers who expect that I’d always pick my own machine should read the conclusion of our previous SBM. Even though my performance standards are high, I’m always interested in value. And going by the charts, Don’s machine looks like the best compromise of fast-enough performance and almost-cheap-enough value.
On the other hand, a look back at some of Hacksaw Don’s installation woes shows that its parts list needs revision before we can make a broad recommendation. And that leaves me, the quiet guy who types a lot, running a quiet machine that computes a lot. I said yesterday that I’d recommend it to anyone who can afford it, and I still do.



































