System Builder Marathon: Price/Performance


As we wrap up another system builder marathon, we realize while nothing in life ever seems to work out perfectly; sometimes things work out pretty well.

This marathon was supposed to be published a number of weeks ago – before the new Radeon 4800 series was launched – and therefore we didn’t have the option of selecting ATI’s excellent new cards for our builds. Unfortunately, our intrepid reviewer Thomas Soderstrom had an unlucky break (involving a few bones and some surgery) and was unable to assemble and test one of the systems, forcing us to defer the marathon. It was further delayed when Nvidia and ATI released their new video cards, the reviews of which of course took precedence over our marathon and pushed us back even further.

While the video cards we chose for these systems might not be our first choice if we could do it all over again (considering what’s available), the rebates available on the GeForce 8800 cards are so prevalent that even today a pair of them in SLI would be $100 cheaper than a pair of 4850s, and realistically the prices will likely drop even further. The rest of the builds were spec’d out well enough that these marathon systems are still viable hardware ensembles.

For more information on the components used to build each of the systems we’re comparing, in addition to the overclocked configurations, check out the sub-$1,000 PC, the sub-$2,000 PC, and the sub-$4,000 PC.

Let’s review the benchmarks we’ve used once more, see how everything stacks up, and analyze the price/performance data to see the value each system offers for the dollar.

Swipe to scroll horizontally
Row 0 - Cell 0 3D-Games
CrysisVersion: 1.1
Row 2 - Cell 0 Video Quality 1: High Details No Anti-Aliasing
Row 3 - Cell 0 Video Quality 2: Very High Details, 4x Anti-Aliasing
Row 4 - Cell 0 Benchmark: Benchmark_CPU.bat
PreyVersion: 1.3
Row 6 - Cell 0 Video Quality 1: Default (No AA, 8x AF)
Row 7 - Cell 0 Video Quality 2: High Quality, 4x AA
Row 8 - Cell 0 Benchmark: THG-Demo
SupremeVersion: 3.220
CommanderVideo Quality 1: Default
Row 11 - Cell 0 Video Quality 2: High Fidelity, High Shadow, 4x AA
Row 12 - Cell 0 Benchmark: Real 60 Game
UnrealVersion: Retail
Tournament 3Texture Detail: 5
Row 15 - Cell 0 World Detail: 5
Row 16 - Cell 0 Field of View: 100
Row 17 - Cell 0 Benchmark: Botmatch (WAR-Torlan, 12 bots, 1 Minute)
WarhammerVersion: 1.6
Mark of ChaosVideo Quality:Default (Highest Settings)
Row 20 - Cell 0 Demo: THG Timedemo (1 Minute)
Row 21 - Cell 0 Audio
iTunes 7.2Version:
Row 23 - Cell 0 Audio CD "Terminator II SE", 53 min
Row 24 - Cell 0 High Quality (160kb/s)
Lame MP3Version: 3.98 Beta 3 (05-22-2007)
Row 26 - Cell 0 Audio CD "Terminator II SE", 53 min
Row 27 - Cell 0 wave to MP3
Row 28 - Cell 0 160kb/s
Row 29 - Cell 0 Video
4.0 XpressImport File: Terminator 2 SE DVD (5 Minutes)
Row 32 - Cell 0 Resolution: 720x576 (PAL) 16:9
Row 33 - Cell 0 Audio: Dolby Digital, 48 kHz, 6-Channel, English
Row 34 - Cell 0 Advanced Acoustic Engine MP3 Encoder (160 kb/s)
DivX 6.6Version: 6.6.1
Row 36 - Cell 0 Profile: Home Theater Profile (720x576)
Row 37 - Cell 0 1-pass, 780 kb/s
Row 38 - Cell 0 Encoding mode: Insane Quality
Row 39 - Cell 0 Enhanced multithreading
XviD 1.1.2Version: 1.1.2
Row 41 - Cell 0 Target quantizer: 1.00 (maximum quality)
Row 42 - Cell 0 Applications
AdobeVersion: 10.0x20070321
Photoshop CS3Filtering 69 MB TIF Photo
Row 45 - Cell 0 Benchmark: Tom’s Guide-Benchmark V1.0.0.4
AutodeskVersion: 9.0
3D Studio MaxRendering One Video Frame
Row 48 - Cell 0 Quality: HTDV 1920x1080 and 1280x720
Grisoft AVGVersion: 7.5.467
Anti-VirusVirus Base: 269.6.1/776
Row 51 - Cell 0 Benchmark: Scanning 3.82GB Application Folder
MaxonVersion: 10.008
Cinema 4DResolution: 1280x1024
Row 54 - Cell 0 Benchmark: Rose Drop, 8 bits (50 Frames)
Rarlab WinrarVersion 3.70 BETA 8
Row 56 - Cell 0 Compression = Best
Row 57 - Cell 0 Dictionary = 4096 kB
Row 58 - Cell 0 Uncompressed Folder Size: 642 MB
Row 59 - Cell 0 Synthetics
SPECviewperf 9Version: 9.0.3
PCMark05 ProVersion: 1.1.0
Row 62 - Cell 0 System, CPU and Memory Tests
Row 63 - Cell 0 Windows Media Player 11.0.6000.6324
Row 64 - Cell 0 Windows Media Encoder
FuturemarkVersion 1.10
3DMark 2006System Test Only
SiSoftwareVersion 2008.1.12.30
Sandra XIICPU Test = CPU Arithmetic, Multimedia
Row 69 - Cell 0 Memory Test = Bandwidth Benchmark
  • L1qu1d
    Really shocked at how much of a benefit the 4000$ computer did when Overclocked:| Especially in Crysis, WOW!
  • zipz0p
    And yet it's still smashed in bang/buck.

    I like the introduction of the high-resolution gaming bang/buck chart - it's a keeper!
  • randomizer
    That extra bang/buck chart was an interesting addition.
  • sublifer
    Great charts. Still though... not a whole lot of value since we've gotten new video cards out. I know its unreasonable to expect the entire thing to be redone with the new components, but at the same time, many people (yes, me too....) would like to see how the outlook might change with the new graphics cards. Is there any way you could substitute appropriately (by that I mean cost) in a few of the systems and compare them with the sbm results?

    Thanks again guys!
  • gwolfman
    What's interesting is that the $2000 build overclocked was able to return slightly above (or slightly below in hi-res gaming) it's value compared to the original build. That means that ever dollar spent is rewarded back with an equal system performance increase. Very interesting.
  • beerzombie
    I think comparing performance/price with overclocked systems when the price of aftermarket cooling solutions are included in the non-overclocked system is a bit misleading. I feel that the reality is that the Sub $4k PC is $256 cheaper when not overclocked, and the sub $2k machine should be $120 cheaper as well as the $1k PC being $26 cheaper. It is just unrealistic to assume someone is watercooling a PC and won't be overclocking it.
  • zipz0p
    Interesting point, BeerZombie. I would like to see the price/performance comparisons again taking that into account!
  • gaiden
    Interesting, according to the Hi-res bang/buck value overclocking a sub$2000 will increase the value by 26% while sub$1000 o.c. will increase by 36% ! This is a very useful stat indeed and i agree with others that it's a keeper for sure. So in conclusion it makes sense that to get the most out of your config vs $ you spend should be anywhere between the $1000-$2000 with o.c. which probably represents majority of the ppl out there. < (pointless, but nice to hear statement) :)
  • Preclude
    The parts are not relevant. What people are missing here is the overall message. They are trying to visually display the aspects of price to performance in the PC market for custom builds. The data will always be relevant no matter what the hardware goes to. There will always be the "best performing budget rig", the "step up medium high end build" and the "Not really worth the money but if you want it you want it" builds. I'm sure some could argue builds between those, but you will always find that builds fall near one of those general areas.

    TL;DR Thanks Toms, anyone who actually builds systems on a regular basis can appreciate this data.
  • Kirth Gersen
    What you can take away from this is that you are very much in the realms of the law of diminishing returns with a $4k system from a gaming perspective. 30FPS is plenty, your eyes can't perceive better, so what is the point of a system which can achieve 80FPS? At least a Ferrari might improve the sex life :).

    A mid range $2k system has always been my price point. For that you can usually build an overclocked system which hits 30FPS on max graphics.