Sign in with
Sign up | Sign in

Overclocking, Test System Configuration, And Benchmarks

Part 4: Building A Balanced Gaming PC
By

Overclocking

The goal here was not to set any air-cooled overclocking records with our hardware samples, but rather to seek out stable clock speeds we could count on through the duration of the series, including these warm summer months. We therefore did not push voltages as high as you’d see in other articles. Load temperatures or diminishing returns from extra voltage were among reasons for dialing back a bit from the maximum stable overclock.

Those of you unfamiliar with overclocking AMD processors can check out these two “How To” guides for overclocking both "unlocked" Black Edition and standard "locked" AMD processors. Here, we used BIOS overclocking to first find the highest stable CPU core speed, and then tweak some additional performance by raising the northbridge (NB), HyperTransport (HT) link speed, and memory frequencies.

Apart from a tedious Phenom II X2 550 BE overclock, which went late into the night, the procedure was straightforward and uneventful. Despite 3+ hours of success in Prime 95’s “Blend” test, the only solution found for complete stability in Grand Theft Auto IV and Need for Speed Shift was to knock the Phenom II X2 550’s core speed down almost 80 MHz.

Readers should keep in mind that, by mixing locked and unlock processors in today’s story, there’s more to look at with each overclock than just the CPU core speed. For instance, while the Phenom II X6 1055T easily hit the highest core speed, less multiplier flexibility led to pairing with the lowest HT link speed and memory frequency.

BFG’s factory overclocks were used for both the GeForce GTX 260 and GTX 285. For other graphics cards, the GPU itself was dialed down about 20 MHz from its limit, while graphics memory was reduced by 20-30 MHz. The lone exception is the Radeon HD 5970, which didn’t get far without additional voltage. Once we added more juice, the card could no longer run Furmark without the VRM temperatures skyrocketing upward to the 120 degree throttling limit. This air-cooled design consideration prevented torture testing, so we chose to play it safe and run all Radeon HD 5970 numbers at the card’s stock speeds.

Be sure to check the test configuration table below for specifics on the overclocks achieved. For comparison purposes, we’ll include data for the overclocked Intel Core i7-920 from Part 3 of the series.

Test Hardware
Processors
AMD Phenom II X6 1055T (Thuban) 2.8 GHz, Socket AM3, 2000 MHz HT Link, 6 MB L3 cache. Overclocked to 3.864 GHz (14*276), 1.363 V, 2484 MHz NB, 1932 MHz HT Link
AMD Phenom II X4 955 Black Edition (Deneb) 3.2 GHz, Socket AM3, 2000 MHz HT Link, 6 MB L3 cache. Overclocked to 3.708 GHz (18*206), 1.45 V, 2472 MHz NB, 2060 MHz HT Link
AMD Phenom II X2 550 Black Edition (Callisto) 3.1 GHz, Socket AM3, 2000 MHz HT Link, 6 MB L3 cache. Overclocked to 3.774 GHz (18.5*204), 1.425 V, 2652 MHz NB, 2040 MHz HT Link
AMD Athlon II X4 640 (Propus) 3.0 GHz, Socket AM3, 2000 MHz HT Link, no L3 cache. Overclocked to 3.645 GHz (15*243), 1.45 V, 2430 MHz NB, 1944 MHz HT Link
Intel Core i7 920 (Bloomfield) 2.66 GHz, LGA 1366, 4.8 GT/s QPI, 8 MB Shared L3 cache, Overclocked to 4.0 GHz (21*190), 1.237 V idle, HTT Disabled
Motherboards
Asus Crosshair III Formula (Socket AM3) AMD 790FX /SB750, BIOS 1702 (5-26-10)
Asus Rampage II Extreme (LGA 1366) X58/ICH10,  BIOS 1504 (07/23/09)
RAM
Corsair 4 GB (2 x 2 GB) DDR3-1600, 8-8-8-24 @ DDR3-1472 (X6 1055T), DDR3-1648 (X4 955 BE),  DDR3-1632 (X2 550 BE), DDR3-1620 (X4 640), 1.65 V
Corsair 6 GB (3 x 2 GB) DDR3-1600, 8-8-8-24 @ DDR3-1524 (Core i7-920), 1.6 V
Graphics
ATI Radeon HD 5970 2 GB Stock @ 725 MHz GPU, 1000 MHz GDDR5
ATI Radeon HD 5870 1 GB OC @ 910 MHz GPU, 1270 MHz GDDR5
ATI Radeon HD 5750 1 GB OC @ 830 MHz GPU, 1320 MHz GDDR5
ATI Radeon HD 4890 1 GB OC @ 980 MHz Core, 1100 MHz GDDR5
Nvidia GeForce GTX 295  1792 MB OC @ 648 MHz Core, 1404 MHz Shaders, 1150 MHz GDDR3
Nvidia Geforce GTX 285 1 GB OC @ 712 MHz Core, 1620 MHz Shaders, 1332 MHz GDDR3
Nvidia Geforce GTX 260 896 MB OC @ 655 MHz GPU, 1404 MHz Shaders, 1125 MHz GDDR3
Hard Drives
Western Digital Caviar Black WD6401AALS, 640 GB, 7200 RPM, 32 MB Cache, SATA 3Gb/s
Power
Corsair CMPSU-850HX 850 W
CPU Coolers
Xigmatek HDT-S1283
Xigmatek Dark Knight-S1283V
Software and Drivers
Operating System
Microsoft Windows 7 Ultimate x64
Graphics Drivers
AMD Catalyst 9.12
Nvidia GeForce 195.62
Benchmark Configuration
3D Games
Crysis
Patch 1.2.1, DirectX 10, 64-bit executable, benchmark tool, Very High Quality, No AA
Far Cry 2
Patch 1.03, DirectX 10, in-game benchmark, Ultra High Quality, 4x AA
S.T.A.L.K.E.R.: Clear Sky
Average of 4 segments 'a-tested object', DirectX 10/10.1, Ultra High Quality, EFD Lighting, no MSAA
Grand Theft Auto IV
Patch 1.0.4.0, in-game benchmark, High Quality Settings, High Textures, Medium View Distance
Fallout 3
Patch 1.7, Custom THG Benchmark "Capital Wasteland" (60 sec.), Ultra High Quality, 4x MSAA, 15x AF
Need For Speed Shift
Patch 1.02, Custom THG Benchmark "Dakota GP" (60 sec.), Highest Quality, 4x AA/ 16xAF
World in Conflict
Patch 1009, DirectX 10, timedemo, Very High Quality, 4x AA / 16x AF
Display all 43 comments.
This thread is closed for comments
  • 3 Hide
    wildeast , August 11, 2010 6:33 AM
    "such as NVidia’s GeForce GTX 400-series and revamp the benchmark suite with some new DirectX 11 titles."
    i'll be waiting for that, and maybe some i5 cpu to see what fit sli best
  • 8 Hide
    jsowoc , August 11, 2010 6:58 AM
    "We set forth to measure the perfect balance in seven different games and four resolutions in this third of many parts." (?)

    I think you copied this paragraph from part 3 and forgot to change it to 4... ;-)
  • 4 Hide
    theshonen8899 , August 11, 2010 7:05 AM
    With the amount of love you guys have for the Athlon x3 I was really hoping to see it on here :\
    I guess I can kind of predict where it'd fall though.
  • 2 Hide
    Darkerson , August 11, 2010 7:23 AM
    I love the in-depth articles like these. Keep 'em coming!
  • 7 Hide
    L0tus , August 11, 2010 7:40 AM
    Brilliant piece.

    I wish I had read this before building my system as I can see that I clearly spent too much on my CPU instead of GPU (i5-750 + HD5770) . Would have done much better with (X2 550 BE + HD5850) !

    ...ain't hind sight a b***h!

    Also interesting to see how GPUs really start to distinguish themselves at higher resolutions. Again, brilliant work.
  • 2 Hide
    TheStealthyOne , August 11, 2010 8:01 AM
    I built a computer for my brother using a Phenom ii 550 paired with a 5770, and it screams! Fantastic gaming chip! It just goes to show you can achieve fantastic performance by planning and balance.
  • -1 Hide
    garlik_bread , August 11, 2010 9:10 AM
    Personally, i'd be interested to see results from a card with less han 1GB RAM on the GPU.

    On the lower end of the spectrum, with the lower resolutions, is the 1GB really necessary?

    Basically, i have a 512MB Asus 5770 and want to validate my purchase :D 
  • -1 Hide
    plasmastorm , August 11, 2010 10:22 AM
    Still running a Maximus formula 775 board with a Q6600, 8gb ram and a Radeon 5850 but this is certainly handy for future reference.
    Probably skipping the i5/i7 generation as I can still play anything at max settings on my 22" monitor while running a 2nd for a film tho :) 
  • 4 Hide
    Tamz_msc , August 11, 2010 11:22 AM
    Please test some newer games, which is essential for an article like this.
  • 1 Hide
    descendency , August 11, 2010 12:27 PM
    plasmastormStill running a Maximus formula 775 board with a Q6600, 8gb ram and a Radeon 5850 but this is certainly handy for future reference.Probably skipping the i5/i7 generation as I can still play anything at max settings on my 22" monitor while running a 2nd for a film tho


    i5/i7 isn't a generation. it's like 5 or so.

    It's the same thing as C2D and C2Q
  • -1 Hide
    jonpaul37 , August 11, 2010 12:35 PM
    plasmastormStill running a Maximus formula 775 board with a Q6600, 8gb ram and a Radeon 5850 but this is certainly handy for future reference.Probably skipping the i5/i7 generation as I can still play anything at max settings on my 22" monitor while running a 2nd for a film tho


    I hear ya man, i have a Q6600 @ 3.6 and a GTX 285 and i can rock anything i play with really nice settings at 1920 x 1080 so it looks like i will be holding out for another year or two...
  • 1 Hide
    Onus , August 11, 2010 1:14 PM
    Very nice. I really like this series.
    Suggestions: there's no need to draw curves; they should be point-to-point lines, as the data is discrete rather than continuous.
    For the RPG, I would suggest Dragon Age: Origins as being more demanding at higher settings, and/or Sacred 2 because of its use of PhysX. The latter runs the risk of becoming an ATi vs. nVidia comparison, but still may be useful.
    It would also be useful to have commentary on what bare minimum lowering of a setting or two is most likely to restore playability without sacrificing appearance too much.
  • 1 Hide
    lunyone , August 11, 2010 1:16 PM
    This totally makes my point, when I say a ~$100 CPU and a $200-$300 GPU are a the best budget gaming machines you can get. I usually make ~$100 CPU choices and ~$100-150 GPU choices when I'm building a budget gaming rig!! :) 
  • -4 Hide
    lemieuxxx , August 11, 2010 2:06 PM
    what about the i3 540. Is it horrable i see its not on here.
  • -2 Hide
    wolfram23 , August 11, 2010 2:46 PM
    Great article. Good to see how a faster CPU can really pull out better FPS, and it seems to make much more difference when having dual GPUs - I can only assume the trend would hold true on dual card (sli/cf) set ups, which makes me even happier to have an i5 750 @ 4ghz with my two 5850s.

    I hope Part 5 has i3, i5, i7, X955, X965, 1055T, 1090T and concentrates on DX11 performance (4xx vs 5xxx). I'd also LOVE to see a more in depth look at CF/SLI configs. There's not a lot of in depth looks at CF5770, CF5850, SLI470... there's some, but not a lot and none comparing these set ups to different CPUs.
  • 2 Hide
    felang , August 11, 2010 4:18 PM
    Catalyst 9.12 and only outdated games... is this January 2010 or what? you should at least test BFBC2, it uses as many cpu cores as you can throw at it...
  • 1 Hide
    Onus , August 11, 2010 4:32 PM
    Felang: they wanted it possible for readers to compare the results with previous articles in the series.
  • -1 Hide
    DXRick , August 11, 2010 5:04 PM
    Very nice article! I noticed that HD5970 and GTX295 benefited the most from the i7-920 CPU. This implies that Crossfire and SLI (multi-GPU setups) scale better with faster quads (and duals?). Thus, it would be nice to see how various CF and SLI setups depend on the CPUs in this test.

    Why did you use the older generation of Nvidia GPUs in this test? We are looking at the GTX460/470/480 now, with numerous test showing how well two 460's in SLI do.
  • 0 Hide
    nforce4max , August 11, 2010 5:04 PM
    Nice article, as for the L3 assessment I do agree that the lack of L3 cache does negatively impact performance but at least its not catastrophic as seen with the early days of the Celeron. Personally I use a meager 8250e that I nuked to 2.57ghz and it gets the job done plus it was dirt cheap. $16 after selling off some junked parts.
  • 1 Hide
    tognetta , August 11, 2010 5:19 PM
    I would like to see the great GTX460 here too ...

    Great job, i am thankful that i read it before building my next gaming machine !
Display more comments