Memory Stress Testing

The SLI version of the Gigabyte nForce4 handled memory exactly as we found on the K8NXP-9, which is very good indeed. We did not see any issues with memory on this board that have been reported as problems on the older nForce3 Ultra Gigabyte boards.

The memory stress tests measure the ability of the K8NXP-SLI to operate at its officially supported memory frequency (400MHz DDR), at the lowest memory timings that OCZ PC3200 Platinum Rev. 2 modules will support. All DIMMs used for stress testing were 512MB double-sided (or double-bank) memory. To make sure that memory performed properly in Dual-Channel mode, memory was only tested using either one dual-channel (2 DIMMs) or 2 dual-channels (4 DIMMs).

Stable DDR400 Timings - One Dual-Channel
(2/4 DIMMs populated)
Clock Speed: 200MHz
CAS Latency: 2.0
RAS to CAS Delay: 2T
RAS Precharge: 10T*
Precharge Delay: 2T
Command Rate: 1T
*Several memory tests have shown that memory performs fastest on the nVidia nForce chipsets at a TRas (RAS Precharge) settings in the 9 to 13 range. Memory Bandwidth tests were run with memtest86 with TRas settings from 5 to 15 at a wide range of different memory speeds. The best bandwidth was consistently at 9 to 11 at every speed, with TRas 10 always in the best range at every speed. The performance improvement at TRas 10 was only 2% to 4% over TRas 5 and 6 depending on the speed, but the performance advantage was consistent across all tests. All benchmarks were therefore run at a TRas setting of 10.

By using two DIMMs in Dual-Channel 128-bit mode, the memory performed in all benchmarks at the fastest 2-2-2-10 timings, at default voltage.

Stable DDR400 Timings - 4 DIMMs
(4/4 DIMMs populated)
Clock Speed: 200MHz
CAS Latency: 2.0
RAS to CAS Delay: 2T
RAS Precharge: 10T*
Precharge Delay: 2T
Command Rate: 2T

Tests with all four DIMM slots populated on the Gigabyte required a 2T Command Rate with 4 DIMMs in two dual channels. This is the pattern seen on other top-performing Socket 939 boards.

Overclocking: Gigabyte K8NXP-SLI Test Setup
Comments Locked

58 Comments

View All Comments

  • BenSkywalker - Thursday, November 25, 2004 - link

    "I would have to say that many of you are in the minority even when it comes to the visitors of this website, let alone computer games in general. Many gamers such as me don't care about the extreme resolutions. Heck, I don't really care that much about SLI."

    You pretty much summed up the issue with your comments right there. Those that aren't interested in extreme resolutions aren't interested in SLI either. Those of us that are interested in extreme resoltution are interested in SLI too.
  • darklight0tr - Thursday, November 25, 2004 - link

    Some of the resolutions numbers being thrown around here are ridiculous! Anything above 1600x1200 I find extreme, and 1600x1200 should only be used with large CRT/LCDs. This is of course coming from me who runs a 19" CRT at 1024x768 because I don't like the looks of my monitor at any higher resolution.

    I would have to say that many of you are in the minority even when it comes to the visitors of this website, let alone computer games in general. Many gamers such as me don't care about the extreme resolutions. Heck, I don't really care that much about SLI.

    I am glad to hear that someone is always pushing the envelope, though.
  • Myrandex - Thursday, November 25, 2004 - link

    Awesome review. I don't know why everyoine is commenting on his resolution choices. Go read teh other SLI review, or read other SLI reviews from other websites such as the [H] or tomshardware. It looked like a sweet motherboard, and Fink this was a goiod review. There is one change I found though:
    Continuing the theme of "more than you might expect", Gigabyte provides eight SATA ports. Four ports are 3Gb/s ports provided by the nForce3 chip, and the other are four 1.5Gb/s ports driven by the PCI bus.


    That should say nForce4 chip I believe. Its on the first or second page. Keep up the good work though,
    Jason
  • Wesley Fink - Thursday, November 25, 2004 - link

    #20 - As I have already stated, 1024 was for comparison to other motherboard benchmarks so, for example, you could see the base K8NXP-SLI is fast as a motherboard compared to other A64 boards. 1280 was included for SLI. Anand reviewed SLI graphics capability at 1600x1200 for those interested in those results.
  • rslayerr - Thursday, November 25, 2004 - link

    #22
    i was coming with that point man...yes i do have a 19" that can do 1600x1200,but at 60hz...wich is no way good since i "feel"the refresh rate thing and that is giving me headache,1280x1024 at 75hz is the minimum (75 hz) aaceptable.85 the best.so westley,yur right man...no point in doing it at 1600x1200 and up...
    beside..some 19" does more hz at 1600x1200...but they are pricy!
  • R3MF - Thursday, November 25, 2004 - link

    with the advent of HDTV spec panels, SLI is truly useful, and we need to know that SLI can drive modern games at these resolution (1920x1200).
  • nserra - Thursday, November 25, 2004 - link

    I never played games with monitor higher then 19", but is it really need AA with 1600x1200 or higher resolution? There are so many pixels that I doubt we see them (except maybe TFT).

    Why every one says that want to test the SLI at 2048x1536? Which monitor will do that with quality and frequency? Who have of those monitors right now?

    1280x1024 75 Hz is not enough, at least 85 Hz is needed, and i think when the resolutions increases you need more Hz or your head will mess up! So where is a 2048x1536 100 Hz monitor? If exists do you really think it cheap?

    Is this card ok for games of just work? http://www.3dlabs.com/products/product.asp?prod=29...

    SLI = 5000$ PC
  • R3MF - Thursday, November 25, 2004 - link

    i can understand the 1280x1024, as many people will have 17-19 inch LCD panels, or otherwise game at that res with their CRT, and for those hoping to get two 6600GT's then it makes a lot of sense, but 1024x768!

    i love anandtech motherboard grouptests, they are useful beyond many other sites. i look forward to an SLI grouptest where i would like to see the following resolutions benchmarked.

    1280x1024 (6[b]6[/b]00GT SLI)
    1680x1050 (6[b]6[/b]00GT SLI + 6800GT SLI)
    1600x1200 (6800GT SLI + 6800[b]U[/b] SLI)
    1920x1200 (6800GT SLI + 6800[b]U[/b] SLI)
    2048x1536 (6800[b]U[/b] SLI)



  • BenSkywalker - Thursday, November 25, 2004 - link

    Wesley- We are still waiting for the 6800NU PCI-E parts to arrive. You may have access to information and when that will be, but everything in the public domain indicates that we will be waiting for the NV41 revision and no firm street date(based on the public information). 6800NU in SLI as of now is something looming on the horizon.

    2x6600GT + NF4 SLI Premium is a bit higher then $400. 6600GT in an SLI configuration makes sense for those looking to do an upgrade now with a future viable path for them to take down the road to up performance again. It makes little sense to pay the premium for NF4 SLI w/dual 6600GTs when you can buy a 6800GT for less and avoid having to pay the premium for the SLI capable mobo. Their performance is close enough, and who has the performance edge depends on the settings although it is the 6800GT more frequently then not.

    I think you are kidding yourself if you think that there are many people concerned over the performance difference of Halo running 1024x768 on any high end vid card setup. The game is quite limited at that setting by the processor/platform, certainly far more then the vid card(excluding a single 660GT). Those that have either a SLI setup or one of the higher end boards are going to be running at settings considerably higher then 1024x768.

    For Aquamark3 I think you will find that pretty much noone cares one way or the other. The game the bench is based on is quite poor and already forgotten by pretty much everyone.

    As far as the 6600GT performing close to the 6800U at 10x7 in D3, that is why I said you might as well have tested 320x240. You are processor/platform limiting the benches making the vid card varriable as small of a factor as possible. Most of the varriation we are seeing could easily be driver overhead- without stressing the parts we have no way of knowing how they are going to stack up in real world useage. Talk to some people who play games regularly and know even a little about computer hardware- ask them why they would spend $400(or signficantly more) on a video card. I can assure you that it is not to run 1024x768.
  • Wesley Fink - Thursday, November 25, 2004 - link

    Comparing Doom3, Far Cry, and UT2004 results from the launch review benchmarks, the Gigabyte is significantly faster than the Asus SLI. However, Anand's benches were run at slower memory timings due to issues with the first SLI board - that were corrected with the second board he received. For that reason it was not completely fair to compare the Asus SLI benches to the Gigabyte SLI.

    The first thing tested on the Gigabyte was to compare stock 6800U results to other A64 tests we have run. We saw the Gigabyte SLI performing the same as the K8NXP-9 nF4, which is the fastest A64 board we have tested so far. To answer the question, the Gigabyte is a lot faster than the Asus as the Asus was tested. We will run benchmarks on the Asus SLI for comparison as soon as we receive the production board here for testing.

    #18 - 2x6600GT it $400 total, 2x6800 is $600 total, 2x6800GT is $800, and finally 2x6800 Ultra is $1000. There are many SLI options other than 2 6800 Ultra and I believe many people do care that 2x6600GT performs faster than a single 6800 Ultra in Aquamrk 3 and Halo at 1024. In fact that $400 combo performs about the same as the more expensive single 6800 Ultra in Doom 3 at 1024.

Log in

Don't have an account? Sign up now