Gaming Performance

While Sandra and Super Pi provided useful information on how memory performs without the influence of other components, our first round of real world tests has shown there is not a noticeable difference in performance between DDR2-533 and DDR2-667 operation on this board. However, our next set of real-world benchmarks consists of gaming benchmarks. Due to the amount of information and the relative closeness of the scores we are not utilizing zero based charts as the line graphs would stack on each other. However, you can click on the graph for a zero based graph of the presented information. Although the data chart tells the story we felt like there should be some separation in the line graphs to accurately depict the information.

We are utilizing Half-Life 2: Lost Coast and Quake 4 in our testing because they are sensitive to memory bandwidth changes. We added Serious Sam II to our benchmark mix because it is typically very GPU dependent and we will see the effects of our memory/GPU choices on these games. We tested at the 1024x768 resolution with High Quality settings in order to reduce the effects of the video card on testing. We found the 7600GS to be GPU limited at 1280x1024 but it still provided a decent gaming experience with scores dropping off around 30% on average but still remaining around the 60FPS level.





Click for Zero Point Charts

Our two memory sensitive games show a continuing pattern with ASRock ConRoe945G-DVI motherboard performing extremely well at both memory settings in Quake 4 and Half-Life 2: Lost Coast. The board finished with top scores in Quake 4 and very good results in Lost Coast, although at DDR2-667 the board did not scale at the same rate as in the other games. The ASRock 945G board finished second in Serious Sam II where the NV 570 SLI outclassed the other boards in our last tests. The onboard Intel GMA950 solution could not complete the Half Life 2 or Serious Sam II tests. It was able to beautifully create a stunning slideshow effect in Quake 4 with an average FPS rate of 10.9 at 1024x768 and 7.6 at 1280x1024. As we mentioned earlier, this graphics solution is best left running TurboTax and playing Internet Tetris.

Application Performance Quick Thoughts
Comments Locked

33 Comments

View All Comments

  • JarredWalton - Friday, September 1, 2006 - link

    The SATA issues at least are chipset/BIOS/motherboard dependent. As this uses an Intel 945G chipset, it should work fine with SATA. NVIDIA chipsets should also work fine, but I don't know what exactly ASRock did with the AM2NF4G-SATA2. Normally, for XP installs you want to set the SATA ports to IDE mode, and once XP is installed with all the drivers you can switch to AHCI mode instead... not that there's much performance difference anyway.

    In terms of stability, have you run without the OC? I really don't trust $50-$60 boards for overclocking. Some will do it, but it's still a bit of an iffy propositions for long-term use. I'd much rather spend $100 on a board I know will OC reliably.
  • yyrkoon - Friday, September 1, 2006 - link

    Yeah, I ran everything per spec. for 3 weeks, and wasnt until recently that I decided to OC it. My thinking is that, if this board isnt going to be stable for more than 3 days, I dont care much for it anyhow. Also, I had planned on replacing it with a ABIT AN9 32x when I bought it. Now Im starting to think maybe I'll replace it with a ABIT AB9 Pro, and a E6x00 CPU.

    Anyhow, the HTT is slightly above 1000MHZ, the memory is running ar 832MHZ, and CPU is reported by the BIOS as running at 2.76GHZ (although CPU-Z says a bit lower), and its running as stable, load, or idle as before. CPU temperatures havent exceeded 122F much, if at all. I have run nvmonitor stability tests, and sandra burn-in test on this system without a hitch, and if you think about it, 16% really isnt a huge OC . . .
  • JarredWalton - Friday, September 1, 2006 - link

    I'm guessing you have a 4200+ chip, right? 11X multiplier with a 250 MHz bus?

    Deoending on CPU, RAM, etc. you're possibly hitting other limits. Just for kicks, try dropping the RAM from DDR2-800 to a lower speed, like set it for DDR2-533 instead of DDR2-667. That should put the RAM at DDR2-611 vs. DDR2-786 (with a 250 MHz HTT speed).

    I would also adjust the memory timings for more conservative values. I'm guessing that with limited memory voltages, you might have problems running any DDR2 memory at 800 MHz. It's a thought anyway. :)
  • yyrkoon - Friday, September 1, 2006 - link

    Also, according to Sandra, my memory bandwidth is slightly above 7000MB/s int, and float. IS this decent ?
  • yyrkoon - Friday, September 1, 2006 - link

    Oh, sorry here:

    Asrock AM2NF4G-SATA2
    AMD 64 3800+ (single core)
    2x1 GB Cosair XMS 6400 DDR2
    eVGA 7600GT KO (stock), Antec SL 450 450W PSU.

    Since on this board there is no divider settings (or atleast none that I know of), basicly, all I've adjusted is memory (533,667,800), HTT link speed, CPU FSB, and multiplier. First I started off at 12x multiplier, and achieved a 230 FSB, anything above, wenbt unstable either in windows, or wouldnt boot. Here, I noticed the memory was quite capable of 880MHZ DDR, but I've had it as high as 902MHZ DDR stable(5-5-5-18). This memory will also run 4-4-4-12 stock volts at stock settings(even though SPD says otherwise). OK, then I dropped the multiplier, and increased FSB in increments of 10's until it failed to boot, then I dropped HTT link speed(800/1600), and memory to 533 standard. After this I increased FSB speed by 10's again, until falure, and eventually arrived at 250MHZ. Dropping the multiplier to 10x would not post with minimum FSB speed I wanted, so I called it quits.

    So in summer basicly, right now, my HTT link is 800/1600, multiplier 11x, and memory is set to DDR2 667 (in the BIOS). nvmonitor reports CPU is running at 2.7x, memory is running at 832, and HT bounces around from high 900's to low 1000's. CPU-Z says the memory speed is 392.xx, CPU is 2.7x, and HT is 1000.

    As far as I know, I've done about all I can, and still maintain a decent clock/ ratio settings. However, I havent OC's since I owned a P55 233mmx CPU (yes, MANY years ago) so I may be missing something. If so, Im all ears :)
  • yyrkoon - Friday, September 1, 2006 - link

    err summer = summary ? :/
  • JarredWalton - Saturday, September 2, 2006 - link

    Your CPU multiplier should only affect things if you set it lower than stock. The Athlon 64 3800+ has a 12X multiplier by default. I generally don't think there's much point in using lower multipliers, as the difference in memory bandwidth has very little impact on Athlon 64 processors. At the stock 12X multiplier, you get nicer memory ratios at DDR2-800 and DDR2-533 settings. Remember that memory speed is actually based off of a divider of the CPU speed. So:

    12 x 200 = 2400 MHz
    DDR2-800 = CPU/6 = 400 MHz (DDR2-800)
    DDR2-667 = CPU/8 = 300 MHz (DDR2-600)
    DDR2-533 = CPU/9 = 267 MHz (DDR2-533)

    Overclocked:
    12 x 233 = 2796 MHz
    DDR2-800 = CPU/6 = 466 MHz (DDR2-932)
    DDR2-667 = CPU/8 = 350 MHz (DDR2-799)
    DDR2-533 = CPU/9 = 311 MHz (DDR2-621)


    Change the multiplier to 11X and you get the following:
    11 x 200 = 2200 MHz
    DDR2-800 = CPU/6 = 367 MHz (DDR2-733)
    DDR2-667 = CPU/7 = 314 MHz (DDR2-629)
    DDR2-533 = CPU/9 = 244 MHz (DDR2-489)

    Notice how certain memory ratios and up with odd final results. That's what you're getting right now most likely. Let's say you go with a 250 MHz bus as an overclock. This is what you get:

    Overclocked:
    11 x 250 = 2750 MHz
    DDR2-800 = CPU/6 = 458 MHz (DDR2-917)
    DDR2-667 = CPU/7 = 393 MHz (DDR2-786)
    DDR2-533 = CPU/9 = 306 MHz (DDR2-611)
  • yyrkoon - Saturday, September 2, 2006 - link

    Jarred, heres a link to a post I made on the forums, if you care to comment further, to see a couple of OC test cases I've done so far

    http://forums.anandtech.com/messageview.aspx?catid...">http://forums.anandtech.com/messageview...amp;thre...
  • yyrkoon - Saturday, September 2, 2006 - link

    Perhaps I should benchmark both, although all sanda CPU tests already indicate that my system performs a slight bit slower than a Fx-57 system.

    The reason why I dropped the multiplier, was in hopes of gaining a higher CPU Frequency, which proved true. My thinking was that if voltage is limiting my increases, perhaps lower the multiplier will relax the CPU some. Since memory bandwidth to a point is rather moot, memory bandwidth isnt a big deal IMO either, how I may do some serious benchmarking comparrison for the hell of it, and may even post them somewhere. I can tell you however that Oblivion does perform better with the OC (was the first thing I checked).
  • Dfere - Friday, September 1, 2006 - link

    Anyone give me a general ballpark on what this integrated video should perform as? 6200 TC ?? 7300?

    Thanks

Log in

Don't have an account? Sign up now