GTX 260 216 overclock

edited July 2009 in Hardware
I ordered this GTX 260 today for $145 shipped after $10 coupon code EMCLSNT35. (Newegg newsletter subscription required to use the coupon) It will be an upgrade for my son's 9600GT.

I will overclock the GTX260 and compare with my HD 4850 512MB (@850/1200) and post the results here. If GTX 260 is much faster than that my son might get the ATI instead :)

Has anyone overclocked this thing and how much? I would appreciate the info. Thanks!

Comments

  • edited May 2009
    I received a bad card :( It is usually fine in 2D, although sometimes system crashes while booting or I see "video driver restarted" message. When I start a 3D application, mosaic patterns appear in the screen and after that either system crashes with a blue screen or the "video driver restarted" message appears again. Replacement is in the mail from EVGA.

    It can very rarely run Crysis GPU benchmark for 3 loops, and something very interesting happened the two times I could complete out of many trials. GPU temperature went upto 85C in the first successful test and it was max 55C in the second; scores were very similar, 42fps @ 1680x1050, DX10, high, noAA. I checked the temperature by touching the card as well, it was burning my fingers in the first test and barely warm in the second. Very puzzling .... but this is a faulty card anyway.
  • edited May 2009
    Oh by the way, I am sure someone, who knows about PC hardware, would suspect PSU problem immediately. I suspected that as well and checked on two different computers; this is not a PSU problem. I am using a completely fine Antec EA650 with max total 45A on the 12v rails.

    Here are my HD4850 Crysis GPU test benchmarks but I will not say it beats GTX 260 just yet.

    850/1200
    Run #1- DX10 1680x1050 AA=No AA, 32 bit test, Quality: High ~~ Overall Average FPS: 45.065
    Run #1- DX10 1680x1050 AA=4x, 32 bit test, Quality: High ~~ Overall Average FPS: 35.17

    625/1000
    Run #1- DX10 1680x1050 AA=No AA, 32 bit test, Quality: High ~~ Overall Average FPS: 36.34
    Run #1- DX10 1680x1050 AA=4x, 32 bit test, Quality: High ~~ Overall Average FPS: 28.82

    System configuration:
    CPU: Q6600 @3.3GHz 9x367
    RAM: 6GB DDR2-733
    MB: P5Q-SE-PLUS
    OS: Vista64
  • scottscott Medina, Ohio Icrontian
    edited May 2009
    Hey Mirage
    Here is the requested info.
    I am running Windows 7 (7100)64 bit
    EVGA GTX260 core 216 ( OC'd )
    Q6600 @ 3.2

    and a screenshot for your viewing pleasure.
    side note: it was running 37ish fps at stock clocks.

    Scott
  • edited May 2009
    scott wrote:
    Hey Mirage
    Here is the requested info.
    I am running Windows 7 (7100)64 bit
    EVGA GTX260 core 216 ( OC'd )
    Q6600 @ 3.2

    and a screenshot for your viewing pleasure.
    side note: it was running 37ish fps at stock clocks.

    Scott

    WOW!!! Thank you so much, you are quick again :) I will use your benchmark as my reference since we have very similar systems.

    I will be able to post my benchmarks next weekend. I can't wait to see how much it will overclock. Nevertheless, the overclocked HD4850 (see above) performs admirably. Don't you think so? I already transferred the HD4850 to my son's PC and I am using the broken GTX 260 for now :(
  • scottscott Medina, Ohio Icrontian
    edited May 2009
    No problem !

    I just noticed you were running the 32 bit test so I reran it.

    Run #1- DX10 1680x1050 AA=No AA, 32 bit test, Quality: High ~~ Overall Average FPS: 45.25

    That was at the same OC.

    And yes that 4850 runs pretty good.

    Scott
  • scottscott Medina, Ohio Icrontian
    edited May 2009
    This is about the best I can get.
    725 Core
    1542 Shader
    1200 Memory

    Run #1- DX10 1680x1050 AA=No AA, 32 bit test, Quality: High ~~ Overall Average FPS: 47.135



    Scott
  • edited May 2009
    scott wrote:
    This is about the best I can get.
    725 Core
    1542 Shader
    1200 Memory

    Run #1- DX10 1680x1050 AA=No AA, 32 bit test, Quality: High ~~ Overall Average FPS: 47.135



    Scott

    Scott, thank you again! This will be great help for me. 725 is a very good overclock :thumbup, that would be a dream overclock for me. You are using the stock cooler I assume. How is the temperature while running the benchmark? BTW, is your GTX 260 a 55nm part or 65m?
  • scottscott Medina, Ohio Icrontian
    edited May 2009
    Thanks
    It will even run OCCT gpu test for over an hour with no errors at those speeds.
    In OCCT the max temp it reached was 71 but the benchmark never got it over 66 (with the fan at 100 ).
    It is the stock cooler.
    It is 65nm.

    Scott
  • edited May 2009
    Temperatures are good, card is stable. I would keep at that speed. I hope the card I receive behaves that well too :) You can use either EVGA's voltage control program or Rivatuner to go even higher (if you can still keep the temperature in check) but I won't have voltage control option with 55nm.
  • edited June 2009
    I have received the replacement card. It is working perfectly fine, and I was able to overclock a little :)

    Here are the results using Crysis GPU benchmark

    579/999 (Stock speed)
    Run #1- DX10 1680x1050 AA=No AA, 32 bit test, Quality: High ~~ Overall Average FPS: 41.44
    Run #1- DX10 1680x1050 AA=4x, 32 bit test, Quality: High ~~ Overall Average FPS: 35.185
    700/999
    Run #1- DX10 1680x1050 AA=No AA, 32 bit test, Quality: High ~~ Overall Average FPS: 46.545
    700/1100
    Run #1- DX10 1680x1050 AA=No AA, 32 bit test, Quality: High ~~ Overall Average FPS: 46.695
    700/1200
    Run #1- DX10 1680x1050 AA=No AA, 32 bit test, Quality: High ~~ Overall Average FPS: 46.77
    675/1152 (SSC speed)
    Run #1- DX10 1680x1050 AA=No AA, 32 bit test, Quality: High ~~ Overall Average FPS: 46.73

    I will keep the card at SSC speed from now on since 700 GPU clock was not stable with Vantage. An interesting observation is about the ineffective memory clock. If you check the results above, you can see that memory overclock is not improving the performance at all.
  • scottscott Medina, Ohio Icrontian
    edited June 2009
    mirage wrote:
    .
    .
    .
    675/1152 (SSC speed)
    Run #1- DX10 1680x1050 AA=No AA, 32 bit test, Quality: High ~~ Overall Average FPS: 46.73
    .
    .
    Nothing wrong with that !!

    I wish my Gigabyte board had SLI. I would buy another one if could justify it for gaming and folding. They are pretty nice cards for the money.


    Scott
  • edited June 2009
    Scott, thanks again for your help with the reference numbers. I hope this thread could be a reference for some others too.
  • scottscott Medina, Ohio Icrontian
    edited June 2009
    Oh ya

    I did play with the EVGA voltage tuner..

    I locked my machine up several times :) Some pretty cool artifacts too !!

    I could not get comfortable with it. I un-installed it.

    Scott
  • edited June 2009
    scott wrote:
    Oh ya

    I did play with the EVGA voltage tuner..

    I locked my machine up several times :) Some pretty cool artifacts too !!

    I could not get comfortable with it. I un-installed it.

    Scott

    So I should not worry too much about my inability to run voltage tuner with a 55nm card. :vimp:
  • edited June 2009
    I think I am done with overclocking this card. I have unlinked the GPU and shader frequencies. The GPU frequency can go upto 725 after unlinking the frequencies. Shader frequency do not want go beyond 1500. However, EVGA Precision utility does not set Shader/GPU frequency ratio below 2. So I can not test any GPU frequency higher than 725 when I set shader at 1450 and I want to have at least 50MHz safety distance from instability limit. Memory frequency does not seem to have much effect on the performance, so I decided not to push beyond 1200. In summary, the final overclock settings are 725/1450/1200. And the final Crysis score is

    Run #1- DX10 1680x1050 AA=No AA, 32 bit test, Quality: High ~~ Overall Average FPS: 47.214

    The GPU temperature saturates at around 60C with fan speed set at 61% (2000rpm) while running the Crysis GPU test for 10 cycles.
  • edited June 2009
    Just a little update about the fan speed settings.

    I wanted to start ramping up the fan speed at lower temperatures than the default 65-70C to keep the overclocked temperatures lower while running the fan at default 40% in 2D mode. EVGA Precision tool's profile based fan control was not exactly the way I wanted to do this. I wanted to control the automatic fan speed parameters to control the fan speed instead of defining speed-temperature steps in pairs. I switched to Rivatuner and used the low-level fan speed controls for that. I just changed the T_min and T_lowlimit values (as shown in the second picture) to 45 which were both 65. The results are summarized below in two temperature monitoring graphs with and without the modifications (left and right, respectively) in the low-level fan control while running the Furmark test after temperature saturates. The maximum temperature was 77C before the modified fan speed control and 67C after modification which starts to ramp up the fan speed at 45-50 instead of 65-70. This is, IMHO, easier and more effective than using Precision tool.

    Note: For Rivatuner to restore the fan settings after suspend mode it should be running at the background. So check the "send to tray on close" and "run at Windows startup" options in the Settings tab.
  • edited July 2009
    I noticed that Rivatuner occasionally fails to reset the fan speeds after wake up. The solution is posted here.
Sign In or Register to comment.