Overclock stability testing.

Discussion in 'Videocards - NVIDIA GeForce' started by wasteomind, Oct 11, 2014.

  1. wasteomind

    wasteomind Guest

    Messages:
    398
    Likes Received:
    14
    GPU:
    EVGA RTX 3090 FTW3
    Its been a couple years since i got new video cards and oced em. What are people using these days to check stability? Last time i went looking for stability people were melting their cards with furmark and heaven marathons.

    I'm going to assume something more reasonable has come along?
     
  2. Extraordinary

    Extraordinary Guest

    Messages:
    19,558
    Likes Received:
    1,638
    GPU:
    ROG Strix 1080 OC
  3. ---TK---

    ---TK--- Guest

    Messages:
    22,104
    Likes Received:
    3
    GPU:
    2x 980Ti Gaming 1430/7296
    Heaven 4.0 very sensitive to bad oc not enough voltage etc
    Valley 1.0 Heats up my cards more than Heaven and good stress tester.
    Pass them both looping overnight and your golden.
     
  4. Pill Monster

    Pill Monster Banned

    Messages:
    25,211
    Likes Received:
    9
    GPU:
    7950 Vapor-X 1100/1500
    This should do it....

    [​IMG]
     

  5. Emille

    Emille Guest

    Messages:
    785
    Likes Received:
    27
    GPU:
    1080 Ti Aorus Extreme
    Don't use furmark or kombustor.


    They are not accurate depictions of real world load, temps or anything else.

    Heaven benchmark uses 99% gpu usage, has good graphics to go with it so you aren't benching something that is technically demanding but looks like crap over and over and has real world temps.

    With my card with default fan profile ( reference cooling ) and my overclock the max temp is 73c...this is a real world example of max gpu usage and temps.

    If you run furmark you will have +20c and any 'instability' it gives off may be unique to that bench and nothing else.
     
  6. wasteomind

    wasteomind Guest

    Messages:
    398
    Likes Received:
    14
    GPU:
    EVGA RTX 3090 FTW3
    That said, my cards will eventually be under water (just need to get some blocks when money frees up), but am I better off leaving the fan profile at default and ocing based off of those thermals or should I up them to 100% (or the highest that it doesn't sound like a jet engine) and tune for that?

    Will temp be a huge factor like it used to be?
     
  7. ---TK---

    ---TK--- Guest

    Messages:
    22,104
    Likes Received:
    3
    GPU:
    2x 980Ti Gaming 1430/7296
    What are your temps on both cards on default fan profile?
     
  8. wasteomind

    wasteomind Guest

    Messages:
    398
    Likes Received:
    14
    GPU:
    EVGA RTX 3090 FTW3
    Hard to say exactly. While running Heaven maxed out at the SC factory clocks of

    • 1266MHz Base Clock
    • 1367MHz Boost Clock
    • 7010MHz Memory Clock
    I'm hitting high 60s low 70s on the default fan profile. There is about a 10c delta between the top card and bottom card in my TH10 case.



    Right now just to try it out i'm using 75% speed on the fans (which is loud), on a 1493 boost clock. Top card is 57c bottom is 47. I reckon I could dial it back some but I'm trying to establish what is stable without an instant crash. What is the the best method for this?
     
  9. ---TK---

    ---TK--- Guest

    Messages:
    22,104
    Likes Received:
    3
    GPU:
    2x 980Ti Gaming 1430/7296
    Do a custom fan profile and the maximum fanspeed that is bearable to you and report back on the temps. Looks like you can knock it down from 75% max based on your temps.
     
  10. eclap

    eclap Banned

    Messages:
    31,468
    Likes Received:
    4
    GPU:
    Palit GR 1080 2000/11000
    Valley and Heaven, as mentioned by TK. Best loop them for a while. I also find 3dmark11 will artifact way before anything else.
     

  11. wasteomind

    wasteomind Guest

    Messages:
    398
    Likes Received:
    14
    GPU:
    EVGA RTX 3090 FTW3
    Going up to +130 on the core +250 on mem crashed the driver. I am backing off to 115 on the core and running again. Lowered fan speed to 50% which is much more bareable and currently seeing temps 64c top, 52c bottom on Heaven maxed out 2560x1440.

    Ok that run lasted longer about 6 mins versus 3mins before. 66c top 52c bottom before driver crashed. Looks like i gotta back off some more. Am I going about this backwards?
     
    Last edited: Oct 11, 2014
  12. ---TK---

    ---TK--- Guest

    Messages:
    22,104
    Likes Received:
    3
    GPU:
    2x 980Ti Gaming 1430/7296
    You are a long way from stable if heaven only woorks for 6 minutes. OC 1 thing at a time ie core first. The way your doing it could be core or vram speed.
     
  13. wasteomind

    wasteomind Guest

    Messages:
    398
    Likes Received:
    14
    GPU:
    EVGA RTX 3090 FTW3
    Ok mem is back to default core is at +110. This puts boost at 1481mhz. Guess I'll see how long it survives.
     
  14. ---TK---

    ---TK--- Guest

    Messages:
    22,104
    Likes Received:
    3
    GPU:
    2x 980Ti Gaming 1430/7296
    Does maxwell oc in 13mhz increments? Knock it down 13mhz or 26 at a time when your oc fails.
     
  15. wasteomind

    wasteomind Guest

    Messages:
    398
    Likes Received:
    14
    GPU:
    EVGA RTX 3090 FTW3
    No crash after 20 mins. Temps 65c and 52c. Yes it appears 13mhz increments give or take on rounding. To test I set a value on afterburner, then launch Heaven to see what the clocks are once loaded. Here is a break down of what some quick testing results in on my cards:

    +139 puts me at 1519mhz.
    +127 puts me at 1506mhz.
    +114 had me a 1493mhz.
    +101 has me at 1481mhz.
    +89 brings me to 1468mhz.
    +76 results in 1455mhz.
    +63 goes to 1443mhz.

    Nothing in between these values seems to result in different clocks, at least as far as the boost clock is concerned as reported via afterburner's g15 lcd app monitoring.
     
    Last edited: Oct 11, 2014

  16. Anarion

    Anarion Ancient Guru

    Messages:
    13,599
    Likes Received:
    387
    GPU:
    GeForce RTX 3060 Ti
    No, it does not use 13 MHz steps.
     
  17. seaplane pilot

    seaplane pilot Guest

    Messages:
    1,295
    Likes Received:
    2
    GPU:
    2080Ti Strix
    Use a synthetic benchmark to obtain the base clocks and voltages then play a game known to be stable for an hr or so. If you are stable up the clocks x10MHz until you crash. Repeat and rinse.
     
  18. -Tj-

    -Tj- Ancient Guru

    Messages:
    18,103
    Likes Received:
    2,606
    GPU:
    3080TI iChill Black
    Try Resident Evil 6 benchmark or if you have HardReset - try its benchmark with 4xmsaa if it passes both few rounds it will pass everything.

    Alien Isolation benchmark is also a good OC test, but personally I found those 2 above the toughest so far.
     
    Last edited: Oct 11, 2014
  19. wasteomind

    wasteomind Guest

    Messages:
    398
    Likes Received:
    14
    GPU:
    EVGA RTX 3090 FTW3
    Will definitely try these two soon/next. Right now I've found that 3dmark Skydiver bench is showing instability where Heaven and Valley are not. Sometimes when it completes it crashes the driver. I had found some settings that seemed to be decent, but then Skydiver smashed them. So I lowered my oc and am now working back up to find out where the instability starts. Once i can pass this with no issues I'll move on to the ones you've suggested.
     
  20. looniam

    looniam Guest

    Messages:
    207
    Likes Received:
    15
    GPU:
    RTX3060 bcuz
    personally i like ~9 runs in mertoLL's bench but nothing seems to be better than a few hours of gaming . . . . . just saying.
     

Share This Page