Stability testing for 7970? This card is a pain to OC...

Discussion in 'Video Cards and Graphics' started by hans030390, Nov 7, 2012.

  1. hans030390

    hans030390 Diamond Member

    Joined:
    Feb 3, 2005
    Messages:
    7,327
    Likes Received:
    0
    I've been trying to OC my Powercolor 7970 (reference design with custom cooler, and this is the revision that includes a long heatsink over the VRMs), but I'm not sure how to go about doing this. With Powertune or Powerplay (I don't remember which it is) on, any program like OCCT or Furmark will throttle over time. You can watch the framerate drop lower and lower. I'm not sure if it's reducing amperage, voltage, clock speeds...either way, doesn't matter.

    That makes it a bit difficult to really test my 7970 for stability. I love testing with OCCT, as it tests for errors for you. Almost every time I thought I had a stable GPU OC from typical stress testing methods (3DMark, Heaven, intense gaming, etc.), OCCT will still detect errors. So, OCCT is my go-to stress testing application for my GPU, as it allows me to easily narrow down on a nearly 100% stable OC. But, since the 7970 throttles on OCCT, I can't a 100% accurate representation on whether or not it's actually stable at full speeds under a full load. Might run for 20+ minutes without errors, but its throttling is so strong that I don't believe this is a good way to test for stability on the 7970.

    Thankfully, I found out how to disable throttling (Powerplay or powertune...whichever) through the MSI Afterburner. Great! My card no longer throttles on OCCT. GPU core temps are just fine. In theory, I should now be able to test my GPU as I'd like...

    Except...Oh, great, now my 7970 crashes after a short period of time (usually just goes to a black screen and the fans spin down to idle levels). I did a bit of further digging...turns out my GPU always does this when the VRM temps hit 115C, and they shoot up to this level fairly quickly in this scenario. This is even with my cooler fans at 100%. They don't get nearly this hot when the card throttles. This is the VRM temp 1 reading, not temp 2. Temp 2 readings are usually great. I'm not sure which VRM(s) correspond to which GPU-Z reading.

    I even tried buying aftermarket heatsinks for the lone VRM that is NOT covered already. All of my VRMs should be good, from what I can tell. So, for one, I have NO idea why my VRMs get so hot on my 7970. And, now, I'm stuck in a situation in which I apparently can't utilize OCCT to test my 7970 for stability no matter what I do. If I let it throttle, I can't get an accurate idea if it's stable at full speeds under a full load. If I don't let it throttle, the VRM temps shoot up too quickly and too high even with aftermarket cooling.

    How is everyone overclocking and testing their 7970s for stability? This is just so frustrating compared to my previous cards...I've never had to deal with ridiculous VRM temps before or PITA throttling.

    As a side note, I flashed my 7970 to the GHz edition BIOS. Works without issues. Before that, I would usually max out the CCC overclock at 1125/1575 with the standard 7970 BIOS. Seemed to run games without issue, but, again, I was never able to fully verify its stability. Just makes it more frustrating because it seems like I have a card that has the potential to overclock decently...
     
  2. Eureka

    Eureka Diamond Member

    Joined:
    Sep 6, 2005
    Messages:
    3,828
    Likes Received:
    0
    7970 is hit or miss. Use Sapphire Trixx to overclock. I can't get past 1050 on stock voltage, I need to up it to about 1.3V to get to 1125, and that's not stable in OCCT (but good enough to game in).
     
  3. Elfear

    Elfear VC&G Moderator
    Super Moderator

    Joined:
    May 30, 2004
    Messages:
    6,465
    Likes Received:
    1
    I'd test with games. I've found some of the error testing programs to be unreliable. IMO, the best ones to use are Crysis, Stalker COP benchmark, and Heaven 3.0.
     
  4. hans030390

    hans030390 Diamond Member

    Joined:
    Feb 3, 2005
    Messages:
    7,327
    Likes Received:
    0
    Eureka, do you test in OCCT as-is, so that the card throttles?

    Elfear, have you found any games that cause the 7970 to throttle like it does in OCCT?
     
  5. cmdrdredd

    cmdrdredd Lifer

    Joined:
    Dec 12, 2001
    Messages:
    24,519
    Likes Received:
    3
    I'd try running Heaven for a long loop. Just start it up set it to max settings at your native resolution and let it run for a few hours. If it crashes or anything like that then you're not stable.
     
  6. zaydq

    zaydq Senior member

    Joined:
    Jul 8, 2012
    Messages:
    782
    Likes Received:
    0
    What voltages are you running through the card?
     
  7. Eureka

    Eureka Diamond Member

    Joined:
    Sep 6, 2005
    Messages:
    3,828
    Likes Received:
    0
    What do you mean? I just use OCCT and run the gpu test. I only let it run about 10 minutes before I'm satisfied, and usually my OCs fail before then.
     
  8. Silverforce11

    Joined:
    Feb 19, 2009
    Messages:
    10,458
    Likes Received:
    1
    Heaven loop. Stable in this = never crashed in any game i play due to OC.

    I am surprised your VRM hit 115C, i have a similar model, twin fan? If you meant ref as in shrouded single fan, then yah, dont push it.
     
  9. Elfear

    Elfear VC&G Moderator
    Super Moderator

    Joined:
    May 30, 2004
    Messages:
    6,465
    Likes Received:
    1
    No games caused throttling until I started hitting 1.3V or higher. It's been awhile though and I don't think I ever tried disabling Powertune at that point.
     
  10. exar333

    exar333 Diamond Member

    Joined:
    Feb 7, 2004
    Messages:
    8,503
    Likes Received:
    2
    This. Same issue with my 7950 I played with. Rock solid up to 950mhz and then no matter the voltage it wouldn't go higher. You may just need to play around, or you may be at your limit. Don't listen to the folks who say '1300mhz is the norm'. It just isn't for all cards.
     
  11. badb0y

    badb0y Diamond Member

    Joined:
    Feb 22, 2010
    Messages:
    3,934
    Likes Received:
    0
    1300 is not the norm for this card

    I suspect most cards will fall between 1100-1200 range.
     
  12. hans030390

    hans030390 Diamond Member

    Joined:
    Feb 3, 2005
    Messages:
    7,327
    Likes Received:
    0
    Well, I believe the stock 7970 GHz BIOS bumps it up to 1.2V max by default, which is normal. That's what my card does now after I flashed it (1.175V stock 7970 BIOS).

    My framerate stops dropping over time in the test due to throttling. It's from Powerplay or Powertune...one of the two. If you disable it in MSI AB, the framerate won't drop over time.

    Yeah, I might just have to do that...

    Yes, twin fan. The PCB looks like the reference PCB, though, from what I can tell. I have no idea why the VRM temp goes so high so quick when I have the throttling disabled. I mean, maybe that is normal. Most people don't seem to disable Powerplay/powertune. It's just the temp 1 reading in GPU-Z. The 2nd temp reading for VRMs is quite cool. Aftermarket heatsink on the uncovered VRM (left side of card) didn't help any.
     
  13. raghu78

    raghu78 Platinum Member

    Joined:
    Aug 23, 2012
    Messages:
    2,859
    Likes Received:
    36
    OP stability testing is better done with normal apps like Heaven 3.0, 3D Mark 11, BF3, Crysis 2. Also your VRM temps are very high at 115c. You have to restrict VRM temps to 90 - 95c. but do not use furmark, occt, msi kombustor. These apps are unrealistic and are intended to max out the chips' power consumption. You don't need to run the HD 7970 Ghz BIOS as it bumps up voltage. Stick to the ref HD 7970 BIOS. If you were able to run at 1125 Mhz with the stock voltage using the ref HD 7970 BIOS thats very good. :thumbsup:
     
  14. aaksheytalwar

    aaksheytalwar Diamond Member

    Joined:
    Feb 17, 2012
    Messages:
    3,384
    Likes Received:
    0
    Just use intensive games for 15-30 mins or so in which temps don't go beyond 115 in this much time. If it is fine then just relax.
     
  15. MrK6

    MrK6 Diamond Member

    Joined:
    Aug 9, 2004
    Messages:
    4,438
    Likes Received:
    0
    Furmark, OCCT, Kombustor and all similar programs are crap for OC testing. They rely on power viruses to put artificially high thermal loads on the GPU's to indirectly test stability. This is a poor method of testing overclocks because A) they don't test all the sub-systems of a modern GPU and B) don't provide an accurate environment for testing. In the end you'll figure out only how good your heatsink is, not your overclock.

    As others mentioned, modern, demanding games are the way to go. Just play or loop something that's DX11 and very demanding, keeping your GPU pegged at 100% usage. My favorite is BF3 atm, but Crysis 2 is also a good choice. Also remember to divide and conquer in your overclocking approach, focusing on the core only first, then the memory only, then finally combining the two. Due to error correction it can be very tricky to pin-point a maximum stable memory overclock with a GPU overclock acting as a confounding factor.

    Also, the GHz edition BIOS is useless. It provides an unnecessarily high amount of voltage at stock that simply wastes power.
     
  16. VulgarDisplay

    VulgarDisplay Diamond Member

    Joined:
    Apr 3, 2009
    Messages:
    6,194
    Likes Received:
    1
    I have never seen anyone claim that 1300mhz is the normal OC on air cooling for a 7970. 1125 is pretty much a given though.
     
  17. hans030390

    hans030390 Diamond Member

    Joined:
    Feb 3, 2005
    Messages:
    7,327
    Likes Received:
    0
    I do want to make it clear that my VRM temps (from reading 1 in GPU-Z) are only high when running OCCT or Furmark WITH throttling disabled through MSI Afterburner (whether that is Powertune or Powerplay, I don't know exactly which). If throttling is allowed to occur, they seem to stay below 90...high 70s or mid 80s seems average.

    I guess I'll be forced to test stability without OCCT. I just love the error checking feature. Has been much more reliable for me than testing through games ever has been, but I'll just have to make do! I'm pretty picking about ensuring 100% stability on my GPU, much like people are with their CPUs.

    I'll also stick with the GHZ edition BIOS. The extra voltage is no problem if throttling is allowed to occur like normal. If anything, it'll allow me to take it a bit further without having to even mess with a voltage setting.
     
  18. AdamK47

    AdamK47 Lifer

    Joined:
    Oct 9, 1999
    Messages:
    11,957
    Likes Received:
    24
    The three of my cards are set to 1000MHz core and 1500MHz memory. With all three of the GPUs undervolted to 1.100V. Runs totally stable, cool, and I don't have to worry about problems while playing anything. I suppose the older I get the less aggressive I get with my overclocks.
     
  19. lavaheadache

    lavaheadache Diamond Member

    Joined:
    Jan 28, 2005
    Messages:
    6,882
    Likes Received:
    0
    I have had 3 7970's so far. Still have 2. My Lightning is capable of 1300+mhz. My original launch Gigabyte reference card could bench at 1300 but it wasn't 100% stable there. I ran it at 1250 with 1.25v for 6 months or so with zero issues. Now my buddy is running it at the same settings. Lastly, I just picked up a new MSI reference card that I have not really had time to OC much yet but did slide both the mem and core sliders all the way to the right(1125/1575) 3 days ago with no voltage adjustments besides the 20% power tune slider. I have played BF3 for around 3 hours trouble free. I don't doubt I will get 1200 mhz out of it.

    I would say since all 3 of the cards I have owned can do a minimum of 20% on the core that getting a card that can't do 1100 is very unlucky.

    ***edit***

    Just remembered that I did purchase a Powercolor Vortex 7970 that was DOA. Card displayed artifacts out of box , with a broken fan blade. meh... Take it as you will.
     
  20. Eureka

    Eureka Diamond Member

    Joined:
    Sep 6, 2005
    Messages:
    3,828
    Likes Received:
    0

    What are the ASIC quality of your cards? I would say that your cards aren't the norm, but then again you also bought cards from higher binned brands (MSI and Gigabyte). These card are sold as overclockers and the company reject chips that don't meet the requirements.
     
  21. lavaheadache

    lavaheadache Diamond Member

    Joined:
    Jan 28, 2005
    Messages:
    6,882
    Likes Received:
    0
    Lightning is 60%, Gigabyte was 66% and my MSI is 80%(it also has a low stock VID).

    I firmly believe asic quality plays a big role in vcore to overclock ratio. My lightning needs extra juice to hit 1125 and 1.275 for 1200mhz but will run over 1300 mhz with 1.33v. It does this with temps in the low 70's though ( the cooler is a beast).

    My old Gigabyte card could run up to 1175 mhz @ stock volts(1.175v) and 1250mhz @ 1.25 volts. It really had a 1:1 ratio like that. My Msi ref card so far is running 1125 mhz 1.125 Default vcore but I suspect it won't do much higher @ stock voltage.