1. Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.

6950 artifacts in aero/browser only at low voltage changes

Discussion in 'AMD / ATI' started by BlackZero, Oct 25, 2011.

  1. BlackZero

    BlackZero New Member

    Joined:
    Jan 28, 2011
    Messages:
    168 (0.13/day)
    Thanks Received:
    103
    Location:
    England
    *The article title should be '6950 artifacts in aero/browser only at low voltage changes'


    So as the title says i have suddenly started getting artifacts (small red/green dots) on the aero tabs and in the browser when ever I change the voltage to a higher number, this happens whether I use a hd 6950 stock/shader unlocked or a 6970 bios. The strange thing is that it will play 3d games pass furmark/atitool without issue, and only gets these little dots all of a sudden in the aero taskbars and in the browser whener the clock/voltage changes to different power states.

    I recently added a full cover waterblock to my gpu so the temps are very low, something like 30c on the gpu and even lower vrm temps according to gpu-z, I did try try the full hd 6970 bios in the past when I first got the card and don't remember getting the coloured dots though that was only for a short time and I only really tested in furmark/games beofre going to a shader unlocked bios instead.

    I also think that this may have something to do with the different power states and clock rates as it only seems to effect items like the browser and aero taskbars. I have also noticed in gpu-z that when browsing the memory clock drops to 625mhz somestimes while the bios only specifies 150 in extreme low power mode and 1250 for all the other modes. maybe this is the reason for the anomolous artifacts?

    I am quite stumped so any help would be appreciated.


    image of gpu-z with the memory clock at 625mhz highlited

    [​IMG]
    Last edited: Oct 25, 2011
  2. cadaveca

    cadaveca My name is Dave

    Joined:
    Apr 10, 2006
    Messages:
    13,743 (4.56/day)
    Thanks Received:
    6,767
    Location:
    Edmonton, Alberta
    Your GPU has a coldbug at air/water temps. Put the stock cooler back on, or go back to the stock BIOS.
  3. BlackZero

    BlackZero New Member

    Joined:
    Jan 28, 2011
    Messages:
    168 (0.13/day)
    Thanks Received:
    103
    Location:
    England
    Well I've moved back to the stock bios now but still get above issues when I increase voltage, only happens when i increase voltage above 1.16v, happens regardless of clocks used. It also only happens when the clocks jump from low power to full power mode and doesn't happen if clocks stay in either mode.
  4. brandonwh64

    brandonwh64 Addicted to Bacon and StarCrunches!!!

    Joined:
    Sep 6, 2009
    Messages:
    18,414 (10.41/day)
    Thanks Received:
    5,983
    Location:
    Chatsworth, GA
    sounds like RMA time, How long have you had this card?
    Crunching for Team TPU
  5. BlackZero

    BlackZero New Member

    Joined:
    Jan 28, 2011
    Messages:
    168 (0.13/day)
    Thanks Received:
    103
    Location:
    England
    I've had it for around 8 months, but Like I said it doesn't happen at stock clocks/voltage only happens when I go above 1.16v and not in 3d only when it changes power states during web browsing/aero.
  6. brandonwh64

    brandonwh64 Addicted to Bacon and StarCrunches!!!

    Joined:
    Sep 6, 2009
    Messages:
    18,414 (10.41/day)
    Thanks Received:
    5,983
    Location:
    Chatsworth, GA
    Maybe it cannot handle the higher voltages. What is stock voltage?
    Crunching for Team TPU
  7. BlackZero

    BlackZero New Member

    Joined:
    Jan 28, 2011
    Messages:
    168 (0.13/day)
    Thanks Received:
    103
    Location:
    England
    yes well i don't know because it will run 3d such as furmark/games etc. without issue, only when changing voltages through powerplay does this happen, stock voltage is 1.1v
  8. TimoX New Member

    Joined:
    Mar 10, 2011
    Messages:
    49 (0.04/day)
    Thanks Received:
    14
    What browser and driver versions?

    Also are you using Hardware acceleration for the browser?
    IF so, what happens if you disable it?


    That 625 memory clock is actually normal for UVD stuff. The drivers halve the clock speeds sometimes. it's half of the default 1250.

    It's easiest to see that in action doing something like playing a video in WMP full screen and running gpu-z.
    The majority of the video usually will sit at 250/150, but sometimes UVD kicks in and it'll fluctuate from 750/1250, to sometimes 500/650, etc...

    I adjusted my UVD lower limits down to 400/1100 and sometimes the memory drops to 550.

    what happens if you completely disable CCC from loading at startup?
  9. BlackZero

    BlackZero New Member

    Joined:
    Jan 28, 2011
    Messages:
    168 (0.13/day)
    Thanks Received:
    103
    Location:
    England
    Tried catalyst 11.9/11.10 pre3 and got same results. Using firefox with hw accleleration disabled, I've even diabled flash hardware acceleration but uvd still kicks in at times. The problem is that before changing the stock cooler to the ek waterblock I used to be able to go upto 1.25 to 1.27v before I got little dots and always assumed it's because of vreg temps so used to lower it to 1.23v when submitting bench results at overclocked frequencies. Now at 1.2v the dots kick in pretty quickly but the vreg temps are much lower as they are water cooled, got me baffled.

    Even at 1.175v I get dots after a while on the desktop regardless of clock rates used, happens only when voltages jump from low power to high power states, the green/red dots then go away after a while.

    I haven't tried with ccc disabled as I needed to use powercontrol for the higher clocks, will try with CCC disabled from starting in msconfig.
  10. BlackZero

    BlackZero New Member

    Joined:
    Jan 28, 2011
    Messages:
    168 (0.13/day)
    Thanks Received:
    103
    Location:
    England
    Tried with ccc/amd external utilities disabled and a fe minutes after enabling the higher voltage got a few artifacts/dots on the lower aero taskbar when browsing through the web, they went away soon after and haven't seen any since (around 20 minutes). so doesn't seem to be related to ccc as such.
  11. TimoX New Member

    Joined:
    Mar 10, 2011
    Messages:
    49 (0.04/day)
    Thanks Received:
    14
    Well, you've ruled out a lot of stuff.

    At this point I'm leaning towards either an odd cold bug type issue cadaveca mentioned, or maybe something on that waterblock is not sitting right.
    maybe one of the memory chips or something isn't quite getting good contact, causing it to run a bit hotter???

    What happens if you change the auto fan to a manual speed, say 50% in CCC.
    just trying to rule out some odd fan speed issue since the fan is no longer connected.

    Maybe one or more of the vrm's isn't getting cooled properly, even though gpuz says it is cool?

    If things used to work fine, and now the only real change is the waterblock, i'd consider pulling it and making sure everything has good/proper contact, just in case somehow one thing, either vrm or mem chip is missing paste/pads on it.
    Double check some of the diagrams out there in case one little thing is missing.
  12. BlackZero

    BlackZero New Member

    Joined:
    Jan 28, 2011
    Messages:
    168 (0.13/day)
    Thanks Received:
    103
    Location:
    England
    Funnily enough i've considered putting the fan to 50% aswell and will try that, also the dots seem to happen regardless of clock speeds and as the mem voltages don't change it probably isn't mem related. The vrm temps are probably measured at a certain point and there's quite a few of them so it's quite possible that not all of them are properly cooled. Considering taking the block apart as a last resort.

    Maybe there's a burn in period for the block as I had to apply a drop of tim to each vrm before adding the thermal pads, this was suggested by ek. but yeah I think i'll have to take it apart to rule everything out.
  13. BlackZero

    BlackZero New Member

    Joined:
    Jan 28, 2011
    Messages:
    168 (0.13/day)
    Thanks Received:
    103
    Location:
    England
    I have since taken my waterblock apart and refitted it a lot more carefully, now my load temps are down by 6-7c but at the same time, compared to 1.17v before, I now get artifacts in aero/bowser at 1.16v.. I am starting to think it probably is some sort of cold bug as reducing temps seem to make it appear faster.

    On the plus side I can now run 880/1425 at 1.1v though it's kinds annoying that having it under water I can't overclock it. I also did a dual boot install of windows xp and upto 1.2v i got no artifacts either in 3d or browsing/desktop. This suggests that the issues in windows 7 are triggered by aero at lower voltages, would it be advisable to disable aero and run it at anything below 1.2v? or are the artifacts at low temps/voltage an indicator of a larger issue and there's a chance of damaging the card, seems to run fine in xp so If I didn't have aero running I would not see any artifacts/issues at such low temps until I went over 1.2v anyway.
  14. TimoX New Member

    Joined:
    Mar 10, 2011
    Messages:
    49 (0.04/day)
    Thanks Received:
    14
    At least it runs cooler. :)

    This is really an odd issue. I'm guessing you already have, but have you double checked other 6950/70 owners using WC to see if they have these issues?

    The odd thing is that those voltage changes aren't that much. I set custom voltages in RBE my 2D voltage is only .875 instead of .900, and my 3D is set at 1.2v. I think my other voltage is 1.0. Even though I'm on the stock cooler, I don't notice any fluctuations or glitches when clocks change. the only way I know clocks have changes is if they stay in 3D for a bit, i hear the fan ramp up since I have AB running it.

    I did notice that in XP, it seemed like the clocks would fluctuate more than in 7, which would keep it slightly warmer, which may explain why it does that for you.

    As for Aero, you could try disabling it to test it out. For me, Aero pretty much stays at 2D 250/150, even using peek. Only if i do something like scroll Fx quickly, do clocks jump to 3D for a bit, then drop back down.

    What drivers in XP, and what are you using to change clocks and voltages with?

    The fact that running it in XP it is fine, leads me away from a hardware defect, and towards an odd software glitch.
  15. cadaveca

    cadaveca My name is Dave

    Joined:
    Apr 10, 2006
    Messages:
    13,743 (4.56/day)
    Thanks Received:
    6,767
    Location:
    Edmonton, Alberta
    It could be a software glitch, for sure. when I ran into similar issues, I did try all sort of things to aviod the problem, but other cards I had did not exhibit the same behavior. I only noticed it when the GPU was less than 40c.

    And yes, it's weird, for sure, and that's why I was so quick to say what I did. I'll never forget that. It's also very similar to other behavior i've seen running sub-zero on AMD GPUs.

    Personally, i think the problem is the memory controller. Load from Aero is much different than XP desktop, which is mostly done by the CPU.
  16. BlackZero

    BlackZero New Member

    Joined:
    Jan 28, 2011
    Messages:
    168 (0.13/day)
    Thanks Received:
    103
    Location:
    England
    I did check with another user with a 6950 watercooled, he's having no issues. In windows xp I was using catalyst 10.11 pre 3 aswell, used sapphire trix to overclock/set voltages. I only get the little artifacts/dots in eitehr the browser or on aero tabs and this only happens when the clocks jump from 2d to 3d while browsing/looking at pictures but nothing in full 3d so it definately seems to be related to some sort of sudden voltage changes not being tolerated or someother software bug.

    I have also put the heating up in my flat just to check and now the card is idling at 34-35c instead of 30-31c and I haven't seen any artifcats even fater flicking through lots of pictures. Very strange issue. Will continue to check at these temps for a while and then check in windows xp again.

    ps. It happens even with clocks at default/stock with only the voltage raised so I don't think it's a memory controller issue.
    Last edited: Oct 28, 2011
  17. TimoX New Member

    Joined:
    Mar 10, 2011
    Messages:
    49 (0.04/day)
    Thanks Received:
    14
    yeah, def a weird one.
    what happens if you disable trixx and don't use it?

    Just trying to rule that out.
    Obviously you'll have stock 3D clocks, but this is just to test/rule out trixx.
  18. BlackZero

    BlackZero New Member

    Joined:
    Jan 28, 2011
    Messages:
    168 (0.13/day)
    Thanks Received:
    103
    Location:
    England
  19. TimoX New Member

    Joined:
    Mar 10, 2011
    Messages:
    49 (0.04/day)
    Thanks Received:
    14
    Well crap, running out of ideas. :laugh::(

    Ok, a few more things.

    1. I think AB has an option to keep constant voltage. it's there in 2.1, not sure about latest beta. Haven't tried it though. This may keep it from doing that if voltage is constant. It'll run a bit warmer, but nothing that should matter with the cooler on there, and maybe just enough to stop the odd artifacting.


    2. maybe force 3D clocks to stay on. This will be similar to above, but it will run a bit warmer than that due to higher clocks, but voltage will be consistent as well.


    3. IF your BIOS supports it, use RBE to adjust 2D and UVD voltages up a bit. Say 1.0 and 1.10 instead of .900 and 1.00


    Any of the above should keep the card slightly warmer, but nothing to be concerned with since you have the WC.


    As for AB, for me once i set it up, I just dropped a shortcut link in my start-up folder, so it loads at start up. In the settings I have it checked to start minimized. It has a check to load at startup as well, but I didn't use that, since sometimes I'm testing stuff and can easily move the startup link out.
    I'm only using it to control the fan, and I have an OC profile in there as well. Don't use it for voltages since I did them in the BIOS.
  20. BlackZero

    BlackZero New Member

    Joined:
    Jan 28, 2011
    Messages:
    168 (0.13/day)
    Thanks Received:
    103
    Location:
    England
    Yeah I've been considering yusing the constant voltage option too, I was also considering upping the low power voltages but even when using a 6970 bios which has uvd at 1.1v I still had the same issues, though it did also have low power voltage at 0.9v so maybe changing that could help.

    Other than that the only component that could possibly cause the issue with sudden voltage changes/demand would be the psu and im case the psu is the oldest component I have as everything else has been recently updated. I've had my toughpower 850w since mid/late 2007 so it's more than 4 years old now and maybe I'll try a different psu just to rule that out. Any suggestions? I'm looking at the corsair HX 850w but would any other similarly priced psu be more stable in terms of ripple etc and quieter?
  21. BlackZero

    BlackZero New Member

    Joined:
    Jan 28, 2011
    Messages:
    168 (0.13/day)
    Thanks Received:
    103
    Location:
    England
    Ordered a Corsair AX 850w, will know soon enough if the psu was to blame.
  22. BlackZero

    BlackZero New Member

    Joined:
    Jan 28, 2011
    Messages:
    168 (0.13/day)
    Thanks Received:
    103
    Location:
    England
    Installed the ax850w and still getting the 2d artifacts in aero tabs/browser, running out of ideas now.
  23. LordJummy

    LordJummy New Member

    Joined:
    May 13, 2011
    Messages:
    1,406 (1.22/day)
    Thanks Received:
    248
    Location:
    US of A
    RMA / Return to shop. Why should you have to deal with this problem? You're the consumer...

    I seriously bought and returned 4 separate 6950's earlier this year. I would test them and bench them to see which ones were better, and return the ones that were not so good. Fry's FTW.
  24. BlackZero

    BlackZero New Member

    Joined:
    Jan 28, 2011
    Messages:
    168 (0.13/day)
    Thanks Received:
    103
    Location:
    England
    but it's not a stock problem, only when overclocked and I've got it watercooled so no more warranty.


    On a side note I got it fixed, or let's say I found a compromise. I changed the 2d clocks from 250/150 to 500/625 using racerx's tool and now no more 2d artifacts. Can't really explain it but it seems it was caused by too low clocks and too high voltage.
  25. BlackZero

    BlackZero New Member

    Joined:
    Jan 28, 2011
    Messages:
    168 (0.13/day)
    Thanks Received:
    103
    Location:
    England
    Just wanted to put in an update for anyone who stumbles across this thread.

    It was definately a cold bug issue when using watercooling, I have since reverted to using constant voltage to keep the card/vrm a little warmer and all is fine I have noticed from gpu-z monitoring that when the vreg temps are around 25c at idle i get the dots but once they are over 30c the 2d dot artifacts go away. Using slightly higher 2d clocks had the same effect, as in they raised 2d temps but it also meant higher power usage and was a Pita having to adjust the 2d clocks with racerx tool on every driver update so I have now set my bios voltage registers to 1.15v for the 2d/uvd/multimonitor and 1.2v for the 3d/boot register. Still get a little white dot or so when first botting up due to low temps at times but it goes away after a short time.

    Using msi afterburner to force constant voltage would probably have the same effect and should be used to ensure everything is working fine before flashing the bios. I'm now slowly moving my way upto 1.25v as using this has no issues in 3d but same little dots in 2d, once I'm comfortable with it i might just flash all the voltage registers to 1.25v and be done with it.
    AsRock, cadaveca and LordJummy say thanks.

Currently Active Users Viewing This Thread: 1 (0 members and 1 guest)

Share This Page