HD6950 occasional microfreeze

Hi,

I've got the same version HD 6950 as you and when I first installed it I noticed the same thing (I think). Occasionally (and completely randomly) it would freeze for the briefest of moments (barely noticeable) and the wouldn't do it again for ages. Whilst trying to flash the BIOS on the card to unlock extra shaders I completlely removed the card from my system (driver sweeper in safe mode etc.) and then re-installed it. It didn't help me flash the card, but I've not seen the freeze issue since. Not sure if this helps you but if you haven't already tried it, give re-installing the card a go!

When you mean reinstalling the card you mean physically removing it from the motherboard then reseating it, or just deleting and reinstalling the drivers?

I physically removed the card from the motherboard, booted from my integrated graphics to safe mode, ran drive sweeper and erased all traces of ATI drivers, put the card back into the motherboard, rebooted then reinstalled CCC and drivers.

I don't know which part of that process sorted the mini-freeze (just re-installing the drivers without touching the card may have worked too) but when I was done it was all working perfectly!

Thanks. But I think I can't do that, because I have a p67 motherboard which requires an actual graphics card to POST - I can't use my integrated SB graphics and I've got no spare card around. I'll try using drive sweeper though and come back to tell if it worked or not.

I also noticed the drivers on the CD worked better than the generic AMD drivers... but they are pretty dated. I hope this won't be a problem later on since upgrading to latest generic drivers causes the gpu to completely freeze on most games. Maybe Sapphire releases its own drivers, I'll have to check.

Okay I found out what was going on - turns out the GPU core clock sometimes goes crazy and varies between 600 and 800 MHz... (and goes at 250 when idle as normal) and the GPU doesn't like it, hence the microfreezes while the GPU sorts its stuff out. This is also what is causing my complete freezes, when the GPU goes real crazy on the clock rates (why was that other thread deleted btw? the issues were distinct!)

So it's either a failing or insufficient PSU, or just a bad GPU. Anyone know which one or how to check it? (preferably avoiding buying more hardware)

My ideas:
- the GPU is bad
- the PSU is overrated and can't power the 6950 - that would be BS though
- the GPU is not properly seated in the PCI-e lane

can't think of anything else. Opinions?

As a temporary fix I allowed 20% more power to the GPU and underclocked its core clock by 20MHz via ATI Overdrive. Well that didn't work out - turns out the best fix is to run a GPU stress-test while gaming to prevent the GPU losing power (wtf?), but obviously it's not that great for gaming. This is just weird.

I'll get a stronger 750W PSU and see if it gets better, perhaps it's not the PSU but power consumption parameters screwing up or whatever. It's definitely a power problem though. The GPU isn't getting what it needs. We'll see.

What PSU are you using? Unless you've got a lot of power draining hardware or a bad quality PSU then you shouldn't need 750W to run a single 6950. I use a corsair tx650w and its plenty powerful enough.

The card will automatically underclock to save power when its not needed so that's not unusual but mine goes straight to 800 core when I'm gaming. How are you monitoring your clock speed during gaming? I use MSI Afteburner on-screen display.

Also, you'll likely encounter problems using outdated drivers. Usually the CDs that come with graphics cards are next to useless because they are so out of date by the time they reach you. If you're having trouble with the very latest ATI drivers then try going back a month but anything older than that isn't a good idea!

Using Thermaltake Litepower 600W PSU. It should normally be able to power my rig but I am now pretty sure it's holding me back somehow. Here is my hardware:
- p8p67 mobo (asus)
- i5 2500k
- 2x2GB 1333MHz DDR3 dual channel G.Skill RAM
- hd6950
- two 500GB Seagate 7200rpm 3.5" SATA hard drives
- 1 asus 1080p HDMI monitor (connected via DVI to the GPU)
- 1 USB speaker set
- 1 external USB hard drive
- 1 USB wireless network adapter

My card properly underclocks at idle to 250MHz core clock. But during gaming it won't stay at 800, it will fluctuate wildly between 600 and 800. Some games couldn't care less, but others are really sensitive to it and hence microfreeze or plain crash.

I use the GPUZ sensors to know my clock speed/temps.

I'm going to use the newest drivers now that I know it's not a driver problem. Unfortunate coincidence made me believe recent drivers were faulty when it's probably my system having some issue.

Btw I'm running Windows 7 x64, so no outdated OS either.

Maybe I should've gotten that Corsair PSU after all, damn it. There's no cutting corners  

There is no PSU voltage information anywhere in Everest or HWMonitor. Nothing in the Sensors tab. I might just have forgotten to connect the PSU sensors to the motherboard or something. But I didn't see any. arghh

This is strange though, why would a power problem cause a stress-testing benchmark to simply slow down and produce no artifacts while completely crashing the GPU in some games?

The fluctuating clocks within a game (other than going all the way to idle clocks on cut scenes) is definitely a sign of power play getting involved, which I believe is caused by not enough power.

I definitely see a correlation between core clocking and GPU crashing. This is what I get:

When the offending program is started with the GPU being in idle mode (250/150), it will instantly crash the GPU driver, which sometimes requires a manual power-off and prevents the program from launching (the GPU state goes invalid and corrupts the program anyway). However, if the program is started while the GPU is in full power mode (800/1250), it will proceed normally.

These are the programs that are known to me to exhibit this behaviour (others just run normally at all times): Just Cause 2, Oblivion, Mozilla Firefox (wtf?)

I am starting to think these particular programs are actually actively checking for this kind of thing and are not happy with PowerPlay screwing with the GPU clocks. I'm thinking disabling PowerPlay should resolve all issues, but editing the profile XML file messed up my driver settings and had me reinstall CCC, so I still need to figure out how to do it.

I am thinking this is not a PSU problem after all, I mean why would these particular programs crash my GPU while a stress-testing program would just carry on with its stuff? It's not like Firefox requires 3D-grade clocks. It just doesn't work out like a power issue in my opinion.

This is so confusing...

This is getting into gigantic proportions - after finishing transfering my 400GB+ of data from my dvds and external hard drives, I realize I can't even launch half my games and programs without having to run a GPU benchmark at the same time. Even Firefox fails.

I've had enough. It's been days and I'm really pissed now. You know what, I'm just going to replace the PSU and RMA the card. If that doesn't work, I'll sell the 6950 on Trademe and get a GTX 560 Ti. Screw this. My patience has its limits.

Nothing against you guys - I appreciate you helping me and it certainly helped me understand some stuff I didn't know before. But I don't feel like it's my job having to spend my holidays adjusting core clocks and downloading driver after driver, tool after tool just to have a stable GPU.

Best answer selected by Bacterius.