• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

Drivers are stupid

Devrij

Devrij

Devrij

Soldato
Joined
23 Jul 2009
Posts
14,134
Location
Bath
So, I'm trying to iron out my gpu overclock, and just when I think I have it perfect: driver stops responding and has recovered message, gpu now clocks to 500mhz and my life is ruined. Reboot and try again.

Now WHY can't the driver recover successfully like it says it does and then continue to function normally at full 3d clocks without requiring a reboot? They used to, but no longer!

If anyone knows how to achieve this, please tell me, but otherwise
/rant

On a side note, it seems my IceQ boost 7950 is temp limited to about 78C at which point this happens. 1200/1450 no problems @1.23V but whether it's that or 1130/1450 @ 1.17V, once that needle hits 78+ it crashes. I'm guessing my VRMs are hot or something but no idea because HIS have disabled the sensor. Hmmm
 
Sounds like you're on the edge of stability so just knock everything back by 25 MHz and enjoy. Not much point running on the edge unless you're benching.

Unless your overclock is perfectly stable on another set of drivers it's unlikely to be that causing it. Mine are rock solid on 13.1 drivers at 1200/1600 but on 13.3 beta's in some games I struggle to keep 1100/1500 stable.
 
It is a driver thing, iirc it happens on Nvidia too, driver crashes and recovers, you'll still need to reboot to get everything back to normal.

Set a custom fan curve to offset the temps?
 
andybird123 said:
the most I've ever had to do is close precision and restart it, doesn't normally need a whole PC restart on nvidia :D
Click to expand...

Most of the time this is the case, although with SLI I sometimes find that 1 card reverts to stock and the other is still OC'd. I restart AB to apply the overclock and it flips it around, so I end up restarting :p

That could just be AB though, not tried EVGA's PX as an alternative.
 
I can't 100% remember with NV, been nearly a year since I had my 670!! I know I had to restart something when the driver recovered to get it to clock correctly again.
 
turn your fan up, my 7970 will do 1200/1650 but it like to be under 63 degrees to do it. Your card has a decent fan, turn it up!! I set a custom fan profile in trixx it's really easy
 
Last edited:
Deceptor said:
Most of the time this is the case, although with SLI I sometimes find that 1 card reverts to stock and the other is still OC'd. I restart AB to apply the overclock and it flips it around, so I end up restarting :p

That could just be AB though, not tried EVGA's PX as an alternative.
Click to expand...

I use precision, when I hit a problem I just turn on "sync", hit "default" which sets them all back to normal and then I click the preset I want to start from and apply and I'm right back to my 24/7 safe OC... from there I can turn off sync if I want to tweak cards individually

if it gets itself in a twist then I have to stop precision and reopen it, but not normally
 
Rusty0611 said:
Sounds like you're on the edge of stability so just knock everything back by 25 MHz and enjoy. Not much point running on the edge unless you're benching.

Unless your overclock is perfectly stable on another set of drivers it's unlikely to be that causing it. Mine are rock solid on 13.1 drivers at 1200/1600 but on 13.3 beta's in some games I struggle to keep 1100/1500 stable.
Click to expand...

Interesting, I'm running the betas so might roll back to 13.1 and see if that makes a difference.

What's odd is that I can drop clocks loads and still get the issue unless I keep the card cool. Currently it runs high 70s at 55% fans, and they get a bit noisy above that. If rolling back doesn't help I'll see what a higher fan speed does. Maybe the VRMs need more airflow or something.
 
Never had an issue that a overclock is stable on one set of drivers but not the other set. The only time ive seen this occur is when changing or updating the software ive used to overclock the cards, afterburner, trixxx etc.
 
LtMatt said:
Never had an issue that a overclock is stable on one set of drivers but not the other set. The only time ive seen this occur is when changing or updating the software ive used to overclock the cards, afterburner, trixxx etc.
Click to expand...

I've never had a single card overclock worse on new drivers, i have had crossfire clock worse/better on different drivers.
 
I can't say I've ever had it make a difference either, but it could just be driver issues with the games (unlikely). I can bench as many runs of heaven as I like at 1200 core, but games or anything that pushes temps harder seem to cause driver crashes on anything over 1140 no matter what volts I run. I'm pretty convinced there's a correlation between temps and these crashes. Maybe I'll run 100% fans and see what happens lol.
 
Devrij said:
I can't say I've ever had it make a difference either, but it could just be driver issues with the games (unlikely). I can bench as many runs of heaven as I like at 1200 core, but games or anything that pushes temps harder seem to cause driver crashes on anything over 1140 no matter what volts I run. I'm pretty convinced there's a correlation between temps and these crashes. Maybe I'll run 100% fans and see what happens lol.
Click to expand...

You're spot on with that. Once you start going above or around 1150 temps play an issue in stability, even more so if you have a high asic card. That's why generall all the best clocking cards have a lower asic as they handle heat better due to leakage.

What temps are getting? I find i always need another notch or two of voltage to be game stable as opposed to bench stable.
 
LtMatt said:
You're spot on with that. Once you start going above or around 1150 temps play an issue in stability, even more so if you have a high asic card. That's why generall all the best clocking cards have a lower asic as they handle heat better due to leakage.

What temps are getting? I find i always need another notch or two of voltage to be game stable as opposed to bench stable.
Click to expand...

At 1200/1450 1.23V I was getting 76C with 100% fans (79C with 60%) while playing Chivalry until it crashed out. BF3 runs a bit cooler, but there are definitely stability issues with the occasional vertex flying out of someone's gun. Under 1150 seems okay though despite similar temps ranging up to 78C

E: more volts doesn't fix the problem either. It still happens at 1150 with 1.2V despite being 100% stable at 1142 with 1.18V
 
Devrij said:
At 1200/1450 1.23V I was getting 76C with 100% fans (79C with 60%) while playing Chivalry until it crashed out. BF3 runs a bit cooler, but there are definitely stability issues with the occasional vertex flying out of someone's gun. Under 1150 seems okay though despite similar temps ranging up to 78C

E: more volts doesn't fix the problem either. It still happens at 1150 with 1.2V despite being 100% stable at 1142 with 1.18V
Click to expand...

Yes you probably need to be keeping the card in the low 70's or even lower at those clocks. Some cards can handle it, some can't. I used to see artifacts at 72c with my core at 1225 on the stock cooler.
 
Temps it is then. Maybe time to pick up an aftermarket cooler then. Might be worth it if I get an extra 100mhz out of it and keep it quiet at the same time.
 
I know it has to do with stability. What's stupid is that when they recover they don't go to full clocks. This wasn't always the case with amd drivers, so I think it's stupid that you are forced to reboot every time you test a clock that turns out to be unstable. It's a minor gripe since SSDs make it more bearable, but it still bugs me that I can't twiddle the voltage a bit and keep testing.

I am, however, stuck on getting my card stable over 1150mhz, which is what is causing me to rage at this particular niggle.
 
Last edited:
You must log in or register to reply here.
Back
Top Bottom