• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

does 1 error in OCCT really mean much?

toyota

toyota

Lifer
I gave up on trying to oc my card since it cant go more than 7% without needing voltage just to be stable in all games. so I decided to stay at stock clocks and undervolt a bit from my stock 1.013 voltage. every game and benchmark was perfectly stable at .963 but I decided to use .975 just to be safe.

well I got curious and decided to run OCCT and got lots of errors after about 5 minutes. I then went with .988 and re ran it and got just 1 error. I did re run it at 1.000 with no errors and that is currently where I am at. I was wondering though since OCCT puts the card in such a massively unrealistic load does that 1 error when using .988 even mean much for actual gaming?
 
Last edited:
I don't use OCCT anymore on gpus, it's similar to the furmark "power virus" and can literally kill a card.

My big test is still crysis demo; if it runs crysis demo it will handle anything that I can ever throw at it, including 4 instances of seti@home running 24/7/365.

btw, I finally threw too much at my gtx 480. 2xSSAA + 4xMSAA + max everything on nwn2 put me down into the low 20's in one particular spot on the skymirror map. I put the vcore back to stock and bumped up the core to 800 just to see what it would do, and my fps went up in an almost linear fashion. And even after a few hours @ 100% it dosn't go over 80c, (76c was the max last night I think), though ambient is a few degrees C lower than usual thanks to our recent cold front. I think I'll try stability testing 850 tonight @ stock vcore on crysis demo/3dmark11, I am seriously amazed at how cool/quiet this card runs.
 
Last edited:
occt and furmark are the only times when my GPU throttles down when I overclock. Tried turning off powertune through ATI tray tools and bam...errors and problems out the wazoo.

No game stresses a GPU as much, not even Crysis 2 or Battlefield 3 maxed.
 
Yeah, those programs aren't like even the most stressful cpu tests (linx, OCCT, etc). On a cpu you can run F&H or many other DC projects and put similar stress on a system, but on a gpu even the toughest game/DC project is nothing like one of those "power virus" applications. Both camps deliberately throttle in furmark, but they probably should add kombustor and OCCT to the list as well.
 
yeah in games I normally hit about 71C but during 10 minutes of OCCT I hit 84C. I could even hear my power supply fan ramping up which never happens. OCCT put more load on just the stock gpu by itself than my whole system has ever used in games even when overclocked. I will probably never run it for more than 5 minutes ever again.

I guess what I am getting at with this thread was to know if OCCT is just detecting those errors because its under very heavy load or not. if its just the insane load on the gpu that is giving 1 error then i see no reason to worry about it and will probably go back to .988 since even .963 was fine in games.
 
Last edited:
So basically what I meant was, if my games run fine for long sessions with no error and 3dmark 11 gets through a few runs I call it good. Furmark is ok though to make sure my fan profiles work.

As for your question directly, I would put my money on OCCT stressing the GPU and memory so much that it errors. Basically forcing an error.
 
Certainly adding the extra vcore will make it "more stable", but since you'll never run that much of a workload through it I wouldn't factor that in when figuring out your max stable or min vcore stable settings. BF3 might be another good one for stability testing, but I dont' have that yet so I can't comment on how it stresses the gpu relative to crysis demo. I do know that crysis stresses it a LOT more than 3dmark11, however, and it's not even close.

cmdrdredd said:
So basically what I meant was, if my games run fine for long sessions with no error and 3dmark 11 gets through a few runs I call it good. Furmark is ok though to make sure my fan profiles work.

As for your question directly, I would put my money on OCCT stressing the GPU and memory so much that it errors. Basically forcing an error.
Click to expand...

That's the issue with those programs, they can/will actually damage your card. Don't use them at all.
 
bryanW1995 said:
Certainly adding the extra vcore will make it "more stable", but since you'll never run that much of a workload through it I wouldn't factor that in when figuring out your max stable or min vcore stable settings. BF3 might be another good one for stability testing, but I dont' have that yet so I can't comment on how it stresses the gpu relative to crysis demo. I do know that crysis stresses it a LOT more than 3dmark11, however, and it's not even close.
Click to expand...

What crysis "demo" are you talking about? The original?

Battlefield 3 locks my GPU usage at 99% and can take over 1.5GB of Vram. Crysis 2 with DX11 patch and high resolution textures on ultra can take 1.8GB of vram and load 99% on the GPU.
 
Yes, the original crysis (not warhead) demo. I've downloaded/deleted it a few times over the years but never actual bought the game. Going through 4 loops of crysis demo for some reason stresses a card more than 99% gpu usage on anything else that I've ever seen other than furmark/kombustor/occt.
 
NO,

Your running stability test with 100 percent load. When you play games, your CPU load assuming a quad core ,, is about 60 to 70 percent BF3 or Crysis 2. It will not crash.

If one hour to 2 hours of gaming doesnt crash it, thats the best test,, ...
 
tweakboy said:
NO,

Your running stability test with 100 percent load. When you play games, your CPU load assuming a quad core ,, is about 60 to 70 percent BF3 or Crysis 2. It will not crash.

If one hour to 2 hours of gaming doesnt crash it, thats the best test,, ...
Click to expand...
there was basically NO cpu usage in the OCCT gpu test. again I ran just the gpu test with error detection. plenty of games will use 99% of the gpu but its no where near as demanding when OCCT is using 99%.
 
bryanW1995 said:
Yes, the original crysis (not warhead) demo. I've downloaded/deleted it a few times over the years but never actual bought the game. Going through 4 loops of crysis demo for some reason stresses a card more than 99% gpu usage on anything else that I've ever seen other than furmark/kombustor/occt.
Click to expand...

I have Crysis and I will say Crysis 2 will stress the card more maxed out. Crysis gets 40fps for me now, Crysis 2 is about 28fps lol.
 
I think you're fine. My old gtx465 would get about 5-7 errors in a 10-15 minute OCCT run, but had absolutely no problems in every single game I played with it.
 
What do you do will those cuda processors? If you computer is purely for entertainment then sure do whatever you feel comfortable with. If your computer is used to make real money then stick with stock. That's my philosophy.
 
Very timely thread. I've been OC'ing my year old GTX 460 (got the bug recently) and have been doing baby-step adjustments to it and error-testing it with OCCT. Someone in Overclock.net recommended the approach. I'd get one error and thinking I needed to be error-free I'd bump up the voltage one notch. Right now I'm at 875/1750/2000 with 1.062 - which I think is pretty reasonable (with zero errors).

I don't think there's any chance of me "killing" it. I don't plan on going much higher (if at all) and I just installed an AC Accelero Twin Turbo II on it. After 15 minutes of OCCT with the fan on auto it was just 64C. Plus Afterburner is limited in voltage adjustment.
 
Last edited:
My experience has been weird with OCCT and Kombustor. I Oced my 6950 to 883/1375 on 1100 mv and it was stable in OCCT and Kombustor for more than an hour. However, when I started playing Witcher 2 and Crysis 2, I would come back to the desktop and the "AMD Display driver stopped working and has recovered" message would pop up. Then I upped my volts to 1125 and it was stable in both games.

with OCCT and Kombustor, I suspect Nvidia and AMD detect that you are running artificial stability tests and the card sips less power and runs slower during these sessions.
 
I don't think Nvidia drivers detect OCCT and throttles. It runs hotter than the most demanding game I have with 100% GPU usage. Can't speak for AMD.
 
One error would bother me, but to each his own. I've run the new version of OCCT with 3-way SLI GTX 580s. If my AX1200 didn't have protection built into it, it probably would have melted within an hour or running it. It can get each GPU over 96C with the fans at near 100%. I ran it for half an hour before shutting it down once. Didn't get any errors, but the amount of heat generated was massive. I was half expecting the whole system to glow orange after a while.
 
1 error does not mean its stable. Gpus need a host of stress tests (different games) because a gpu is very complex. A shader intensive game in dx9 vs a memory intensive game in dx11 will cause your gpu to potentially be stable in 1 and not stable in another.

Just because you use the most demanding bench out there (metro 2033, furmark) and pass does not mean you are stable. And crysis would hardly be demanding for 2011. Maybe 2 years ago.

I play lots of games. I've been stable in furmark, metro, crysis, 3dmark, dozens of other games, but you will find that 1 odd game that you hardly ever play gives you artifacts until you downclock a little. That's how it goes with gpus. There is not 1 benchmark that can stress all the components of a gpu until they find a way to run dx9, dx10, dx10.1, dx11 pathways all at once, which will never happen.
 
does 1 error in OCCT really mean much?
Click to expand...
If you want real stability you should overclock to where there are 0 errors and then scale back the OC a bit since on occasion a program will stress your hardware in ways the burn-in test didn't.

Applying that to an undervolt like you are doing, scale back the undervolt a little. Since you get 0 errors at 1.000 and the default is 1.013... well honestly in such a situation I wouldn't undervolt at all.
 
Leyawiin said:
Very timely thread. I've been OC'ing my year old GTX 460 (got the bug recently) and have been doing baby-step adjustments to it and error-testing it with OCCT. Someone in Overclock.net recommended the approach. I'd get one error and thinking I needed to be error-free I'd bump up the voltage one notch. Right now I'm at 875/1750/2000 with 1.062 - which I think is pretty reasonable (with zero errors).

I don't think there's any chance of me "killing" it. I don't plan on going much higher (if at all) and I just installed an AC Accelero Twin Turbo II on it. After 15 minutes of OCCT with the fan on auto it was just 64C. Plus Afterburner is limited in voltage adjustment.
Click to expand...

Your "thinking" is wrong. Both AMD and Nvidia throttle all of their cards in furmark. OCCT and Kombustor do the same thing to the cards that furmark does, and it is entirely possible to fry a card using those programs according to both camps. Intel/AMD don't throttle their cpus in linx/ibt/etc etc or even tell you not to use them, because they use programs like that themselves in testing. GPU's are NOT cpu's, so don't come back here complaining/whining when you kill your gtx 460.

tigersty1e said:
1 error does not mean its stable. Gpus need a host of stress tests (different games) because a gpu is very complex. A shader intensive game in dx9 vs a memory intensive game in dx11 will cause your gpu to potentially be stable in 1 and not stable in another.

Just because you use the most demanding bench out there (metro 2033, furmark) and pass does not mean you are stable. And crysis would hardly be demanding for 2011. Maybe 2 years ago.

I play lots of games. I've been stable in furmark, metro, crysis, 3dmark, dozens of other games, but you will find that 1 odd game that you hardly ever play gives you artifacts until you downclock a little. That's how it goes with gpus. There is not 1 benchmark that can stress all the components of a gpu until they find a way to run dx9, dx10, dx10.1, dx11 pathways all at once, which will never happen.
Click to expand...

Other than the furmark comments, I agree with this. Multiple games that stress different parts of the card are the best way to verify that it's working properly.
 
Last edited:
bryanW1995 said:
Your "thinking" is wrong. Both AMD and Nvidia throttle all of their cards in furmark. OCCT and Kombustor do the same thing to the cards that furmark does, and it is entirely possible to fry a card using those programs according to both camps. Intel/AMD don't throttle their cpus in linx/ibt/etc etc or even tell you not to use them, because they use programs like that themselves in testing. GPU's are NOT cpu's, so don't come back here complaining/whining when you kill your gtx 460.
Click to expand...

1. I'm very cautious and don't take big, stupid leaps in Mhz or voltage, plus I only run OCCT for ten minutes at a time
2. It ain't going to overheat with that third party HSF on it.
3. If it does get fried its a year old relatively cheap card (so I'm not going to whine and bitch to anyone)
 
You must log in or register to reply here.

TRENDING THREADS

Back
Top