• 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.

Overclocking gpu and occt GPU error checker

Nerusy

Nerusy

Nerusy

Soldato
Joined
12 Jul 2010
Posts
2,893
What is up with that problem, even though the overclock is stable it dishes out thousands upon thousands of errors.

So, what is your overclock and how many thousand errors you get, does anybody have a GPU overclock that has 0 errors, is it even possible?
 
What card and what overclocks with what drivers? It varies a lot with everyone depending on how good the chip is, the cooling, etc.
 
Zotac amp! edition, max temps during test 62c

on 701 clock it shows 0 errors.
on 741 shows 1000 rising steadily so far, first few minutes, the thing is, crysis is stable so its strange that occt shows errors.
 
Its simple, the likelyhood is that even if they are not visible, the card is NOT 100% stable at the overclocked frequency; it's generating errors against the expected, and pre-calculated data. Similar to how an overclocked CPU might not crash, but it might fail prime/linpack or do wierd things because its not 100% stable and is generating errors at low level.

As you say yourself, at 700Mhz it is stable, at 741 it is producing errors in its output, and is not. Just because you can play Crysis doesnt mean the overclock is 100% stable, I remember lots of cards dying under similar circumstances when Doom3 came out and lots of overclocked cards couldn't handle it and gave up the ghost.

One of the main reasons for testing programmes like OCCT is for exactly the reason you're commenting on, to warn you the card is producing errors before you can see them, or it becomes fatal in the long run for the card.

Upping voltage to the core MIGHT help reduce the level of errors, but thats a whole extra can of worms.
 
Last edited:
And there I thought overclocking gpu's is easier than cpu.

Thanks for the insight, So I take it on stock voltage, max core and sharder overclock with 0 errors would mean that it doesn't degrade the chip or increasing clock and shader and memory on stock voltages still increase the degradation rate?
 
If you are getting errors then your card isn't stable, simple as that.

Add more vGPU until you get 0 errors on a 30min test, but keep an eye on your max load temps.
 
You haven't mentioned what card you have, but I'm guessing it's the gtx 470.

IF so when upping the voltage to obtain stability, you can take it upto 1.087 without any worry of killing the card, but of course as already mentioned, keep the heat as low as you can. 80c and lower preferably under stress.
 
robin66 said:
You haven't mentioned what card you have, but I'm guessing it's the gtx 470.

IF so when upping the voltage to obtain stability, you can take it upto 1.087 without any worry of killing the card, but of course as already mentioned, keep the heat as low as you can. 80c and lower preferably under stress.
Click to expand...

I thought that overvolting will no matter what result in the hardware degrading faster?
 
Alexrose1uk said:
Its simple, the likelyhood is that even if they are not visible, the card is NOT 100% stable at the overclocked frequency
Click to expand...

this is correct. the gpu is generating errors but they are not visible to the eye. expect computational errors if the card is used for doing calculations such as folding etc. i always use occt for gpu overclocking. and i run the artifact tester on my cards for 12 hours to make sure the card is stable.
 
Not fast enough for it to matter at those voltages. Most sites I've been to in the past couple of days have all said that 1.087 on air is fine if you keep the heat under control. Anything over that voltage, and you'll need to run your card on water.

Up your clocks on the standard voltage until you get errors in OCCT, I find that 30 minutes of testing is enough also. Then slowly up the voltage until you get stability and then once stable, just overclock somemore at about 10 mhz increments. keep repeating this until you get to the temps your are happy with, about 80c and lower I go for under stress.

I made this point yesterday about backing off once you find the limit on your card. Myself I tend to back down by anything between 25 to 50 mhz. Some people like to keep it on the edge so to speak. But I'd atleast back down by 10 mhz once you find your safe limit.
 
Last edited:
Phwawww
470s in SLI in my akasa Eclipse 62, with the side fan panel, so 120mm intake fan sat right above both cards.
Bottom card reach's 92 & the top card reach's 107 :D:D
All stock & default fan profile...
 
You must log in or register to reply here.
Back
Top Bottom