Help overclocking my 6970

smarteyeball said:

1.4 -1.5v is a bit more than a 'tiny' boost

1.425-1.45v is a bit more realistic. Just make sure you keep an eye on the temps.

ok, thank you

smarteyeball said:

1.4 -1.5v is a bit more than a 'tiny' boost

1.425-1.45v is a bit more realistic. Just make sure you keep an eye on the temps.

ok, so i bumped it up to 1.425v and ran a stability test, the heat peaked around 55.6 degrees. it seems as though im getting less taring and stuttering

smarteyeball said:

What stability test?

Is it the GPU that's tearing and stuttering, if so the GPU OC is not stable. CPU overclocks don't cause screen tearing or stuttering.

ok, well all i know is the computer will lock up for a milisecond and all of a suddent everything is trying to catch up to iether the gpu or cpu. i think it was cpu as i dont see the problem anymore, i could see it being the gpu possibly, but it is weird, if i get a gpu crash, i get vertical lines of grey and white. if i have a cpu crash i get horizontal lines of rainbow colors then a BSOD. as for the stability test, amd overdrive has a stability check for the cpu, it runs a "cpu, fpu, and calculation" test which will max the heat output and is kindof like a bench mark but is more to say if it is running O.K. i wouldnt know how to test the stability of my GPU when overclocked unfortunately, if you know of a way i could do this if you let me know id appreciate it

smarteyeball said:

What stability test?

Is it the GPU that's tearing and stuttering, if so the GPU OC is not stable. CPU overclocks don't cause screen tearing or stuttering.

i actually had another question, im running my graphics clocks at cor: 930, mem: 1425, and gpu: 1175. as youve pointed out before, somthign in my graphics is unstable and i agree, the question is, is it more likely to be my ram or my core? or would i really have to run a stabillity test to know for sure (well duh, but you know what i mean)? i want games to run smoother, and this plus+ cpu OC has helped tremendously, but i can still go smoother, what part is more important to the frame rate? i would imagine it would be the memory clock, because that is the ammount of frames that can be cached, but also because the game is FPS, its kinda spntanious which makes me believe more gpu may be in order, anyways, which can i lower, and which can i increase, or rather, what could i do that would increase the over all frame rate?

smarteyeball said:

Sorry, just noticed this.

It's hard to pinpoint the exact culprit. Probably the core speed. The best way to work out it if it's core or mem is to dial one back a step or two and re-test until it's replicable.

ie:

900 / 1425

If that eliminates tearing, then you simply keep raising the core (or mem, depending on which value you're testing) until you can reproduce the tearing/choppines.

Overall, the core clock has the biggest impact on framerate. But both have an impact on each other.

Since the core is doing the lions share of the processing, the the faster that is, the more that can be processed. However the memory is almost as important since it's 'feeding' the core. If too slow, it can't feed the data to the core quick which leads to diminishing returns. (frame dips, choppy)

The diminishing returns also apply to the 'voltage wall'. If either, or both, the core and memory aren't receiving sufficient voltage then performance will suffer as it will be unsustainable which results in the choppy/frame dipping. It's like an engine running on lower octane fuel - the engine is capable of X amount of performance, but with a lower octane it simply can't deliver the full performance capability.


For testing I personally use a combo of apps and games.

I start with Kombustor to check the heat and artifacts (flickering dots/long black shadows etc)

Then I use some benchmark apps like 3D Mark Vantage, 3D Mark 11, Heaven Benchmarks to check for artifacts. (Vantage and 3D Mark have 'free' and 'full' versions. The full versions can be free if you know where to look)

Lastly I test 'real world' performance by playing games.

Of equal importance to checking for artifacts I also pay strong attention to the framerate counters in these apps. I take note of how much they fluctuate at certain spots. With the exception of Kombustor, framerates fluctuate a lot in these apps. That's why I choose a particular spot and compare each run.

If there are no artifacts, yet the framerate goes up and down and is not constant, then I know I'm in diminishing return territory. From there I either lower the clock speed or up the voltage until the framerate is stable and constant. Otherwise performance in games won't be smooth.

thank you very much. although, i will not be able to use 3d mark, it does not support 5760x1080, last time i ran it it screwed up my resolution horibly, had to restart without even seeing my score and then reset my drivers (not fun). so if i understand you correctly, just start slowly lowering the core and increasing the ram. id like to play with the voltage a little also, im assuming the same rules apply as to CPU, only go up in .12-.25 increments. but i may need help finding out why TriXX wont let me edit my voltage, understood, it is in beta and it has not gained a reputation, i would like to point out, afterburner didnt work at all. so i need somthing to allow me to edit voltage on the fly easily, ore help finding out why sapphire trixx wont let me so i possibly could

smarteyeball said:

Sorry, just noticed this.

It's hard to pinpoint the exact culprit. Probably the core speed. The best way to work out it if it's core or mem is to dial one back a step or two and re-test until it's replicable.

ie:

900 / 1425

If that eliminates tearing, then you simply keep raising the core (or mem, depending on which value you're testing) until you can reproduce the tearing/choppines.

Overall, the core clock has the biggest impact on framerate. But both have an impact on each other.

Since the core is doing the lions share of the processing, the the faster that is, the more that can be processed. However the memory is almost as important since it's 'feeding' the core. If too slow, it can't feed the data to the core quick which leads to diminishing returns. (frame dips, choppy)

The diminishing returns also apply to the 'voltage wall'. If either, or both, the core and memory aren't receiving sufficient voltage then performance will suffer as it will be unsustainable which results in the choppy/frame dipping. It's like an engine running on lower octane fuel - the engine is capable of X amount of performance, but with a lower octane it simply can't deliver the full performance capability.


For testing I personally use a combo of apps and games.

I start with Kombustor to check the heat and artifacts (flickering dots/long black shadows etc)

Then I use some benchmark apps like 3D Mark Vantage, 3D Mark 11, Heaven Benchmarks to check for artifacts. (Vantage and 3D Mark have 'free' and 'full' versions. The full versions can be free if you know where to look)

Lastly I test 'real world' performance by playing games.

Of equal importance to checking for artifacts I also pay strong attention to the framerate counters in these apps. I take note of how much they fluctuate at certain spots. With the exception of Kombustor, framerates fluctuate a lot in these apps. That's why I choose a particular spot and compare each run.

If there are no artifacts, yet the framerate goes up and down and is not constant, then I know I'm in diminishing return territory. From there I either lower the clock speed or up the voltage until the framerate is stable and constant. Otherwise performance in games won't be smooth.

gpu voltage problem fixed, just had to update trixx

smarteyeball said:

Sorry, just noticed this.

It's hard to pinpoint the exact culprit. Probably the core speed. The best way to work out it if it's core or mem is to dial one back a step or two and re-test until it's replicable.

ie:

900 / 1425

If that eliminates tearing, then you simply keep raising the core (or mem, depending on which value you're testing) until you can reproduce the tearing/choppines.

Overall, the core clock has the biggest impact on framerate. But both have an impact on each other.

Since the core is doing the lions share of the processing, the the faster that is, the more that can be processed. However the memory is almost as important since it's 'feeding' the core. If too slow, it can't feed the data to the core quick which leads to diminishing returns. (frame dips, choppy)

The diminishing returns also apply to the 'voltage wall'. If either, or both, the core and memory aren't receiving sufficient voltage then performance will suffer as it will be unsustainable which results in the choppy/frame dipping. It's like an engine running on lower octane fuel - the engine is capable of X amount of performance, but with a lower octane it simply can't deliver the full performance capability.


For testing I personally use a combo of apps and games.

I start with Kombustor to check the heat and artifacts (flickering dots/long black shadows etc)

Then I use some benchmark apps like 3D Mark Vantage, 3D Mark 11, Heaven Benchmarks to check for artifacts. (Vantage and 3D Mark have 'free' and 'full' versions. The full versions can be free if you know where to look)

Lastly I test 'real world' performance by playing games.

Of equal importance to checking for artifacts I also pay strong attention to the framerate counters in these apps. I take note of how much they fluctuate at certain spots. With the exception of Kombustor, framerates fluctuate a lot in these apps. That's why I choose a particular spot and compare each run.

If there are no artifacts, yet the framerate goes up and down and is not constant, then I know I'm in diminishing return territory. From there I either lower the clock speed or up the voltage until the framerate is stable and constant. Otherwise performance in games won't be smooth.

i tryed going to 1200 from 1175 on the voltage, and i traded 25 with the core and ram, so i whent from 930/1425 to 905/1450. havent gotten to try it out yet, but when i do i hope that works out. i found increasing the voltage on my cpu caused more problems than it solved, it would lock up suddenly periodically and the temp would ramp up suddenly, so i ran it back down to 1.4. anyways now that you know what i did with my voltage, feel free to tell me anything you would do differently, esspecially if you think i ramped my voltage up to high