How do YOU test/find your stable 680 OC?

[Whach]

Hi all.

As the title says, what's your procedure/philosophy and what do you consider a stable 680OC? Using which programs, i.e. dedicated stress benches vs actual game play? (or both)

I try to find the best GPU OC first, then the memory. I don't fiddle around too much with the Power Target settings too much, and don't change the voltages (just my preference).

I have been using Heaven on the highest settings (tessellation extreme, 8xAA etc) and have found that some OC settings don't loop for long. In my case 2-4 loops of consecutive benchmarks at a certain OC.

However, I have also found that those exact same settings are perfectly playable for hours on end in demanding games like BF3, Crysis, Crysis 2, Metro etc.

I have used the OCScannerX utility without any hitches or reported errors, but find Heaven to be more "sensitive"?

Based on this, I find myself asking whether an OC is considered to be more stable/important/relevant if a heavy benchmark can loop continuously AND play games, or if just games by themselves?

I would assume the prior, but I also would consider the latter too as playing benches is not a primary use scenario. IDK.

What do you guys/gals think? Look forward to you opinions and thoughts =)

 

[Homer]

1. I'd recommend raising your power target to max. That will help eliminate any possible instabilities overclocking (at least what you can get out of 1.175v).
2. Heaven benchmarking is a pretty good start. Use max settings at the highest possible resolution (if you can do 5760x1080 or something larger, do it)
3. Watch temps while stress testing. If you get too close to 70*C at a desirable fan level, you might as well downclock a bit otherwise it will throttle for you.
4. Heavy games like BF3 are good stability tests as well (usually if you can run for a few hours your OC is stable)


One question I have about benches though... even if I hit some clocks fully stable stress testing, I will sometimes actually lose score in Heaven and 3dmark at a certain OC. Anyone know why this would be?

 

[psyside]

1. I'd recommend raising your power target to max. That will help eliminate any possible instabilities overclocking (at least what you can get out of 1.175v).

What is the advantage of raising the TDP if your card doesn't even hit 90% power usage (like mine 680DCII@1280 core)

Is there something i don't know about it?

And about the question, its unstable memory overclock, if you got any...if not its just slightly unstable overclock overall.

 

[psyside]

Delete.

 

[Whach]

1. I'd recommend raising your power target to max. That will help eliminate any possible instabilities overclocking (at least what you can get out of 1.175v).

Wouldn't it be better to find the maximum OC at a default power target to find a minimum limit, and then increase the power target after to break the default TDP & to find the upper limit? or the other way round?

 

[psyside]

Wouldn't it be better to find the maximum OC at a default power target to find a minimum limit, and then increase the power target after to break the default TDP & to find the upper limit? or the other way round?

IMHO raising TDP if your not hitting max, it will cause more instability then without it.

Maybe a slight bump 5-10%, is the best metod?

BTW its so strange to get Heaven crash in 20 seconds, and being able to play BF3 40 mins and then crash, imho Heaven is not so good benchmark like many thinks, its maybe bad coding more then finding instability? really strange, because even Crysis need alot more time to crash, and its alot more demanding.

 

[Whach]

Just out of curiosity, how did you find your optimal memory OC. looking for artifacts is PITA = /

 

[psyside]

I don't, i got my card 2 days ago, still havent played alot with it regarding overclocking

As far as i know, you should use the exact same settings/benchmark and chek when the fps start to drop, or you can play some demanding singleplayer game and chek out when fps dips start to appear (memory error checking)

Best way is to use benchmark, for most reliable test.....but if your bored with it just play games and chek out fps stability.

 

[sk3tch]

And about the question, its unstable memory overclock, if you got any...if not its just slightly unstable overclock overall.

Thanks for this advice. In my haste to get the fastest and most memory bandwidth with my 4-way setup I had my memory OC too high. Even though it was "stable" - it was exhibiting strange performance as Homer outlined (decreased bench scores and decreased FPS in games). I reset it back to stock and it is actually better now. Need to go back and find the sweet spot between +0 and +400 mem clock offset.

I'm dumping these FTW+ 4GB cards, anyway. I'll try with my Classys.

 

[TroyX]

i find memory oc to be pointless.

 

[wadec22]

i find memory oc to be pointless.

depends on your setup. since i prefer a single gpu setup and drive three monitors, memory ocing actually helps a lot.

as far as stress testing my oc's....
1st I loop heaven a couple times
2nd I play BF3 for awhile
3rd I play about 15min of Crysis 2. Crysis 2 seems to be the ultimate stress test from what i've seen.

 

[fomoz]

One question I have about benches though... even if I hit some clocks fully stable stress testing, I will sometimes actually lose score in Heaven and 3dmark at a certain OC. Anyone know why this would be?

From personal experience this can be two things. Either your average Kepler boost is lower because your cards are throttling at some points of the bench (70C and 80C) or you OC'd your RAM so much that you're getting micro stuttering, but without visible artifacts.

i find memory oc to be pointless.

I get a higher FPS boost from memory OC than from core OC.

 

[psyside]

Thanks for this advice. In my haste to get the fastest and most memory bandwidth with my 4-way setup I had my memory OC too high. Even though it was "stable" - it was exhibiting strange performance as Homer outlined (decreased bench scores and decreased FPS in games). I reset it back to stock and it is actually better now. Need to go back and find the sweet spot between +0 and +400 mem clock offset.

I'm dumping these FTW+ 4GB cards, anyway. I'll try with my Classys.

No problems.

 

[Blkout]

i find memory oc to be pointless.

Its not pointless at all on the GTX 6xx series. They benefit greatly from memory overclocking compared to the current gen AMD cards.

 

[Whach]

Does anyone know hot to reset the default fan curve in precision X? I was messing around with it, but neglected to take note of how the original one looked like =/

 

[wadec22]

Does anyone know hot to reset the default fan curve in precision X? I was messing around with it, but neglected to take note of how the original one looked like =/

if i remember correctly its really basic, 30C is 40% fan speed, 40C is 50%, 50C is 60%, etc...

 

[Whach]

if i remember correctly its really basic, 30C is 40% fan speed, 40C is 50%, 50C is 60%, etc...

I don't remember it being linear like that when I first saw it. I remember it to be a curved profile. Well, at least I think it was. Oh well, maybe ill just re-install it to have a look =/

 

[wadec22]

I don't remember it being linear like that when I first saw it. I remember it to be a curved profile. Well, at least I think it was. Oh well, maybe ill just re-install it to have a look =/

just double checked, it is that linear. slightly more aggressive than msi afterburner, probably to help with higher default clocks.

 

[Whach]

ahhh. thanks for checking = )