How to overclock a GPU?

[tuffluck]

i have a sapphire, vapor-x radeon 5770 and am wondering the best way to OC it is? there are no good OC guides for it! is it safe to OC the GPU just like it can be safe to OC the CPU? how much will it help me?

 

[RazberyBandit]

Start by reading here:
http://www.tomshardware.com/reviews/overclock-graphics-card,1916.html
It's old, but there's a wealth of information there that should get you started.

 

[flyinfinni]

The easiest way to OC is to use CCC overdrive utility. Run a program like GPU-Z to watch your temps and loads etc. All you need to do is just increment your clocks a little at a time and test between each increase. Do just one clock first (either core or memory) and step it up 5-10 MHz at a time and run a test (Furmark is a good one, or just play some games or run 3dMark Vantage). keep increasing and testing until you start to see artifacting or a crash, or your temps get too high, and then back down 10-20 MHz and theres your OC. Repeat this for both Core clock and Memory clock. Should get you a nice boost.

 

[tuffluck]

The easiest way to OC is to use CCC overdrive utility. Run a program like GPU-Z to watch your temps and loads etc. All you need to do is just increment your clocks a little at a time and test between each increase. Do just one clock first (either core or memory) and step it up 5-10 MHz at a time and run a test (Furmark is a good one, or just play some games or run 3dMark Vantage). keep increasing and testing until you start to see artifacting or a crash, or your temps get too high, and then back down 10-20 MHz and theres your OC. Repeat this for both Core clock and Memory clock. Should get you a nice boost.

thanks! very helpful. what temps should i try to stay within? i think the vapor-x does 40c idle and 60c max, but that is with the fan at 40%.

 

[flyinfinni]

You'll want to keep your temps below 100C for sure, and I personally try to stay under 85C for normal use. You probably won't have a problem with temps on a 5770 though- they're not super hot to begin with.

 

[tuffluck]

You'll want to keep your temps below 100C for sure, and I personally try to stay under 85C for normal use. You probably won't have a problem with temps on a 5770 though- they're not super hot to begin with.

so you don't have to stress them for 10 hours or anything like with the cpu to make sure they are stable? just do futuremark or 3dmark and if it doesn't artifact it is okay?

can you update drivers and the OC still be okay? do i need to up the GPU core voltage? how much mhz do you think i can safely add on to the GPU and memory? thanks again!

 

[flyinfinni]

Generally if you can get through 3dMark without having any problems, or run 10 min of Furmark, you should be alright. What will really tell you though is if when you are gaming for extended periods of time and start having problems, you might need to back it off a little, but torture testing for many many hours like with a CPU really isn't necessary in my opinion- you'll know when you are gaming if its ok or not, and its pretty quick to fix (especially since you aren't generally messing with voltages and stuff)

 

[tuffluck]

i got it from 860/1200 to 905/1245. it ran fine with no artifacting during the test, but the gpu got up to 87c during the test of 10 minutes. is that okay? i ran 3dmark06 and didn't notice a huge difference in the score or fps but during 3dmark06 it only got up to 59c. how hot can i get it in testing and how hot should i aim for it under load? is it weird that didn't make a huge impact on the 3dmark06 score?

 

[flyinfinni]

You should be ok with that. I mean, furmark is an artificially massive load on the GPU, so it'll get hotter in that than it ever will with an actual game, which is why I run 3dMark and play some games as well when I test.

 

[tuffluck]

okay, great. i was doing 68c while playing bf:bc2. i think i will leave the OC where it is, huh? or can/should i keep going? playing bf:bc2 at medium settings on 1680x1050, i get 62 average fps for about the 10 minutes i played. that's great, i was getting 50 before...which seems like way too big of a difference for just a small percentage OC. what do you think?

e6750@3.4/2gb ram otherwise.

 

[tuffluck]

Best answer selected by tuffluck.

 

[tuffluck]

i tried the auto-tune feature and that was garbage seeing as once it spent 30 minutes to set, i ran 3dmark06 and it crashed almost immediately. i put them back on the 905/1245 that i had set them earlier and called it a day.

fyi, when i compare my 3d marks in 3dmark06 to the 3d marks that TH posts in their graphics card area, i get lower scores even with the OC than the other 5770's listed. the 3.0 score is like 7,6xx for 5770's and i am getting like 7,4xx. do you think that's probably typical, where there is a 5% difference between some cards? it's just that there are 3 5770's and they all score around the same, but mine scores lower than all of them, even overclocked.

 

[EXT64]

You have to compare other components as well, like CPU, Ram amount and speed, etc.

 

[tuffluck]

You have to compare other components as well, like CPU, Ram amount and speed, etc.

3dmark06 is meant to do nothing but confuse! i thought they broke out 2.0, 3.0, and cpu score so that they were directly correlated to what they say they are regardless of other components.

also i find it weird that TH doesn't say what processor, RAM, mobo, etc they used to do those benchmarks?

 

[flyinfinni]

The CPU makes a big difference in the 3dMark 06 and 3dMark Vantage scores. Look just at the GPU scores and see how big of a difference you see there as opposed to the total score.

 

[tuffluck]

The CPU makes a big difference in the 3dMark 06 and 3dMark Vantage scores. Look just at the GPU scores and see how big of a difference you see there as opposed to the total score.

i am talking about the gpu scores ONLY. the 7,6xx is for the 3.0 score on the GPU for the 5770s, and my 3.0 score only gets 7,450 overclocked. my total score is 14,471, so i know the CPU plays a role in that. that's what i was asking...if it's weird that my GPU ONLY score is lower on an OC'd 5770 compared to normal clocked 5770s?

 

[flyinfinni]

Well, try backing down your overclock a little. It could be that even if its stable, its not performing properly. If you go too far it can lead to some performance drop. Did you ever bench your card before you OCed it?

 

[tuffluck]

Well, try backing down your overclock a little. It could be that even if its stable, its not performing properly. If you go too far it can lead to some performance drop. Did you ever bench your card before you OCed it?

yeah, the 3.0 score was 7,342 with no OC, and after the OC it was 7,448, but i'm seeing several on TH list that are doing 7,600 supposedly stock.

is it just possible some cards come a little under/over performing from the factory, or do you think there is something wrong with my card?

 

[flyinfinni]

what CPU are you using? Its possible its bottlenecking your GPU a little, but still- seems a little odd.

 

[tuffluck]

what CPU are you using? Its possible its bottlenecking your GPU a little, but still- seems a little odd.

e6750@3.4. have you ever run the 3.0 test and compared it to the TH results to see if it is the same? maybe cpu plays a part...look at this guy. same card, 5770, and hitting drastically lower numbers than mine.

http://www.tomshardware.com/forum/274740-33-5770-higher-3dmark06-score-4890

 

[tuffluck]

i compared my score with others that have a similar config to me on the ORB of futuremark...it looks okay i think. i'm sure a diff processor would change the gpu score a bit...could be a slight bottleneck, but nothing immensely grave to worry about. <3%.

 

[flyinfinni]

Thats true- its very possible that the CPU is a little bit of a bottleneck. One way to check that is to try and OC your CPU a bit and see how much of a difference it makes in your scores.

 

[tuffluck]

stock @ 2.67 it was an SM 3.0 score of 6,619, so yeah with the CPU @ 3.4 and getting 7,342 i would say the CPU definitely plays a part in it.

i am looking at the i7's paired with the 5770 and the 3.0 score is capping at around 8,000, so my bottleneck doesn't seem to be too huge (7%).

 

[flyinfinni]

Yes- thats pretty good indication of a CPU bottleneck. Unless you can OC your CPU a little more, then you're gonna be stuck there, though bumping the resolution up will shift that bottleneck more over to the GPU from the CPU, so if you are gaming at a higher resolution than you are benching with 3dmark06 (generally you will be) it shouldn't be too much of an issue.

 

[RazberyBandit]

That's the biggest problem with the freebie versions of 3D Mark - you can't adjust the resolution to get a true representation of what factors are making what difference at the actual resolution you use when gaming. A CPU bottleneck at 1280x1024 does not indicate there will be any such thing happening at higher resolutions, like 1920x1080 for example. Will performance still benefit from an overclocked CPU at higher resolutions? Usually, yes, but the difference in performance is not nearly as much as it is at lower resolutions.