EVGA GTX 970 SC Overclock help

Nikko G

Honorable
Aug 25, 2015
136
2
10,685
So currently my GTX 970 (EVGA SC Edition) has about 100 Mhz on the core and 200 Mhz on the memory added, this is with power limit to only 100. Though when I try to raise the core clock to something like 110, power limit to 110 and even 25+ Mv on the voltage it is still unstable. Currently my boost is suppose to be 1417 on the core but it ends up being 1467 and considering its only a 10+ boost to the already 100+ on the core shouldn't a 110 power limit and +25 Mv be enough to keep that stable? My temps are fine hovering between 65-70C.
 
Solution
Ok, like I said, why dont you try running the memory clock at stock?

If the core clock cant go higher than +100, then it might need even more voltage (even though it shouldnt).

The thing is, that not all cards overclock the same, even if you have the exact same card as someone else, their gpu might be better for overclocking than yours. These are standard variations from one gpu to another, not all of them will have the same tolerances.

A quick google search reveals that this card can reach about +150 MHZ base clock, according to techpowerup's sample.
Other users reported +120 MHZ without overvoltage, +104, etc.
Most of them seem to be able to be successfully stable with +200 on memory, just like you.

My understanding is...
Which program are you using to OC? I assume EVGA Precision, right?

Also, how are you testing stability and what exactly happens when the card is not stable, as you said?

Anyway, the 200MHZ memory OC could be your problem. Sure, your core temps are fine, but what about memory?
Why dont you leave the memory stock for now, and focus on the Core?
See if it makes a difference.
 


Yes I am currently using EVGA Precision X, does the program matter? Also I am using Unigine Valley and 3dmark Firestrike. When I put the Core clock a little bit higher than 100 the programs crash after a certain amount of time. Tried playing a game like Witcher 3 and that was playable for about 10 minutes then crashed as well. No artifacts that I can spot at all in any of the benchmarks or game.
 
Ok, like I said, why dont you try running the memory clock at stock?

If the core clock cant go higher than +100, then it might need even more voltage (even though it shouldnt).

The thing is, that not all cards overclock the same, even if you have the exact same card as someone else, their gpu might be better for overclocking than yours. These are standard variations from one gpu to another, not all of them will have the same tolerances.

A quick google search reveals that this card can reach about +150 MHZ base clock, according to techpowerup's sample.
Other users reported +120 MHZ without overvoltage, +104, etc.
Most of them seem to be able to be successfully stable with +200 on memory, just like you.

My understanding is that its probably your sample that cant overclock very well, try increasing the voltage a bit more, but dont overdo it.
 
Solution


Yep I do understand that not all chips are the same when it comes to overclocking, but what I find weird is that it runs at 100 on the core perfectly fine without changing the power limit and such, but as soon as another 10 is added it just crashes, no matter the power limit. I guess I'll try keeping memory low and try increments of 10 on the core and see how that goes.
 


Yes, your goal now should be to push the core even more (since thats what you want), so ignore the memory for now. Leave it at stock. You can always overclock it afterwards, when you've found a good setting for the core clock.

I am honestly out of ideas though. I dont really have a good answer for you I'm afraid.
 


Yep thanks for the help anyways, btw whats a safe voltage for this GPU?
 

1250mv should be pretty damn safe. Not sure how high you can go though, I havent pushed mine at all. I like being close to stock voltage.
 
For your card:
Max out the power target slider. Most SC's are good start with +100 on the core, and +300 on the memory. Adjust in small increments from there.

There's nothing at all in the stock bios that will allow you to harm your card in any way, shape or form. The voltage is locked in the bios to max out at 1.262v. You'll likely see a max of 1.256v. SO AS LONG AS YOUR TEMPS ARE GOOD, move the slider all the way up and run with it.

Also...download GPUz and open the sensors tab. Use either Unigine Heaven or 3D Mark Firestrike to put stress on the GPU and test the overclock. That's not likely to give you "game stable" overclocks, as each game is a little bit different, and each driver version even more different, but they'll give a basic understanding of how to fine tune your clocks for the best performance, so you can do it for the games you want to play.

Now that you have GPUz downloaded, and the sensors tab open, watch the core clock, memory clock, voltage and pay particular attention to the perf cap performance line. The perf cap performance line will tell you exactly what's capping the performance of the GPU, so you can work around it. VREL, isn't anything to worry about. Really common. VOP, means you need more voltage to go any further. And PWR, means Power. You're at the cards maximum power target, and clocking any higher won't do you any good, because the card will just throttle itself to get itself back under the power limits.

There are ways around the power limits, but.....that's a whole different animal. Tackle this one first.

Oh...and IF you're running SLI. Open two instances of GPUz. One for each GPU. That will be important to dialing in your overclock to work well on both cards. The selector at the bottom of GPUz will allow you to select which card each window will monitor.

EDIT: I always advise for people to try to keep their temps lower than 75c. OBVIOUSLY, the lower the temps, the better. Maxwell LOVES to run cool, so the cooler it is, the more stable it'll be and the higher it'll boost.