780m won't hold OC.

Hey guys,

Haven't posted in a while but I am having problems again with my 780m in SLI. I have had this exact same problem in the past. Reinstalling drivers solved it before but it doesn't work now.

Basically what happens is that my core clock of my second 780m won't overclock with an unlocked vBIOS. It simply stays at stock clock of 850. It won't budge. At times when gaming it will spike up to that OC value (typically around 950-1000) but it simply won't hold. Memory overclocks just like normal, it is just the core clock that is uncooperative. It's just steady at 850 with temps usually 5-10 degrees less than my other GPU which also consumes significantly more power (when playing BF4). Could this be a symptom of a bad card? It just seems so odd that it refuses to cooperate now but in the past the problem somehow solved itself.

Any ideas as to what could be causing this?

Thanks guys,

T

 

I know this might seem like a silly question but did you flash both cards?

 

That's a really important question. You won't believe how many times I've seen that haha.
We don't recommend overclocking the GPU manually but we can advise you to constantly check on the temperatures if you are doing it. Here is how!

 

Try forcing the max P-state of the 2nd card with Nvidia Inspector and see if that helps, syntax like this (assuming card ID 1 is the one you're interested in):
nvidiaInspector.exe -forcepstate:1,0

 

Yup. Both have been flashed since August. It's really a weird problem. It's almost as if k-boost doesn't work on the second card. I'm gonna do some 3DMark tests this afternoon to see if my old scores are noticeably better than they 're now with my temperamental card.

I'll give that p-state a try in Inspector.

 

Have you tried other games?

 

I've tried other games like CounterStrike and Crysis. It does the same thing. I did a test on 3DMark11 and my scores haven't changed to when I first got it. My graphics score has actually improved a little. Before I couldn't break 16k. I was expecting a performance decrease. This all seems so strange to me. If you look at HWINFO on the screenshot it shows you how my different GPU's behave. I guess its not so much a problem as it is an annoyance to my obsessive compulsive tendencies lol. My physics score is embarrassing.

 

I think it's a mixture between the game and driver.

 

From the screenshot, it seems that your secondary GPU reached 88 degrees max. temperature. I don't think you can reach that kind of temperature at stock clocks, even with 1.075V. Can you run some on-screen monitoring during the benchmark to look at the real-time GPU clocks? Rivatuner statistics server 6.0.0 should do nicely for that purpose.

 

During the tests, the overlay shows that the core clock of GPU1 is jumping all over the place in benchmark 1,2 and 4. It jumps between 849 and 1060 (my OC), although mostly between 900-950. GPU2 sits at 1060 through each bench. In the Physics test GPU1 held at 1060 through the bench while GPU2 sat at 849 (not sure if this means anything). In the combined test both clocks were steady at full OC. Temps did not exceed 86C through the entire bench. Memory is steady throughout all of the bench as well.

Any ideas?

 

If you did not extend the power limit ("power target" in Precision), that same behavior will occur. I believe it is safe to set the power all the way to the maximum of 146%.

 

Yes, I have my power target to maxed all the time, even running stock clock speeds. I haven't seen any adverse effect with temps or anything from that.

Physics test is CPU and doesn't use SLI/CF or even a single GPU, so what you observed during the Physics test is expected without any GPU load.

 

I've always left that maxed as well. I was doing a lot of testing yesterday and noticed that sometimes my cards hold the clock and sometimes they don't. Doesn't seem to matter what the clock is, its kind of odd. Despite that I achieved my highest graphics score yet at 17.2k on 3DMark11 last night. I guess it's not so much a problem as I thought it was. Basically the clocks do whatever they want at idle but usually sort themselves out under load.

Thanks for the info guys.

 

triple check the contact your heatspreader on the gpu is making. I have a suspicion maybe your not making full 100% contact.

 

If it's boost causing issues you could disable it and edit the main clocks.