Nvidia GTX 1080 8GB overclock

liberty610

Distinguished
Oct 31, 2012
466
5
18,815
Hey guys,

I'm attempting my first overclock, and I decided to try it with my MSI GTX 1080 8 GB card.

I some what understand what overclocking is. I have been watching several tutorials on YouTube on how to use MSI afterburner. Several of them said that I could boost the core voltage all away up to a hundred with this card while trying too find the sweet spot for overclocking. But I get confused on where I should dial it back to, and I have heard that running it at 100% core voltage is not good and I noticed that my temperatures went up to almost 89 Celsius when I did. I don't necessarily need to overclock the card because it runs pretty solid out-of-the-box, but I'm an enthusiast when it comes to the stuff and I want to learn how to do it.

So am I currently doing this right while avoiding the core voltage? I have the core voltage set to zero, power limit is 121 with temp limit linked/locked to that with a high of 92c.

I have a custom fan curve, and my temps are staying at a solid 73c to 76c.

I am currently running Valley Benchmark to test my over clock. I started with the core clock boost first. And I was able to get it to +150 without any issues, anything above that caused a crash. So far, I am working on the memory clock boost, and for some reason I'm able to get it to +700 without any crashes. Is this normal? I have read that you can get the memory clock a lot higher than the core clock, but +700? Is this normal? Am I doing something wrong?

Any tips or help would be great. I also plan on getting 3D Mark to treat it further. Thanks in advanced.
 

Eximo

Titan
Ambassador
Sounds like you did pretty well, depending on what the original boost clock was.

Most Pascal cards max out around 2100Mhz on the core. The memory is more flexible, but I've not seen most people exceed 10,500 or so. So if you really managed a +700 that is pretty good.

Those temperatures are quite good. If you are satisfied with the stability and overclock then there is no need to add excess voltage for no reason.
 

liberty610

Distinguished
Oct 31, 2012
466
5
18,815
Yea, I have left the core voltage alone at 0. In Valley, my base core clock was 1987 Mhz and the memory clock was 5005Mhz. The core seemed to cap out at +150 on afterburner which is 2126 Mhz now on the core. I stopped pushing the memory clock, because +700 seemed like a lot. I don't seem to be having any issues in Valley yet, and I ran a bench test in Grand theft Auto 5 with it. I couldn't really tell any issues. There MAY have been a little lag on one of the loading scenes, but I am not sure. I am going to try 3D Mark and see how it acts.

I am getting a solid 55 to 60fps on GTA 5 at the resolution of 3440 x 1440 (I have an ultra wide monitor). Should I keep boosting the memory clock? And how long should I run Valley in order to test the stability of the OC? I was letting it run a good 20 mins or so at 700+ and didn't seems to have any issues.
 

liberty610

Distinguished
Oct 31, 2012
466
5
18,815
Oh, the memory in Valley is now at 5705 Mhz with the +700 on MSI Afterburner, and Valley is showing it as 5705 Mhz while running. And you are suppose to times that number by 2, correct? If so, I am getting 11,410 on the memory. The base without an OC was 5005 Mhz,
 

liberty610

Distinguished
Oct 31, 2012
466
5
18,815
Okay then, sounds good! I seem to be stable with it. I played GTA 5 for a good 45 minutes without any issues.

So here's my next question... How much of a boost should I be getting performance wise? And what settings should I have Valley benchmark set at to test it? The main reason I ask is, I have an ultra wide monitor that runs 3440 x 1440 resolution. This resolution setting isn't available in Valley. I have to tell Valley to 'match system' in order to run at that resolution, because the max resolution setting Valley offers is lower.

So, when doing the base clock test compared to the overclock test with the plus 150 on the core and the plus 700 on the memory and with Valley settings set to ultra and the resolution set too 'match system' so it runs at 3440 x 1440, my benchmarking score wasn't that much different at all. Maybe 1fps faster for the average, and my max fps was like 3 higher.

I'm assuming that since I'll be gaming in the 3440 x 1440 resolution, that's the resolution I should be running the benchtest, right? It just seems like I should be getting a slightly higher fos boost them that, no?
 

Vellinious

Honorable
Dec 3, 2013
984
2
11,360


Use Superposition or Firestrike. Valley is a horrible GPU test as it's CPU limited. It's more a CPU benchmark now.
 

liberty610

Distinguished
Oct 31, 2012
466
5
18,815


This is the first and only time I have heard of this. I'll def. look into it.
 

Vellinious

Honorable
Dec 3, 2013
984
2
11,360


It used to be decent...but with today's high end hardware, it's very CPU limited. Heaven is too, but not to the extent that Valley is. Superposition or the 3D Mark benchmarks, like Firestrike and Timespy are much better for GPU only benchmarks.
 

adamscurr

Distinguished
Apr 6, 2013
232
0
18,760
I would have to say that timespy and firestrike are more overall system benchmarks since it tests a lot of parts of your system. It does give you separate scores for the cpu and gpu, but you get a combined score in the end... I like it... If you really want to test your system, run the 4k test... :)

Adam
 

liberty610

Distinguished
Oct 31, 2012
466
5
18,815
I downloaded 3d mark the other day to try the free portion of it, but the other tests are locked I guess? And I am currently getting this error when I try to run Timespy:

DXGI call IDXGIFactory2::CreateSwapChainForHwnd failed [142213121]
 

Vellinious

Honorable
Dec 3, 2013
984
2
11,360


The graphics score is the only thing that matters.