[SOLVED] Multiple Pascal Cards Stuck @ Base Clock

[ryoungblood]

SOLVED: Google Chrome, Slack Desktop Client, Skype, Remote Desktop (Windows RDP), when open, throttles a GPU to its base clock. My hypothesis is that Windows has a priority system and will assign GPU use to active applications using DirectX. You can identify offending applications by individually opening and closing them while monitoring GPU power utilization. For my 1080, I see power use drop to a forced 40% or lower when one of these applications are running (active or background). Upon closing the application, GPU power use will increase to its non-limited level and will regain clock set by OC software or other.

Well, it's happening again, and now it's happening frequently on 3/6 of my Pascal cards.

It's an intermittent issue and I can't seem to figure out what is causing it. Right now, across 3 machines (see attached picture), I have 2/6 cards stuck at their base clocks. I have Windows set to maximum performance, and NVIDIA set to prioritize performance as well. The cards are downclocked when running below 65 degrees and drop down to roughly 40-50% power utilization, so it's not a temp issue. GPU-Z on each machine points at Util.

All 3 computers are running Windows 10 Pro 64-bit and running 372.90 drivers (best for F@H). I have the issue on both EVGA XOC and MSI Afterburner. All power levels are set to their maximum for each card for MSI Afterburner. All builds have at least a 750W PSU.

I doubt it's a CPU issue, because it's happening on across 3 different CPUs, including a i7 6700k, i5 2500k, and an AMD FX-8320E.

4 of the cards are EVGA (1x 1080 FTW Hybrid, 1x 1070 SC, 2x 1070 SC Black)
2 of the cards are ASUS 1070 STRIX OC

See picture for reference. View attachment 187872

OLD POST:

I'm running into an issue trying to overclock a 1080 with EVGA Precision 16. I have a 1070 in the same machine that will overclock fine and remains stable at 2025 MHz, but the 1080 won't go above base clock of 1721 MHz.

GPU-Z is saying the card is capped due to utilization, but I'm not buying it. I'm running a i7-6700k @ 4.4GHz that is stable at ~40% utilization, so I don't think the bottleneck is there. NVIDIA control panel set to prioritize performance, Windows power management is the same. PSU is rated for 750W.




Any thoughts on getting this thing running faster?

 

[ATMINSIDE]

What are you using to load the GPU?

 

[JrClocker]

If you are going to overclock the Pascal cards, you need to be using the EVGA Precision X OC (or MSI Afternburner).

I use EVGA Precision X OC myself.

 

[ryoungblood]

Folding @ Home

EDIT: Downloaded XOC and it's showing the same thing. Though it did run at 2088~ for 3 or 4 minutes before dropping back to 1721.

 

[Alaric]

I've had pretty decent results with Galax's Xtreme Tuner Plus. I did one click on the 5% OC button and got 2114 MHz. I haven't needed to try any higher clocks (1080p) but it seems to have the necessary goodies. Voltage for the GPU, memory, and % of over voltage.

 

[ryoungblood]

It's starting to happen on my other machine with the two 1070. Given that it seems to be related to power utilization, I think it's an issue with Windows 10.

 

[Johan45]

Have you raised the power limit on the card? Most of those EVGA cards have a dual BIOS one is usually for OC with higher limits have you tried tht?

 

[ryoungblood]

I raised the voltage to 100% via Precision XOC but am still seeing intermittent clock speeds above 1721.



Do I really need to push this thing above 100%?

 

[ATMINSIDE]

Done a clean install of the driver?

 

[ryoungblood]

Just a few minutes ago. Running 372.90 which is the last stable driver for F@H.

 

[Johan45]

Not voltage , the power target is key. I would leave voltage at stock to begin with since it will push you closer to the power target limit. Have you tried the seconday BIOS. On my 980Ti classified it raised the power target to 140%

 

[ryoungblood]

I'll try that.

Secondary power mode appears to keep the OC stable. Thanks for your help, all.

 

[Vellinious]

Use the voltage / frequency curve tool. It'll help you lock in a voltage for a specific clock, and take some of the guess work out of boost 3.0. Not ALL, but at least SOME. PCX kinda sucks for this. I switched to MSI's AB. The graph is much easier to work with.

 

[ryoungblood]

Multiple Pascal Cards Stuck @ Base Clock

I posted over a week ago about my EVGA 1080 FTW Hybrid downclocking itself to its base clock. Well, it's happening again, and now it's happening frequently on 3/6 of my Pascal cards.

It's an intermittent issue and I can't seem to figure out what is causing it. Right now, across 3 machines (see attached picture), I have 2/6 cards stuck at their base clocks. I have Windows set to maximum performance, and NVIDIA set to prioritize performance as well. The cards are downclocked when running below 65 degrees and drop down to roughly 40-50% power utilization, so it's not a temp issue. GPU-Z on each machine points at Util.

All 3 computers are running Windows 10 Pro 64-bit and running 372.90 drivers (best for F@H). I have the issue on both EVGA XOC and MSI Afterburner. All power levels are set to their maximum for each card for MSI Afterburner. All builds have at least a 750W PSU.

I doubt it's a CPU issue, because it's happening on across 3 different CPUs, including a i7 6700k, i5 2500k, and an AMD FX-8320E.

4 of the cards are EVGA (1x 1080 FTW Hybrid, 1x 1070 SC, 2x 1070 SC Black)
2 of the cards are ASUS 1070 STRIX OC

See picture for reference.

 

[EarthDog]

Merged the duplicate threads... best to continue on with this thread as it has all previous troubleshooting, etc... so we don't have to start over.

 

[ryoungblood]

Sorry about that.

I've changed the title of this thread and added the main text to the first post.

 

[ryoungblood]

SOLVED: Google Chrome, Slack, Remote Desktop (Windows RDP), when open, throttles a GPU to its base clock.

 

[ATMINSIDE]

Wow, interesting! Thanks for following back with that.

 

[ryoungblood]

It appears the Slack desktop client and Skype do the same. I have a feeling it's related to software using DirectX.

My hypothesis is that Windows has a priority system and will assign GPU use to active applications using DirectX. You can identify offending applications by individually opening and closing them while monitoring GPU power utilization. For my 1080, I see power use drop to a forced 40% or lower when one of these applications are running (active or background). Upon closing the application, GPU power use will increase to its non-limited level and will regain clock set by OC software or other.