HELP! Clock issue on GTX960 ... 200+Mhz above normal

Hi,

My GTX960 card here > http://www.gigabyte.fr/products/product-page.aspx?pid=5439#sp

It goes way to high in clock frequency … 1354Mhz ? instead of BASE: 1127 MHz / BOOST: 1178 MHz (1152 / 1203 MHz in OC Mode)

I only noticed now because i wanted to stop GPU from going up and down while on desktop or in game in order to reduce frame latency. As with my 7600GT it was 1-2ms frame time, with my GTX960 it’s above 10ms (up to 25ms).

It also seems PowerMizer is not working as i didn’t find a solution to Disable Adaptative Clocking, or to set my frenquency as i would like it to be.

Thanks.

I’ve noticed that the default clocks under Linux tend to be a bit higher than the reported stock clocks for the same GPUs under Windows, at least for Maxwell.

Frame latency is a tough nut to crack on Linux in general, due to Xorg adding a few ms of overhead, and most compositors not getting out of the way for OpenGL applications the way that they theoretically should. It is possible, though, that in your case you’re seeing TDP or temperature throttling due to the unsustainable overclock.

There is a way to allow Nvidia settings to control overclock, though – make an /etc/X11/xorg.conf containing only this:

Section “Device”
Identifier “Device0”
Driver “nvidia”
VendorName “NVIDIA Corporation”
Option “Coolbits” “8”
EndSection

And you should have another option in PowerMizer to adjust clocks up and down in terms of Mhz. See if that helps.

I’m using openbox, nvidia 355.11, xorg 1.17.2, kernel 4.2. I even tried ck-kernel 4.1.6 but… same.

So my problem is when i launch 3D Game, Max Performance on PowerMizer , GPU Clock goes at 1354 instead of staying at 1164Mhz.

I’m looking for a solution to set GPU CLOCK in order to disable all adaptative thingy.

nvidia-settings -a [gpu:0]/GPU2DClockFreqs=1164,3505
nvidia-settings -a [gpu:0]/GPU3DClockFreqs=1164,3505

It is not working… i tried many things and nothing worked as it should.

Maybe i missed something…

Did you see what I said about making an xorg.conf that will allow to you to adjust clocks in powermizer? Do that.

Yes, i did but it’s limited to -90 Mhz for me. So it’s not really a solution…

I added in nvidia-setting-rc file :

[gpu:0]/GPUGraphicsClockOffset[3]=-90
[gpu:0]/GPUPowerMizerMode=1

but GPU is still going higher than it should in 3D fullscreen, well now it’s 1264Mhz in 3D and 1164Mhz in 2D…

I found this to stop GPU from going up and down, so by default it will be 1354Mhz all time -90Mhz.

Option         "RegistryDwords" "PowerMizerEnable=0x1; PerfLevelSrc=0x2222;PowerMizerLevel=0x1; PowerMizerDefault=0x1; PowerMizerDefaultAC=0x1"

1264Mhz for 2D and 3D… :)

Thanks anyways… i guess i will have to wait nvidia to fix it.

ps: Maybe i can report to nvidia by email the problem ?

They have a bug report page, but it’s buried on their website, and typically doesn’t get looked at by anyone with experience with the Linux driver – you’ll have better luck here, just wait for a developer to come to this thread.

1264 might still be stable, though, if the manufacturer overclock goes up to 1200 – are you sure it’s still causing a problem?

I just found this as it got moved UP on forum

https://devtalk.nvidia.com/default/topic/862732/linux/980ti-boost-clock-issue/post/4626420/#4626420

and it seems he has the same problem as i do…

nvidia-smi -q -d CLOCK

==============NVSMI LOG==============

Timestamp                           : Fri Sep 25 14:35:22 2015
Driver Version                      : 355.11

Attached GPUs                       : 1
GPU 0000:01:00.0
    Clocks
        Graphics                    : 135 MHz
        SM                          : 135 MHz
        Memory                      : 405 MHz
    Applications Clocks
        Graphics                    : 1164 MHz
        Memory                      : 3505 MHz
    Default Applications Clocks
        Graphics                    : 1164 MHz
        Memory                      : 3505 MHz
    Max Clocks
        Graphics                    : 1430 MHz
        SM                          : 1430 MHz
        Memory                      : 3505 MHz

After some tests with unigine heaven benchmark… GPU is at 1354Mhz at 60°c, 55°c at 1264Mhz.

Can someone from NVIDIA fix it please, i only want to run it at BASE clock and the BOOST is wrong.

My card is not made for it, gigabyte says BASE: 1127 MHz / BOOST: 1178 MHz…

I want it stable, lower temperature, and that GPU dosn’t eat my PSU as it would affect my system.

Thanks.

Don’t graphics card these days have a power target? Meaning that as long as there is thermal headroom they’ll try to increase the clockspeeds.

EDIT: I had to use an URL shortener for this Youtube link, because the forum tried to embed it which didn’t really work. Even when putting URL tags around it.

So, if it’s GPU BOOST 2.0 fault, how i turn it off ?

How can i set max temp if it’s based on it ?

ps : when i use max power saving, my PCIeLinkSpeed is reduced, is there a way to set it to max ?

Option  "RegistryDwords" "PowerMizerEnable=0x1; PerfLevelSrc=0x2222; PowerMizerLevel=0x3; PowerMizerDefault=0x3; PowerMizerDefaultAC=0x3"

As i could just overclock my GPU/MEM at normal clocks speed, and my problem would be fixed for good! (i think i can’t overclock when it’s like that, need to test it again, later, lol)

60c is not hot for a modern GPU. I think blackout is correct that the card might actually be totally stable at this point unless you’re seeing obvious throttling behaviour.

60 Degrees Celsius is nothing. That’s basically luke warm far a GPU. Even up to 90 C° is basically ok.

One way to do this is to use Maxwell BIOS tweaker via wine and nvflash for linux.

In there you can limit all power states, frequencies and etc. There are youtube guides on how to do that on windows, virtually the only difference on linux is running maxwell tweaker via wine and using linux native nvflash in linux shell instead of windows cmd.exe.