Message boards :
Number crunching :
Titan V core clocks in compute mode
Message board moderation
Previous · 1 · 2
Author | Message |
---|---|
![]() ![]() ![]() Send message Joined: 29 Apr 01 Posts: 13164 Credit: 1,160,866,277 RAC: 1,873 ![]() ![]() |
I power limit the 1080 TI to 215W and the RTX 2080's to 200. I don't see any observable increase in compute times between power limited and full power settings. The clocks still stay up where GPU Boost lets them go. Seti@Home classic workunits:20,676 CPU time:74,226 hours ![]() ![]() A proud member of the OFA (Old Farts Association) |
Tod Send message Joined: 17 Apr 99 Posts: 27 Credit: 143,685,603 RAC: 0 ![]() |
Good to know :-) My electricity bill is big enough as it is. lol |
![]() ![]() ![]() Send message Joined: 19 May 99 Posts: 766 Credit: 354,398,348 RAC: 11,693 ![]() ![]() |
Hi All, Since the topic has migrated off the Titan V, is there a simple way to speed up the GTX 10x0 series cards on linux to get more throughput. I'm running the special app and seeing 98-99% GPU utilization via nvidia-smi, but power for some cards is well below max. Perhaps it is being temperature throttled? ![]() |
Ian&Steve C. ![]() Send message Joined: 28 Sep 99 Posts: 4267 Credit: 1,282,604,591 RAC: 6,640 ![]() ![]() |
Hi All, what are your temps? unless you're over 85C, you likely aren't being throttled due to temps. power use will usually be below max some cards more than others (1080tis use more of their power limit than say a 1060). the WUs don't load the card fully for 100% of the run time. loading is low in the beginning and only really ramps up in the latter half of the processing run time. if you're seeing 98+% GPU utilization, you're fine. but to speed things up, you can always overclock. specifically the memory. I clock all my cards' memory to what their P0 values would be, but in the P2 state. so, 1080ti : +1000 MHz mem (11000 MHz effective) 1080 : +1000 MHz mem (10000 MHz effective) 2070 : +400 MHz mem (14000 MHz effective) 1060: +400 MHz mem (8000 MHz effective) and I OC the GPU core about 75-100 MHz to get them all around 2000MHz this is just a small OC really. I could probably push it more, but these are all very stable for me. probably a good idea to use the keepP2 script from petri to avoid system crashes from temporary "over-overclock" when switching WUs. Seti@Home classic workunits: 29,492 CPU time: 134,419 hours ![]() ![]() |
![]() ![]() ![]() Send message Joined: 29 Apr 01 Posts: 13164 Credit: 1,160,866,277 RAC: 1,873 ![]() ![]() |
I agree with Ian's assessment. You really can't do much more with the special app other than run the -nobs parameter which you are doing. The next thing is to reclaim the P0 memory clocks that Nvidia penalizes you when you run any kind of compute load on consumer cards. Get the memory clocks back to where they would be if you were running a game by adding an + offset to the P2 state in Nvidia X Server Settings PowerMizer page. If you do, I too would recommend running Petri's keepP2 script in the background to prevent the card's from crashing in the brief transition to P0 state when a task finishes and before the next one starts. Seti@Home classic workunits:20,676 CPU time:74,226 hours ![]() ![]() A proud member of the OFA (Old Farts Association) |
![]() ![]() ![]() Send message Joined: 19 May 99 Posts: 766 Credit: 354,398,348 RAC: 11,693 ![]() ![]() |
Ian, Keith, Thanks for the suggested values. I tried setting them via the nvidia-smi -i 1 (etc) -ac 11000,2000 command, but it said not supported, even for the 1080 ti. The Nvidia X Server Settings PowerMizer page doesn't seem to have any options, other than setting Preferred Mode. All options there mainly have me at Level 2 out of 3, which an occasional bump up to 3. BTW, My 1080 is at 82C, 1080ti at 76C. What am I missing on how to set the overclock? ![]() |
juan BFP ![]() ![]() ![]() ![]() Send message Joined: 16 Mar 07 Posts: 9786 Credit: 572,710,851 RAC: 3,799 ![]() ![]() |
What am I missing on how to set the overclock? Go to gpuug.tech there are few scripts to make that automatically for you. If you OC not forget to set the fan control to keep your GPU on a safe zone. ![]() |
![]() ![]() ![]() Send message Joined: 29 Apr 01 Posts: 13164 Credit: 1,160,866,277 RAC: 1,873 ![]() ![]() |
Ian, Keith, The old -ac parameter is not valid on Pascal or Turing cards, only on Kepler and Maxwell. First thing is you have to set Persistent mode and then unrestricted mode with: #!/bin/bash /usr/bin/nvidia-smi -pm 1 /usr/bin/nvidia-smi -acp UNRESTRICTED Next follow it up with the overclock for each card in this format as an example: #!/bin/bash /usr/bin/nvidia-settings -a "[gpu:0]/GPUPowerMizerMode=1" /usr/bin/nvidia-settings -a "[gpu:1]/GPUPowerMizerMode=1" /usr/bin/nvidia-settings -a "[gpu:2]/GPUPowerMizerMode=1" nvidia-smi -i 1 -pl 200 /usr/bin/nvidia-settings -a "[gpu:0]/GPUFanControlState=1" /usr/bin/nvidia-settings -a "[fan:0]/GPUTargetFanSpeed=100" /usr/bin/nvidia-settings -a "[fan:1]/GPUTargetFanSpeed=100" /usr/bin/nvidia-settings -a "[gpu:1]/GPUFanControlState=1" /usr/bin/nvidia-settings -a "[fan:2]/GPUTargetFanSpeed=100" /usr/bin/nvidia-settings -a "[gpu:2]/GPUFanControlState=1" /usr/bin/nvidia-settings -a "[fan:3]/GPUTargetFanSpeed=100" /usr/bin/nvidia-settings -a "[gpu:0]/GPULogoBrightness=20" /usr/bin/nvidia-settings -a "[gpu:1]/GPULogoBrightness=20" /usr/bin/nvidia-settings -a "[gpu:2]/GPULogoBrightness=20" /usr/bin/nvidia-settings -a "[gpu:0]/GPUMemoryTransferRateOffset[4]=400" -a "[gpu:0]/GPUGraphicsClockOffset[4]=60" /usr/bin/nvidia-settings -a "[gpu:1]/GPUMemoryTransferRateOffset[3]=800" -a "[gpu:1]/GPUGraphicsClockOffset[3]=30" /usr/bin/nvidia-settings -a "[gpu:2]/GPUMemoryTransferRateOffset[3]=2000" -a "[gpu:2]/GPUGraphicsClockOffset[3]=30" In this example I have restricted power usage on the RTX 2080 to 200 Watts. Then I turned all the gpu fans up to 100% speed. Then I turned down the logo brightness on each card and finally followed everything up with my desired overclocks on each card. The cards in use in this example are a RTX 2080, GTX 1080 and then a GTX 1070 as gpu0, gpu1 and gpu2. Gpu0 was a bit of learning experience since I had to discover that since it has two discrete fans on it, they have to be enumerated separately which throws the fan count off by 1. Thus the strange fan control settings for each card. Seti@Home classic workunits:20,676 CPU time:74,226 hours ![]() ![]() A proud member of the OFA (Old Farts Association) |
![]() ![]() ![]() Send message Joined: 19 May 99 Posts: 766 Credit: 354,398,348 RAC: 11,693 ![]() ![]() |
Keith, Thanks, I tried that and things failed at the TargetFanSpeed for fan:0 with this error: ERROR: Error assigning value 100 to attribute 'GPUTargetFanSpeed' (SETI64-Ubuntu:1[fan:0]) as specified in assignment '[fan:0]/GPUTargetFanSpeed=100' (Unknown Error). Here was the script I had built so far: #!/bin/bash nvidia-smi -pm 1 nvidia-smi -acp UNRESTRICTED /usr/bin/nvidia-settings -a "[gpu:0]/GPUPowerMizerMode=1" /usr/bin/nvidia-settings -a "[gpu:1]/GPUPowerMizerMode=1" /usr/bin/nvidia-settings -a "[gpu:2]/GPUPowerMizerMode=1" /usr/bin/nvidia-settings -a "[gpu:3]/GPUPowerMizerMode=1" /usr/bin/nvidia-settings -a "[gpu:0]/GPUFanControlState=1" /usr/bin/nvidia-settings -a "[fan:0]/GPUTargetFanSpeed=100" /usr/bin/nvidia-settings -a "[gpu:0]/GPUMemoryTransferRateOffset[3]=200" -a "[gpu:0]/GPUGraphicsClockOffset[3]=50" I'm running Nvidia driver 410 (open source) ![]() |
Ian&Steve C. ![]() Send message Joined: 28 Sep 99 Posts: 4267 Credit: 1,282,604,591 RAC: 6,640 ![]() ![]() |
pretty sure you nee to run the coolbits command and reboot before you'll be able to change the fan speeds and whatnot. run this, then reboot: sudo nvidia-xconfig --thermal-configuration-check --cool-bits=28 --enable-all-gpus then you will be able to run overclocking and fan speed commands in the terminal or via the nvidia-settings GUI Seti@Home classic workunits: 29,492 CPU time: 134,419 hours ![]() ![]() |
![]() ![]() ![]() Send message Joined: 19 May 99 Posts: 766 Credit: 354,398,348 RAC: 11,693 ![]() ![]() |
|
![]() ![]() ![]() Send message Joined: 29 Apr 01 Posts: 13164 Credit: 1,160,866,277 RAC: 1,873 ![]() ![]() |
Be prepared for the coolbits tweak to change gpu enumeration. It always does for me. Others state they never have issues. Just a warning, just in case. Seti@Home classic workunits:20,676 CPU time:74,226 hours ![]() ![]() A proud member of the OFA (Old Farts Association) |
![]() ![]() ![]() Send message Joined: 19 May 99 Posts: 766 Credit: 354,398,348 RAC: 11,693 ![]() ![]() |
|
![]() ![]() ![]() Send message Joined: 29 Apr 01 Posts: 13164 Credit: 1,160,866,277 RAC: 1,873 ![]() ![]() |
Good to hear Roger. Now you just need to test for stability for overclock amounts the cards are happy with. Seti@Home classic workunits:20,676 CPU time:74,226 hours ![]() ![]() A proud member of the OFA (Old Farts Association) |
![]() Send message Joined: 6 Jun 02 Posts: 1668 Credit: 623,086,772 RAC: 156 ![]() ![]() |
Hi everyone, I have not had much time to twiddle with the TITAN V. My first impressions and after thaughts after a quick test were that the P0 enables the graphics processing units too and the power usage jumps to twice the normal and no speed gain is really gained in Seti. I'm runnig still at P2 with keepP2 with +220 for the GPU and +220 for the memory. I'm waiting for the Tuesday outages to align with my free time. Being a teacher the rest of the spring is going to be quite hectic at work. Well -- the summer is coming at a fast pace. There is a noticeably less snow here now. The roads are ice free again! I'll keep reading about your experiments and results with the new options. Thank you for your good work! Petri p.s. A new version (0.98) is still being tested by me and TBar and a couple of not so random Setizens. Memory usage will drop. Some speed gain can be expected. To overcome Heisenbergs: "You can't always get what you want / but if you try sometimes you just might find / you get what you need." -- Rolling Stones |
©2025 University of California
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.