NVIDIA GPU Fan Control using GUI in Linux

Message boards : Number crunching : NVIDIA GPU Fan Control using GUI in Linux
Message board moderation

To post messages, you must log in.

Previous · 1 · 2 · 3

AuthorMessage
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 13161
Credit: 1,160,866,277
RAC: 1,873
United States
Message 1882594 - Posted: 7 Aug 2017, 8:14:23 UTC

So your thinking is logical. Then logic would seem to point at the pump control somehow getting put into an invalid state that equates to the observed pump rpm. But 400 rpm is NOT one of the predefined pump speeds. The slowest predefined pump speed is Quiet which runs at around 915 rpm. I will need to pull the link control cable and cause the system to black screen on purpose. Shouldn't be too difficult...... just choose 3.9 Ghz on the overclock which I have never been able to make stable because that is where my chip hits the wall.
Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)
ID: 1882594 · Report as offensive
Grant (SSSF)
Volunteer tester

Send message
Joined: 19 Aug 99
Posts: 13720
Credit: 208,696,464
RAC: 304
Australia
Message 1882596 - Posted: 7 Aug 2017, 8:49:00 UTC
Last modified: 7 Aug 2017, 8:49:53 UTC

USB is a very useful interface. However badly behaving drivers, and devices, can cause all sorts of weird grief.

My i7's network port got taken out by a lightning strike to the phone line- it found it's way to ground through the modem & out it's network port and through the computer. So I have to rely on a USB wireless network adaptor. Every so often I loose my internet connection (sometimes the indicator on the Task Bar shows a problem, other times it shows as normal). Pull the adaptor out, plug it back in a few seconds later & the Internet works again.
Then there was the time the system started locking up at random times, one time it managed to throw up a message on re-booting about an "Unrecognised USB device" before crashing & burning again. I hadn't had been having any internet issues, so I unplugged my Kindle. Issue solved. The battery on the Kindle won't fully charge & somehow it had gotten itself all confused, and it was bringing the computer down at random times through the USB connection.
Grant
Darwin NT
ID: 1882596 · Report as offensive
Profile Brent Norman Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester

Send message
Joined: 1 Dec 99
Posts: 2786
Credit: 685,657,289
RAC: 835
Canada
Message 1882597 - Posted: 7 Aug 2017, 9:30:40 UTC - in response to Message 1882594.  
Last modified: 7 Aug 2017, 9:34:49 UTC

Scratching head, it maybe that the PSU is detecting a 'sleep' state from the motherboard causing it to reduce power. Maybe?

EDIT: You said it also does this after repowering from a BSOD? Would the "paper clip" PSU test show different results?
ID: 1882597 · Report as offensive
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 13161
Credit: 1,160,866,277
RAC: 1,873
United States
Message 1882645 - Posted: 7 Aug 2017, 16:25:58 UTC - in response to Message 1882597.  

Scratching head, it maybe that the PSU is detecting a 'sleep' state from the motherboard causing it to reduce power. Maybe?

EDIT: You said it also does this after repowering from a BSOD? Would the "paper clip" PSU test show different results?

Yes, it will show the problem after a blue screen of death or black screen of death. If the system crashes during the early morning while I'm sleeping, I will find it back at the desktop with nothing running except SIV which is normal. The LED on the pump will be deep blue because the CPU is at room ambient of 25° C. If I then reboot the system to go into the BIOS to change whatever I changed in my overclock to put it back at one of my stable profiles, then the system quickly overtemps in the BIOS with the pump LED changing to deep red and the radiator fans go racing to full speed. The system then will shut down from the cpu telling it too. The cpu temp in the BIOS on the monitor screens shows 75° C (remember this chip has the +20 offset) and the AIO fan speed (where the pump is telling the system its speed) is only reading 400 rpm. Then I have to try restarting the system multiple times or just be patient enough for the water temp to drop enough before the system makes it back to the Windows 10 desktop where the AIO water temp will then rapidly decline to ambient, the pump is then at normal 2850 rpm, and the radiator fans return to low speed.
Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)
ID: 1882645 · Report as offensive
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 13161
Credit: 1,160,866,277
RAC: 1,873
United States
Message 1882647 - Posted: 7 Aug 2017, 16:31:34 UTC

I forgot to mention that I can enter and leave the BIOS and boot into Windows all day long without any issues, [IF] the system was shut down or restarted from a normal Windows shutdown. Only a immediately prior BSOD can cause the system to overtemp in the BIOS.
Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)
ID: 1882647 · Report as offensive
Profile Brent Norman Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester

Send message
Joined: 1 Dec 99
Posts: 2786
Credit: 685,657,289
RAC: 835
Canada
Message 1882681 - Posted: 7 Aug 2017, 21:01:11 UTC - in response to Message 1882645.  

I believe you said you can set a fan/pump curve in SIV (I'm thinking of PrecisionX type setting). Can you set a minimum of say 60% cooling so that it never goes to 'quiet' mode? After all it's a cruncher and should never get cold :D

That way even if it goes to a minimum value, the CPU should just get uncomfortably warm, and not hot.

I can't believe that Corsair would put out a product that powers up at 'last known' state and not some default value. That is just stupid!
ID: 1882681 · Report as offensive
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 13161
Credit: 1,160,866,277
RAC: 1,873
United States
Message 1882683 - Posted: 7 Aug 2017, 21:29:01 UTC - in response to Message 1882681.  
Last modified: 7 Aug 2017, 21:38:19 UTC

I believe you said you can set a fan/pump curve in SIV (I'm thinking of PrecisionX type setting). Can you set a minimum of say 60% cooling so that it never goes to 'quiet' mode? After all it's a cruncher and should never get cold :D

That way even if it goes to a minimum value, the CPU should just get uncomfortably warm, and not hot.

I can't believe that Corsair would put out a product that powers up at 'last known' state and not some default value. That is just stupid!

The fans are already on an aggressive fan curve and they mostly stay at 100% all the time unless I am idle and not crunching. The pump is set for maximum pwm of 100% at 255 all the time. There is no curve profile on the pump.

Well, if you lurk a bit on the Corsair Cooling forums a bit, you will see that most of the posts involve issues with their Link software and how poorly engineered it is. Ray Hinchliffe, developer of SIV has constantly pointed out to Corsair all the bugs in their software and has fixed almost all of them for SIV users or has developed work-a-rounds for bugs that should have been fixed by Corsair but have been ignored or never acknowledged by them for years.

The Asetek hardware and firmware is especially stupid. One of the reasons he always recommends the CooliT hardware and firmware instead since it is properly written.

Corsair Cooling Forum

Using SIV to Control Corsair Link Hardware.pdf

[Edit] Was cruising around the Corsair forum and stumbled onto this post by Ray. That reminded me of the stupidity of my Corsair H100iV2 AIO that is on the linux cruncher. Forgot about its quirks. I guess the pump is running at full speed after all when psensors and GKrellm monitors report its pumps speed at 1430 rpm. It is really running at 2860 rpm.



[red-ray]: The H100iV2 is silly and reports half the pump speed
The H100iV2 is silly and reports half the pump speed, so the actual pump speed is about 2860 RPM.

I suspect this is far faster than needed, but you need to USB connect the cooler and use software to set a lower speed.

Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)
ID: 1882683 · Report as offensive
Profile Jeff Buck Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester

Send message
Joined: 11 Feb 00
Posts: 1441
Credit: 148,764,870
RAC: 0
United States
Message 1884361 - Posted: 17 Aug 2017, 2:20:24 UTC
Last modified: 17 Aug 2017, 2:21:53 UTC

I've uploaded a new version (v0.4.1) with just a couple minor changes, for anyone still interested in trying this out.

W3Perl identified a text alignment issue for one of the captions when the app was run on Mageia 6 systems. That has been resolved. He also requested a "Revert" button, to revert to the last saved settings after trying out one or more new settings. That button has been added. (It will only appear after at least one setting has been changed subsequent to the last save.)

NOTE: The distribution zip file no longer includes the compiled GUI file, gpufancnvgui.pyc, just on the off-chance that there might be any conflict when a file compiled on my Ubuntu 14.04 system runs under another Linux OS. A new gpufancnvgui.pyc should be created the first time the app is run. For current users of an earlier version, it may be advisable to delete any existing gpufancnvgui.pyc that you have, although Python should recognize the older datestamp and recompile automatically. A compiled file isn't necessary for the app to run, though it theoretically improves load time after the first execution, for whatever that's worth. :^)

EDIT: The download link remains the same as in my first post in the thread.
ID: 1884361 · Report as offensive
Stephen "Heretic" Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 20 Sep 12
Posts: 5557
Credit: 192,787,363
RAC: 628
Australia
Message 1884454 - Posted: 17 Aug 2017, 14:21:30 UTC - in response to Message 1881977.  

Well, Stephen did raise his hand.......but perhaps he was just indicating he needed to go to the bathroom. ;^)


. . Sorry gentlemen. La_Bamba died, she started throwing computational errors and resetting. I have tried a few things but no joy, it seems to be either the CPU or MoBo that has gone to cyber-heaven, so I bought a new rig. An ASUS prime x370 Pro and a Ryzen 7-1700, but no joy there either, it will not boot up ... :(

. . The long and the short of it is that I have not gotten around to trying it out, I am still running with the original terminal script-based version which is doing the trick. I did download the files but I have not d/l'd the other app you say is necessary to run it (which I cannot now remember now, was it Python?). When I resolve my other issues I will jump back in on that one. I am sorry if I let you down ... :(

Stephen

:(
ID: 1884454 · Report as offensive
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 13161
Credit: 1,160,866,277
RAC: 1,873
United States
Message 1884493 - Posted: 17 Aug 2017, 17:02:06 UTC - in response to Message 1884454.  

Python is almost always included automatically in every distribution I can think of. You don't need to download anything if I remember correctly. If you type python in Terminal, that should present you the Python command line >>> Type quit () or CTRL-D to quit back to Terminal. That will tell you if you have the Python package already.

I can't remember Steven, was this the same AM4 setup you had before? I thought you had troubleshot and determined everything had to go back to vendor. I would have thought you would have had replacements already. Or maybe I am confusing you with someone else. I will have to go back to the original thread I guess.
Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)
ID: 1884493 · Report as offensive
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 13161
Credit: 1,160,866,277
RAC: 1,873
United States
Message 1884494 - Posted: 17 Aug 2017, 17:04:02 UTC

I haven't gotten around to trying version 0.4 yet. Downloaded already. Was busy recovering a hard drive yesterday. Should get to a test today.
Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)
ID: 1884494 · Report as offensive
Profile Jeff Buck Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester

Send message
Joined: 11 Feb 00
Posts: 1441
Credit: 148,764,870
RAC: 0
United States
Message 1884521 - Posted: 17 Aug 2017, 18:24:56 UTC - in response to Message 1884454.  
Last modified: 17 Aug 2017, 18:25:36 UTC

. . The long and the short of it is that I have not gotten around to trying it out, I am still running with the original terminal script-based version which is doing the trick. I did download the files but I have not d/l'd the other app you say is necessary to run it (which I cannot now remember now, was it Python?). When I resolve my other issues I will jump back in on that one. I am sorry if I let you down ... :(

Stephen

:(
No worries, and certainly no letdown, Stephen. I wrote the thing purely for my own benefit, so if anyone else finds it useful, that's just a bonus!

As Keith noted, Python generally gets installed automatically as part of most flavors of Linux. However, you probably will need to add the "python-numpy" package. Keith also found that his setup was missing "python-qt4", although that seemed to come standard with mine. Check with Package Manager for both of those.

@Keith
The v0.4.1 distribution just has minor changes. If you didn't see the caption alignment problem before, you shouldn't see anything different in that area. (The "Fan Control:" text was overlaying the "Off" button on Mageia 6.) The "Revert" button is primarily useful if you're playing around with fine-tuning your settings but want a quick way to jump back to the last settings that you saved.
ID: 1884521 · Report as offensive
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 13161
Credit: 1,160,866,277
RAC: 1,873
United States
Message 1884530 - Posted: 17 Aug 2017, 19:16:02 UTC - in response to Message 1884521.  

Thanks for reminding me Jeff that I did have to install those two packages. Package installation is so simple in Linux that one forgets that you even have done it. I take that there is a further revision to the 0.4.1 release from your last post.
Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)
ID: 1884530 · Report as offensive
Profile Jeff Buck Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester

Send message
Joined: 11 Feb 00
Posts: 1441
Credit: 148,764,870
RAC: 0
United States
Message 1884536 - Posted: 17 Aug 2017, 19:23:18 UTC - in response to Message 1884530.  

I take that there is a further revision to the 0.4.1 release from your last post.
Nope, v0.4.1 is the latest (for now). The two minor changes I mentioned are included in that release.
ID: 1884536 · Report as offensive
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 13161
Credit: 1,160,866,277
RAC: 1,873
United States
Message 1884562 - Posted: 17 Aug 2017, 21:14:12 UTC

Jeff, followed the readme to delete the previous gui file and removed the cfg file from the archive and copied everything over to desktop folder. Stopped the app and restarted and up comes v0.4.1 and everything running as it should. So Python recompiled the gui file as you said it would. Good Job! Thanks
Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)
ID: 1884562 · Report as offensive
Profile Jeff Buck Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester

Send message
Joined: 11 Feb 00
Posts: 1441
Credit: 148,764,870
RAC: 0
United States
Message 1884603 - Posted: 17 Aug 2017, 23:41:24 UTC - in response to Message 1884562.  

Jeff, followed the readme to delete the previous gui file and removed the cfg file from the archive and copied everything over to desktop folder. Stopped the app and restarted and up comes v0.4.1 and everything running as it should. So Python recompiled the gui file as you said it would. Good Job! Thanks
You're welcome! And thank you for reporting your observations.
ID: 1884603 · Report as offensive
Profile Jeff Buck Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester

Send message
Joined: 11 Feb 00
Posts: 1441
Credit: 148,764,870
RAC: 0
United States
Message 1886688 - Posted: 28 Aug 2017, 22:15:46 UTC

In view of the recent discussion of 19-slot MBs and the like, I realized that somewhere along the line somebody here might just be insane determined enough to go for the gold (or at least a silver or bronze). And suddenly the 6 GPU upper limit I had built into this app didn't look like such a good idea, at least for the long haul. So I've done just a bit of redesign that should allow the app to handle however many NVIDIA GPUs some future champion decides to throw at it.

The individual radio buttons have been replaced with a spin box and the app should automatically detect how many NVIDIA GPUs there are on your system. The rest of the functionality remains the same. This redesign actually results in a slightly more compact window for the app, like so:



The new version, v0.5.1, is available from the same download link as before, on DropBox.

The distribution files no longer include either the compiled version of the GUI (Python should recompile on first execution) or a default configuration file. The defaults are now built-in, which should alleviate any risk of existing users overlaying their current config file with a newly distributed default one. Existing users shouldn't need to make any configuration changes due to the new release. Just copy the new files over the old ones and keep on truckin'.
ID: 1886688 · Report as offensive
Profile Keith Myers Special Project $250 donor
Volunteer tester
Avatar

Send message
Joined: 29 Apr 01
Posts: 13161
Credit: 1,160,866,277
RAC: 1,873
United States
Message 1886729 - Posted: 29 Aug 2017, 3:20:52 UTC

Just a quick FYI, Jeff. The v0.5.1 works fine. Only complaint is the spin button only works if you place your cursor in the number box, not intuitively using the up/down arrows. A bit slow to respond also.
Seti@Home classic workunits:20,676 CPU time:74,226 hours

A proud member of the OFA (Old Farts Association)
ID: 1886729 · Report as offensive
Profile Jeff Buck Crowdfunding Project Donor*Special Project $75 donorSpecial Project $250 donor
Volunteer tester

Send message
Joined: 11 Feb 00
Posts: 1441
Credit: 148,764,870
RAC: 0
United States
Message 1886731 - Posted: 29 Aug 2017, 3:43:37 UTC - in response to Message 1886729.  

Just a quick FYI, Jeff. The v0.5.1 works fine. Only complaint is the spin button only works if you place your cursor in the number box, not intuitively using the up/down arrows. A bit slow to respond also.
Thanks, Keith. I did notice that slow response, but it still should work using just the up/down arrows. On my machines, it just seems to take a second or two longer than seems normal. I'm going to play with it some more, though, because I had some sort of recursion problem with it initially, that caused it to treat a single click as two clicks, cycling from 0 to 1 to 2, for instance, from just a single click. I put a couple extra lines in the code to work around that, but....?? It should behave exactly the same as the control interval spinner, but it doesn't (yet). I expect I'll have a sudden D'oh moment at some point, and discover that it's some Python gotcha that I overlooked.
ID: 1886731 · Report as offensive
Previous · 1 · 2 · 3

Message boards : Number crunching : NVIDIA GPU Fan Control using GUI in Linux


 
©2024 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.