LTT Official Folding Month V (2022)

[Heats with Nvidia]

6 minutes ago, Alex Atkin UK said:

The BIOS fan profile?  Yeah I gave up on that almost immediately as it absolutely bakes the card.

Its currently running the BIOS fan profile and it currently sits at 80°C core 90°C memory and 94°C hot spot and runs at 1800MHz.

Although the room it is in currently only has 15°C, i think thats not too bad, especially for a dying fan. So i think, replacing the thermal pads and thermal paste really helped.

[Alex Atkin UK]

4 minutes ago, Heats with Nvidia said:

Its currently running the BIOS fan profile and it currently sits at 80°C core 90°C memory and 94°C hot spot and runs at 1800MHz.

Although the room it is in currently only has 15°C, i think thats not too bad, especially for a dying fan. So i think, replacing the thermal pads and thermal paste really helped.

Yeah mine is running in a 21C room and its certainly an interesting comparison.  Mine doesn't seem to be reporting memory temps?

[dogwitch]

9 minutes ago, Heats with Nvidia said:

Its currently running the BIOS fan profile and it currently sits at 80°C core 90°C memory and 94°C hot spot and runs at 1800MHz.

Although the room it is in currently only has 15°C, i think thats not too bad, especially for a dying fan. So i think, replacing the thermal pads and thermal paste really helped.

which gpu is that temps on?

 

[Heats with Nvidia]

1 minute ago, dogwitch said:

which gpu is that temps on?

 

Asus RTX2080 Turbo

[Alex Atkin UK]

The stock fan profile maxes out at around 50% to maintain quiet operation, it kinds sucks as it needs about 80% hair dryer volume to remain sanely cooled.

[RollinLower]

17 minutes ago, Alex Atkin UK said:

The stock fan profile maxes out at around 50% to maintain quiet operation, it kinds sucks as it needs about 80% hair dryer volume to remain sanely cooled.

same under linux on my quadro. GPU gets to 80C but fan stays at max 60%. also no easy way to change that in Nvidia-smi that i can tell sadly.

 

right now i mitigated it somewhat by having the rest of the server it's in just blast air at the card.

[Gorgon]

2 hours ago, RollinLower said:

same under linux on my quadro. GPU gets to 80C but fan stays at max 60%. also no easy way to change that in Nvidia-smi that i can tell sadly.

 

right now i mitigated it somewhat by having the rest of the server it's in just blast air at the card.

1. disable Wayland:

sudo nano /etc/gdm3/custom.conf
uncomment: #WaylandEnable=false

    2. Configure Coolbits:

sudo nvidia-xconfig -a --cool-bits=12 --allow-empty-initial-configuration

    3. Disable Modeset:

sudo nano /usr/lib/modprobe.d/nvidia-kms.conf
	change:
	options nvidia-drm modeset=1 to options nvidia-drm modeset=0

   4. Reboot

   5. Find user id for gdm3 Xauth:

sudo ps -ef | grep auth

   6. Use UserID for commands:

sudo DISPLAY=:0 XAUTHORITY=/run/user/<UserID>/gdm/Xauthority nvidia-settings \		
 -a [gpu:0]/GPUGraphicsClockOffsetAllPerformanceLevels=50 \
 -a [gpu:0]/GPUFanControlState=1 \
 -a [fan:0]/GPUTargetFanSpeed=75 \
 -a [fan:1]/GPUTargetFanSpeed=75

The above is for headless operation and took me months to figure out. For systems with a connected display once you get Coolbits set you can just use the nvidia-settings GUI to adjust the Fan Speed and apply a GPU offset etc.

 

If you have multiple GPUs then things get funky, Pascal GPUs only have 1 fan target per GPU (i.e. if there are 2 or more physical fans they are all controlled via a single Fan object). Turing and later have multiple fan targets per GPU but they are all exposed up to a maximum of 7 in increasing numerical order so, for example, if you have 2 RTX 2000-series dual axial GPUs the first GPUs fans will be FAN0 & 1 and the 2nd GPUs fans will be FAN2 & 3.

This should show the enumerated fans:

sudo DISPLAY=:0 XAUTHORITY=/run/user/<UserID>/gdm/Xauthority nvidia-settings -q fans

[RollinLower]

8 minutes ago, Gorgon said:

1. disable Wayland:

sudo nano /etc/gdm3/custom.conf
uncomment: #WaylandEnable=false

    2. Configure Coolbits"

nvidia-xconfig -a --cool-bits=12 --allow-empty-initial-configuration

 

    3. Disable Modeset:

sudo nano /usr/lib/modprobe.d/nvidia-kms.conf
	change:
	options nvidia-drm modeset=1 to options nvidia-drm modeset=0

   4. Reboot

   5. Find user id for gdm3 Xauth:

ps -ef | grep auth

 

   6. Use UserID for commands:

DISPLAY=:0 XAUTHORITY=/run/user/<UserID>/gdm/Xauthority nvidia-settings \		
 -a [gpu:0]/GPUGraphicsClockOffsetAllPerformanceLevels=50
 -a [gpu:0]/GPUFanControlState=0 \
 -a [fan:0]/GPUTargetFanSpeed=75 \
 -a [fan:1]/GPUTargetFanSpeed=75

The above is for headless operation and took me months to figure out. For systems with a connected display once you get Coolbits set you can just use the nvidia-settings GUI to adjust the Fan Speed and apply a GPU offset etc.

as always you're a wizard. will see if this works on the server, currently running 22.04 headless.

[Gorgon]

16 minutes ago, RollinLower said:

as always you're a wizard. will see if this works on the server, currently running 22.04 headless.

works on my 22.04 rigs which are all headless - correction this should be 1 to enable Manual Fan Control (0 is Automatic)

/GPUFanControlState=1

[Mxyzptlk]

Discovered last night that one of my foldings 3080 that is 'remote' hasn't been folding for at least a few weeks.....  Oopsie...

[rkv_2401]

On 11/8/2022 at 6:37 AM, BiotechBen said:

How do you cancel a WU? They get pushed to you and even if you tell it "fold for this condition" it just ignores it and pushes it anyway. Based on the logs I was looking through, it looks like it's trying to fold two proteins at the same time, as if it was designed to be run on a dual GPU card 

Same, I tried setting it to fold for cancer WUs because of thermal limitations but I just get the most random projects, my laptop with it's fan at 100% runs at 70-73c on Cancer WUs but I get some other work units that push it to 85c and it got so hot today on 18601 (at the lowest clock speed for the GPU) that it triggered BD-HOT and switched off. How am I supposed to deal with a 20+c temperature swing between work units? 

[Mxyzptlk]

31 minutes ago, rkv_2401 said:

Same, I tried setting it to fold for cancer WUs because of thermal limitations but I just get the most random projects, my laptop with it's fan at 100% runs at 70-73c on Cancer WUs but I get some other work units that push it to 85c and it got so hot today on 18601 (at the lowest clock speed for the GPU) that it triggered BD-HOT and switched off. How am I supposed to deal with a 20+c temperature swing between work units? 

I typically have all my rigs set to prioritize for Cancer, but there has not been many Cancer WU's lately. So in place of Cancer WU it brings up whatever is next. 

[Senzelian]

The hotel is paying for electricity ...

 

[Alex Atkin UK]

36 minutes ago, rkv_2401 said:

Same, I tried setting it to fold for cancer WUs because of thermal limitations but I just get the most random projects, my laptop with it's fan at 100% runs at 70-73c on Cancer WUs but I get some other work units that push it to 85c and it got so hot today on 18601 (at the lowest clock speed for the GPU) that it triggered BD-HOT and switched off. How am I supposed to deal with a 20+c temperature swing between work units? 

I've actually set all mine to Alzehimers  but I just don't have enough GPUs to save you.

[Gorgon]

21 hours ago, GOTSpectrum said:

@cbigfootconfirms points are good, I'll get a post out soon hopefully before I cresh, if not it was be asap tomorrow afternoon 

Is it afternoon yet?

[cbigfoot]

13 minutes ago, Gorgon said:

Is it afternoon yet?

here it is ist ~10 PM there...lol...but the #s are in the spread sheet...

[Gorgon]

1 minute ago, cbigfoot said:

here it is ist ~10 PM there...lol...but the #s are in the spread sheet...

OK - didn't see that last night - good to know

[GOTSpectrum]

46 minutes ago, Gorgon said:

OK - didn't see that last night - good to know

So I did write a blog post but I forgot to post it on here. Whoopsiedoodle

[cbigfoot]

7 minutes ago, GOTSpectrum said:

So I did write a blog post but I forgot to post it on here. Whoopsiedoodle

so you did 

 

[GOTSpectrum]

2 minutes ago, cbigfoot said:

so you did 

 

This just goes to show how important it is to follow the blog AND the thread 

 

[Gorgon]

31 minutes ago, GOTSpectrum said:

So I did write a blog post but I forgot to post it on here. Whoopsiedoodle

I suspected that might have been the case

[GOTSpectrum]

points and ranks are done, 

 

Blo0g post and ranks on the sheet will come in the next hour or two, but first, I need to cook a pizza

[cbigfoot]

Point Validation complete...

[Slave2school]

Man Down!!!

 

1000w Cougar CMX ex mining rig psu died powering the V56 and ryzen 1700 folding in my daughter's pc.  Up and running again now with a spare 500w EVGA that will most likely die as well unless I keep the gpu in power saving mode.

[GOTSpectrum]

1 hour ago, GOTSpectrum said:

points and ranks are done, 

 

Blo0g post and ranks on the sheet will come in the next hour or two, but first, I need to cook a pizza

So I may have fallen asleep at my desk. Blog post is almost done