• Welcome to Overclockers Forums! Join us to reply in threads, receive reduced ads, and to customize your site experience!

Maybe shutting down my 4P

Overclockers is supported by our readers. When you click a link to make a purchase, we may earn a commission. Learn More.

Junebug

Member
Joined
May 22, 2006
Location
Asheville NC
It may be the end of an era. I may have to shutdown my 4P (Adak's old Tanker). I love this beast and for a while it sat happily cranking out 150-200K points a day. At that rate I could live with the electric costs. But in the past month or so a new set of WU (like 14025) have come out that take 3 to 4 days to complete and only yield 760 points per day (you ready that correctly).

Right now its still heating season, so I have the beast running in my office keeping me warm while I work. But I think come Spring I will have to shut it down unless things change.

Does anyone know whats up with the resource intensive WU that yield very little in points?

Regards,

Junebug

PS... gulp35 should hit 200 million in the next few days!
 
Sorry to hear, I'm glad that Adak's Tanker got to fold on while under your care. :salute:

The Supermicro in my sig is only good for CPU folding for "old times sake". It does catch Rosetta@Home duty a couple times a month.

Electricity consumption is an issue here too, and being on a fixed income I have to go sparingly on my folding binges.
 
Do not run CPU wu very often or have p14025 in my HFM history, but do have a p14024 run with CPU8 on i9-7940X that ran 2min 38 sec TPF, or 4.4 hours total for 19,243 est. credit.

Likely something is not working correctly on Tanker.

ppd calculator shows 3.0 days on a 14025 would be 716 ppd and the wu only has 1183 base credit, so appears that you are still receiving credit for the quick return bonus. Just not much due to being slow.

Rebooting the host often returns things to normal. Sometimes need to delete the files in the work folder.

Bad fan(s) or dust could cause thermal throttling.

Bad wu usually get some partial credit, but the quick return bonus will be lost if too many failures.

Does the log file show anything unusual?
 
Last edited:
Whats the PCIe config look like on that board? It might be able to find a home on my farm if it could could power at lest two GPUs at x16?

Edit to say, at the moment, I don't really consider power/cooling a problem, come summer, I'll have to evaluate.
 
Are 14025's a7 cores? I see some people complaining there isn't enough a7 cores available, maybe you are picking up a4 core WUs?
 
Hi all,

I do think Tanker is working correctly. Yesterday it picked up a WU and was running at 160k ppd. But later yesterday, it picked up a 13797...now cranking along, a 4 day run, for 1046 ppd. I looked,and this is supposed to be a A& core so I do not think that is it. Not sure why Tanker finds these so hard to digest.
 
Hi all,

I do think Tanker is working correctly. Yesterday it picked up a WU and was running at 160k ppd. But later yesterday, it picked up a 13797...now cranking along, a 4 day run, for 1046 ppd. I looked,and this is supposed to be a A& core so I do not think that is it. Not sure why Tanker finds these so hard to digest.

A7 or A&? :rofl:
 
Last edited:
I looked up the original build thread for Tanker and see that there are 4x6272 opteron, which support AVX, but noticed Ubuntu 11 was used. Apparently there is an issue with a7 core on the older linux versions. The older A4 core runs fine. If the A7 wu have been failing, the successful completion rate may have fallen below 80% and the QRB will not be received, even for the good A4 wu. All the more reason to check the log for issues...

FahCore_a7 can not start on Ubuntu Server 12.04.5 LTS
https://foldingforum.org/viewtopic.php?f=105&t=29271

If the OS has never been upgraded, consider installing a newer version. I am using Ubuntu 16.04.1 on all my hosts, which keeps me on kernel 4.4.0-xxx and avoids issues the older graphics drivers have with newer kernels (more efficient GPU folding on the older drivers).

Perhaps a moderator at FF can check on the status of some of the returned wu and the status of the QRB qualification. What user name is Tanker folding under?


What troubleshooting have you done ?

check the logs for issues ?

rebooted ?

check cpu, thread and memory usage (top) ?

check temperatures (sensors) ?

If you find that all 64 CPU cores are not being used, try running two or more instances of FAH with less CPUs to get full utilization. 2x32, 2x24+16, 4x16. Whatever works best.
 
Last edited:
I I am using Ubuntu 16.04.1 on all my hosts, which keeps me on kernel 4.4.0-xxx and avoids issues the older graphics drivers have with newer kernels (more efficient GPU folding on the older drivers).

What version of nvidia driver are you finding the best?
 
linux PPA driver 390.25 on 1080Ti (requires newer driver for support). linux RUN driver 370.28 on all the others. A little harder to install the RUN driver but worth extra 5% or more. I can help with the driver and coolbits installation, if needed.
 
Hi Guys,

Sorry for the slow reply, I have been out of town. OK, I looked at the logs. The last slow one I ran was an 8219, run 4, clone 93, gen 95. It took about 10 minutes per frame. It successfully completed it.

I am running Ubuntu 15.10. Tanker came with 12.10 on it and I kept that on until last December when I spent some time trying to upgrade it to a current version. Long story short, the most recent version I could get to run FAH was 15.10. FAH control still does not run.

Right now it is running a 13741, making around 70K points per day. When I look at the CPU usage, probably 60-70% of the cores are pegged at 100% and the rest seem to be picking up and work, running for a while, then loosing work dropping back to zero. Cycling back and forth.

Luckily it is running a relatively fast WU right now. The next time it bogs down I will look and let you guys know what it is doing.

Thanks again for the support!

Junebug
 
If you find that all 64 CPU cores are not being used, try running two or more instances of FAH with less CPUs to get full utilization. 2x32, 2x24+16, 4x16. Whatever works best.

^ I 2nd this option :) Drop down to 2 WU x32 cores. Run this for a couple of WU to find if both groups are working correctly.
If this is a fast WU and you are only pulling ~70% usage, I would look for a problem. I would start with something easy, like splitting my server.
From my CPU folding (Limited Ubuntu knowledge) I would have to call a bad memory chip, connected to 1 of the CPU's.

I first stopped CPU folding because of heat and went to GPU folding. I then have stopped this also, to concentrate on benching.
Benching is a sprint :) Folding is a marathon :thup:
 
I think this is a great idea. Unfortunately, I never could get FAHControl to install no matter what I did after I changed to Ubuntu 15.10. Any good suggestions on how to do this? Or is there an alternate Linux distribution that is friendlier with FAH?
 
I have been using Ubuntu 1604.1 with FAH v7.4.16 for all my hosts. Have not run Ubuntu 15, but FAHControl requires older version of python than ships with Ubuntu 16.
Try installing python-support_1.0.15_all.deb and see if that gets FAHControl running. You are not alone , there are lots of posts at folding forum related to linux fahcontrol and the python issue.
http://launchpadlibrarian.net/109052632/python-support_1.0.15_all.deb
 
Hi Guys,

I have another slow WU. An 8217 cranking along at 2724 ppd. The really strange thing is now all 64 cores are reporting at 100%.

Thanks HayesK, I will try installing the python and see if that will not make FahControl work and then try breaking down the cores into smaller groups.

Thanks again guys!

Junebug
 
Hi Guys,

I have another slow WU. An 8217 cranking along at 2724 ppd. The really strange thing is now all 64 cores are reporting at 100%.

Thanks HayesK, I will try installing the python and see if that will not make FahControl work and then try breaking down the cores into smaller groups.

Thanks again guys!

Junebug

I would not only upgrade to 16.04, add python but also upgrade fahclient to 4.4.15.

To get python 1.0.15:

wget "http://launchpadlibrarian.net/109052632/python-support_1.0.15_all.deb" <-- remove the " " marks
sudo dpkg -i python-support_1.0.15_all.deb
 
Takes me about an hour to setup Ubuntu 16.04.1 with hard disk partition adjustments, software updates, and customization. Faster on SSD, slower on spinner. Could get folding on a SSD in less than 30 minutes, without any updates or customization. I install Ubuntu from a live image on a thumb drive and keep a copy of the files required for setting up FAH on another thumb drive.

My notes for installing Ubuntu 16.04.1, slightly edited to remove my personal info. Not really intended to be a detailed "how to", but more of an outline with useful notes and linux command syntax.

Code:
# linux commands 02-19-2018, edited for forum post 02-26-2018.txt

# misc terminal cmds
top
sensors
sudo nautilus
nvidia-smi
nvidia-settings

# GPU LED brightness, off=0
nvidia-settings --assign GPULogoBrightness=0

# create xorg with coolbits=28 for all gpus
# need to reboot to load xorg changes
sudo nvidia-xconfig -a --enable-all-gpus --cool-bits=28

# use gedit to change host and hostname
# reboot after editing host and hostname
sudo gedit /etc/hosts
sudo gedit /etc/hostname
reboot

# cmd to stop, start or restart fahclient
sudo /etc/init.d/FAHClient stop
sudo /etc/init.d/FAHClient start
sudo /etc/init.d/FAHClient restart

# # Ubuntu install from Live CD or USB drive
# Make fresh install ubuntu 1604.1 to avoid kernel update newer tha 4.04.0-xxx
# manually adj existing partitions on dual boot sys
# edit NTFS partition to make room for ubuntu
# using 15000 MB for XP64 partition
# using 15000 MB for swap partition
# use "-" key to delete existing linux partitions, sda5 ubuntu and sda6 swap 
# use "+" key to add partition sda5 ext4 with mount / and sda6 swap
# use "+" key to add partition sda6 swap
# no updates or extra software
# enable autologin

# boot into Ubuntu desktop
# unlock extra items from launcher sidebar
# add terminal and software updates to launcher sidebar
# edit settings for display, disable lock

# copy F60-folding and deb folders to home
# change property of nv.run to executable
# copy Folding folder, and setup files to home
# make shortcuts to Folding folder and linux commands file

# use sudo nautilus to copy files to system folders
# copy fah config file to computer etc/fahclient
# copy gpu.txt to computer var/lib/fahclient
sudo nautilus

# share Folding folder
# reboot

# use gedit to make grub visible and reduce timeout
# unhide grub boot menu by adding # at start of line GRUB_HIDDEN_TIMEOUT=0
# adjust the 10s default for the grub menu in line GRUB_TIMEOUT=10
# the default progam can be adjusted in line GRUB_DEFAULT=0
sudo gedit /etc/default/grub
sudo update-grub
#reboot

# run software updater
# reboot
# run terminal update, upgrades, autoremove and clean
sudo apt-get update
sudo apt-get install -f
sudo apt-get upgrade --fix-missing
sudo apt-get dist-upgrade --fix-missing
sudo apt-get autoremove
sudo apt-get clean
reboot

# run the software updater again and the terminal updates to comfirm all up to date.
# set software updates: check "never", "download", "every two weeks", "never"  
# change property of usr/bib/update-notifier to not executeable

# install sensors
sudo apt-get install lm-sensors --fix-missing

# optional - detect the motherboard and device sensors
# sudo sensors-detect
# reboot

# install nvidia driver PPA
# sudo add-apt-repository ppa:graphics-drivers/ppa
# update the software list
sudo add-apt-repository ppa:graphics-drivers/ppa
sudo apt-get update

# install nvidia driver from ppa to oldest kernel
# grub startup menu, select advanced, select other kernel(s)
# sudo apt-get install nvidia-304
# sudo apt-get install nvidia-340 (compat with 750Ti)
# sudo apt-get install nvidia-384 (compat with 1080Ti)
# sudo apt-get install nvidia-387
# sudo apt-get install nvidia-390
sudo apt-get install nvidia-390 --fix-missing
reboot

# cmd to remove nvidia ppa driver
# if machine has booted to login screen, 
# Press Ctrl+Alt+F1 to Enter text mode
sudo apt-get purge nvidia*
reboot

# add nvidia settings to launcher sidebar
# view gpu info in terminal
nvidia-smi

# install opencl index detection
sudo apt install ocl-icd-opencl-dev --fix-missing

# create xorg with coolbits for all gpus
# reboot after creating xorg
# use sudo nautilus and browse computer to /etc/X11 to edit or replace xorg
sudo nvidia-xconfig -a --enable-all-gpus --cool-bits=28
reboot


# FAH installation
# confirm config.xml file in etc/fahclient to ensure client starts paused
# confirm gpu.txt in var/lib/fahclient to ensure gpus detected
# install python deb file for FAHControl to work
# install the FAHControl and FAHClient deb files
# enable autostart for fahclient
# add FAHControl to launcher sidebar

# Boinc installation
# install boinc and boinc manager
# add boinc manager to sidebar
# change view to advanced
# attach projects to boinc manager using project URL
# overclockers has teams for Rosetta, Seti and World Community Grid
# Rosetta URL:  http://boinc.bakerlab.org/rosetta
# World Community Grid URL:  www.worldcommunitygrid.org
sudo apt-get install boinc-client boinc-manager --fix-missing


# procedure to install nvidia display driver using RUN file
# prior to installing the driver using run file,
# paste all the cmds below into terminal,
# hit enter, but exit before password
# nvidia driver 370.28 supports 1070, 1060, 980Ti and older 
# nvidia driver 370.28 is not compatible with linux kernels newer than 4.4.0-xxx
sudo service lightdm stop
sudo service lightdm start
sudo sh NVIDIA-Linux-x86_64-370.28.run
sudo sh NVIDIA-Linux-x86_64-370.28.run -K
sudo sh NVIDIA-Linux-x86_64-370.28.run -uninstall

# blacklist nouveau display driver
# use sudo nautilus to copy nouveau blacklist file disable-nouveau.conf to /etc/modprobe.d
sudo nautilus
# open terminal and update all kernels
# reboot. Should have poor resolution desktop if successful.
sudo update-initramfs -u -k all
reboot

# install nvidia.run driver to default kernel
# Press Ctrl+Alt+F1 to Enter text mode
# stop the display manager
# install nvidia.run display driver
# ignore error message about 32 bit lib
# do not need to update xorg file
# reboot
Ctrl+Alt+F1
sudo service lightdm stop
sudo sh NVIDIA-Linux-x86_64-370.28.run
reboot


# install nvidia.run driver to other kernel(s)
# grub startup menu, select advanced, select other kernel(s)
# machine will boot to login screen
# Press Ctrl+Alt+F1 to Enter text mode
# stop the display manager
# recompile nvidia.run driver (-K)
# ignore error message about 32 bit lib
# do not need to update xorg file
# restart the display manager
Ctrl+Alt+F1
sudo service lightdm stop
sudo sh NVIDIA-Linux-x86_64-370.28.run -K
sudo service lightdm start

# un-install nvidia.run driver
# Press Ctrl+Alt+F1 to Enter text mode
# stop the display manager
# uninstall nvidia.run display driver (-uninstall)
# do not need to update xorg file
Ctrl+Alt+F1
sudo service lightdm stop
sudo sh NVIDIA-Linux-x86_64-370.28.run -uninstall
reboot


#
# end of install notes
# misc info below
#


sudo dpkg -i --force-depends fahclient_7.2.9_amd64.deb
sudo dpkg -i --force-depends fahcontrol_7.2.9-1_all.deb

sudo apt-get install linux-image-generic-ck linux-headers-generic-ck

sudo apt-get -f dist-upgrade

sudo apt-get -f install

sudo apt-get install update-manager-core
sudo do-release-upgrade

sudo dpkg --configure -a

sudo apt-get install nvidia-cuda-toolkit
 
Thanks for the guide HayesK! It took me a few days to find some time to follow your instructions. In the mean time, I bounced between units making 1590K per day and 0.7kppd.

So I upgraded to 16.04, installed the python stuff, and installed FAH. FAH is up and running making 80kppd right now. But again no joy on getting FAHControl to work.

I am hoping the fresh install will work better, but I still have no way to reconfigure to smaller CPU blocks. Is there some way I can manually rewrite a config file (like the old days when I was running Knoppix on my server farm). With the new FAH, I would not have a clue how to modify correctly and not fubar it all up.

Regards,

Jeff
 
If you have another system with fahcontrol working, you could make the desired changes, then copy the modified config file to the 4p. Be sure to make a copy of the original config before making changes, then copy back after done.

The config file can be edited with gedit. The safest way to edit the config file would be to make a copy of the existing file to your home folder. Make an extra copy just to be safe. Right click and select open with gedit. Make the desired changes, then use "sudo nautilus" to copy the modified config file to etc/fahclient.

I normally copy my default config file to etc/fahclient using "sudo nautilus". While running sudo nautilus, you can rename or delete the files in the system folders, so be careful. You could even edit the config file in the fahclient folder by right click and open with gedit, but safer to make the changes to a copy, then rename the original and copy in the revised file.

If you need an example config, I could make one with several cpu slots to see what the config file structure looks like with the extra cpu slots.

Really need to get fahcontrol running. What is the issue? Does the fahcontrol deb file install ok or does it fail to connect to the client?

when there were dependency issues with 7.2.9, was able to force the package installation with
sudo dpkg -i --force-depends fahcontrol_7.2.9-1_all.deb
the package would need to be in your home folder and the correct name used with the command.
 
Back