Anandtech vs Tom's Hardware Folding@Home Coronavirus Race thread

Page 7 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

Icecold

Golden Member
Nov 15, 2004
1,090
1,008
146
It seems like WU's may be getting harder to come by. I checked on my 2 machines that are folding this morning and both were idle. Generally a few pauses/starts gets a new WU but I did that a ton of times and only 1 of them got one.

It may be a good time to switch at least one of them over to Linux. My main hurdle before was I couldn't get FAHControl to work because of it needing python-gnome2 which isn't part of recent Linux distro's. Is everybody just running FAHControl on another(Windows?) machine to connect to all their Linux machines, or there's an easy workaround I wasn't finding regarding python-gnome2?
 

blckgrffn

Diamond Member
May 1, 2003
9,110
3,028
136
www.teamjuchems.com
Depending on what those units are priced at, they are good units (made by CHW): https://www.anandtech.com/show/12821/the-bitfenix-formula-gold-650w-psu-review/5

Another option would be the Pbhanteks AMP Series and the Thermaltake ToughPower GF1. Finally the Seasonic Focus GX models are priced pretty decently as well

Nice, those are great links! The Bitfenix units are at NewEgg for $75 shipped for Gold 650W Non-Modular and $105 shipped for Gold 850 Modular. I found this list referenced a lot:


Based on that the Bitfenix units seems to be a great value per watt and for relative ranking. They surely aren't pretty, but it appears their value is on the inside of the unit as opposed to the aesthetics.



I know this might be a little off topic, but I feel like I am probably not the only one needing a PSU as an outcome of this race. There are still weeks left and lots of time for hardware to bake :D
 

Markfw

Moderator Emeritus, Elite Member
May 16, 2002
25,478
14,432
136
Nice, those are great links! The Bitfenix units are at NewEgg for $75 shipped for Gold 650W Non-Modular and $105 shipped for Gold 850 Modular. I found this list referenced a lot:


Based on that the Bitfenix units seems to be a great value per watt and for relative ranking. They surely aren't pretty, but it appears their value is on the inside of the unit as opposed to the aesthetics.



I know this might be a little off topic, but I feel like I am probably not the only one needing a PSU as an outcome of this race. There are still weeks left and lots of time for hardware to bake :D
I have been folding with 8=16 cards for the last 16 years, tell me about baking hardware !

I used to use exclusively Corsair AX series PSUs. Now that I have cooked 2 of them(an ax1200i and a ax860i), switching to EVGA (10 year warranty) platinium or better.
 
Last edited:
  • Like
Reactions: blckgrffn

StefanR5R

Elite Member
Dec 10, 2016
5,459
7,715
136
It may be a good time to switch at least one of them over to Linux. My main hurdle before was I couldn't get FAHControl to work because of it needing python-gnome2 which isn't part of recent Linux distro's. Is everybody just running FAHControl on another(Windows?) machine to connect to all their Linux machines, or there's an easy workaround I wasn't finding regarding python-gnome2?
Back in winter 2016--2017 when I started folding, I had Gentoo Linux only both on the folding computers and on my "console" PC. There was no ready-made way to run FAHControl on Gentoo already at that time, and I made only a few lazy, unsuccessful attempts to make it work since. Back then, I configured, ran, and controlled FAHClient entirely in the text console. This was with CPU folding only, on dual-processor computers originally built for physics simulations.

Later, when @biodoc and others helped me to get GPU folding set up on PCs which I had put together for just this purpose, I used Mint 18.3 which supports FAHControl out of the box. I still have Mint 18.3 on these computers and don't plan to replace it before I retire the hardware itself.

Though as a small inconvenience to me, these PCs which besides folding can also run FAHControl, are tied to monitor/mouse/ keyboard on a KVM switch placed elsewhere away from my desk. I could set up remote desktop (which I last used at home several years ago), but didn't feel a pressing need so far.

Occasionally I use a Windows 7 laptop to run FAHControl (or similarly: BoincTasks; which alternatively works on Linux+Wine but is sluggish there). Hmm, now here is a thought: Perhaps FAHControl for Windows could be run on Linux+Wine, when desparate. :-)

Which distribution + version do you use? There should always be a way to get something like python-gnome2 working, but it certainly could get somewhat involved, depending on the distro and on individual background.
 

biodoc

Diamond Member
Dec 29, 2005
6,257
2,238
136
@Icecold , just to add to @StefanR5R comments, FAHControl works on linux mint 19.2 and 19.3.

After installation of linux mint:
sudo apt update
sudo apt upgrade
reboot
go to menu , administration, device manager to install nvidia driver then reboot
sudo apt install ocl-icd-libopencl1
sudo apt install ocl-icd-opencl-dev
Edit: Install Fahclient and FAHConrol
cd /var/lib/fahclient
sudo wget http://fah-web.stanford.edu/file-releases/public/GPUs.txt

Now restart the FAHClient:
sudo service FAHClient restart
 

Soulkeeper

Diamond Member
Nov 23, 2001
6,712
142
106
I use slackware.
I had to create my own package for fah based on the ubuntu/debian ones.
Just like I had to create my own ROCm packages from source.

If any fellow slackware users need help let me know.
 
  • Like
Reactions: TennesseeTony

borandi

Member
Feb 27, 2011
138
117
116
My WU pickup today has been a fair bit worse than the past couple of days. Can get to 40 attempts before work comes through. Really requires babysitting.
 
  • Like
Reactions: Ken g6

Icecold

Golden Member
Nov 15, 2004
1,090
1,008
146
@Icecold , just to add to @StefanR5R comments, FAHControl works on linux mint 19.2 and 19.3.

After installation of linux mint:
sudo apt update
sudo apt upgrade
reboot
go to menu , administration, device manager to install nvidia driver then reboot
sudo apt install ocl-icd-libopencl1
sudo apt install ocl-icd-opencl-dev
Edit: Install Fahclient and FAHConrol
cd /var/lib/fahclient
sudo wget http://fah-web.stanford.edu/file-releases/public/GPUs.txt

Now restart the FAHClient:
sudo service FAHClient restart
Got through all the steps up to installing FAHClient and FAHControl on a fresh install of Linux Mint 19.2 Cinnamon. Installing FAHClient gets the following message/error:


Setting up fahclient (7.5.1) ...
update-rc.d: error: no runlevel symlinks to modify, aborting!
dpkg: error processing package fahclient (--install):
installed fahclient package post-installation script subprocess returned error exit status 1

Any ideas?

EDIT: It seems to work regardless of that error. I haven't been able to get a WU yet, but I was able to get FAHControl to detect my GPU and communicate with FAHClient, show the correct user name, etc.so I think I should be good. I appreciate the help!
 
Last edited:

rise

Diamond Member
Dec 13, 2004
9,116
46
91
Hmmm, I haven't had a WU in a while, like since this morning on 2 machnes anyway
 

Sushisamurai

Member
Jan 21, 2015
47
7
71
I just started this folding@home, wasn't picking up a task initially but I hear my GPU's spinning up. I'm guessing it's working now? I also presume as long as my GPU's are increasing in that % that the %'s are meaning some form of work is being done. I'll leave it on tonight and see what happens. Thanks anandtech, happy to support!
 

Kiska

Golden Member
Apr 4, 2012
1,007
290
136
Plenty of people even here have offered systems and resources to help.
FAH has also got Microsoft and Github working with them to help.
The project doesn't need more network bandwidth, they need more storage bandwidth. LTT's pipe (a) won't do much and (b) pales in comparison to what help they're getting from MS.

Yay LTT workunit server coming online :D
Here is the video:
 

StefanR5R

Elite Member
Dec 10, 2016
5,459
7,715
136
Yay LTT workunit server coming online :D
Here is the video:
Video summary: They bring the pictured 6-GPU computer online as folding client, call on their viewership to fold for their team, and say they are working with the F@h project to deploy a WU server in LTT's office (8:26...9:31 in the video). They'll make another video when that happens. (Or if that happens? The result files then still need to be shipped to F@h's infrastructure. On a truck loaded with harddisks maybe?)

Edit, relevant foldingforum thread:
Do you need help?
Joe_H said:
The people in charge are currently in talks with some large providers, the requirements are steep. A Work Server needs to handle high levels of I/O, high network traffic, have about 50-100 TB of fast storage.
bruce said:
Traditionally, FAH has purchased enough servers to meet the research needs of Grad Students at the various campuses that are part of the Consortium. The whole world is suddenly interested in COVAID-19 and FAH has received unanticipated advertising from several popular websites so suddenly we need 6x of the infrastructure and of the staff. Such rapid growth could not really be anticipated or planned for.

FAH doesn't bug people for money but the information is available if you look for it.
https://foldingathome.org/about/donate/
Unrestricted funds for "infrastructure" are harder to come by than research funds, too.
Joe_H said:
I know they have already made some arrangement with Azure, there are two new servers listed on the server status page in the process of being configured hosted there. As that is publicly visible, I am mentioning it here. As I understand it there are multiple negotiations going on.
sukritsingh said:
Hi all! If you are interested in contributing hardware or server side power to our effort, please reach out to me at sukrit.singh@wustl.edu.
Generally the minimum requirements we have for setting up a work server are:
Debian Linux machines with 8GiB RAM, 100TiB SSD storage, and 1GiB/s network.

If you think you could help out please send me an email! Thanks so much for your support and help!


Server stats is showing three new servers in the azure.com domain now.
 
Last edited:

biodoc

Diamond Member
Dec 29, 2005
6,257
2,238
136
Got through all the steps up to installing FAHClient and FAHControl on a fresh install of Linux Mint 19.2 Cinnamon. Installing FAHClient gets the following message/error:

Setting up fahclient (7.5.1) ...
update-rc.d: error: no runlevel symlinks to modify, aborting!
dpkg: error processing package fahclient (--install):
installed fahclient package post-installation script subprocess returned error exit status 1

According to info in this thread, this only happens when you "uncheck" the box to have the client to start automatically during the installation dialog.
The fix is:
sudo update-rc.d FAHClient defaults
sudo dpkg --configure -a

I usually install FAHClient using the defaults (run automatically during start up). If I want to change it to manual start up then:
sudo service FAHClient stop
sudo systemctl disable FAHClient.service
reboot

Then to start, stop or restart the FAHCient:
sudo /etc/init.d/FAHClient start
sudo /etc/init.d/FAHClient stop
sudo /etc/init.d/FAHClient restart
 
  • Like
Reactions: StefanR5R

StefanR5R

Elite Member
Dec 10, 2016
5,459
7,715
136
Then to start, stop or restart the FAHCient:
sudo /etc/init.d/FAHClient start
sudo /etc/init.d/FAHClient stop
sudo /etc/init.d/FAHClient restart
Or
sudo service FAHClient {start|stop|restart|status}​
In my experience, stop and more so restart can be unreliable. I have gotten cautious and first stop, then check the process table, then start if I want to restart.
 
  • Like
Reactions: biodoc
Nov 18, 2016
42
39
91
Well well well. after two months of quarantine and boredom (hello from italy! :p) a good old fashioned race was just about what i needed! can't play Dirt Rally 24/7 after all, and this time i'm bringing some decent guns. for the last christmas race i was still sporting a 1070, but now i have a heavily overclocked, fully loaded 2080 waiting to go. just lemme sort out FaH again and i'll be in as soon as possible. :cool:

I even thought about slotting in the 1070 again but i don't think my PSU would handle both so, sadly, 2080 alone it is. won't be able to run it 24/7 for safety, can't exactly get new parts these days if something breaks down, but anything counts i guess.
 
Nov 18, 2016
42
39
91
woah, just woah. is turing so incredible at computing? i'm going through my first WU and i'm scoring around 2100k PPD, way above the average for my card. i mean i did read that FaH has been implementing better modeling software but this is beyond my wildest expectations, so much that i'm thinking it's a glitch lol. it's a net fourfold increase in throughput against the 1070... for comparison the 3D rendering performance is just about double and a half more depending on the game.

also, is there a way to be sure i'm working on COVID? there's no selectable options in the 'cause' dropdown, although from what i read in their blog i kinda guess they went full steam on COVID simulations to begin with.
 

borandi

Member
Feb 27, 2011
138
117
116
Here's a list of rough PPD and cost per card. 2080 Ti can get anywhere from 2m-2.5m

There isn't a specific option for COVID, but they are being given the highest priority. So as long as you stick on 'any', then you'll mostly get COVID work. Even if, out of the 5-6 WUs a day, one of them isn't COVID, it means someone else will from the chain.

.
 
  • Like
Reactions: Assimilator1
Nov 18, 2016
42
39
91
yep, i've seen that chart. i guess their 2080 estimate is a bit conservative. they probably didn't test on an overclocked model also, mine is running full steam at 2010 clock and my memory is +1450mhz over base speeds, even faster than a 2080S stock. i remember that core clock was kinda more important than memory speed back in the day, can't feed the beast without memory throughput tho. well, all the better for us. Tom's will be that much more behind in the end. :D
 

StefanR5R

Elite Member
Dec 10, 2016
5,459
7,715
136
Many of the current workunits have higher than usual PPD. With 1080 Ti limited to 180 W (instead of the default 250 W), I am getting 1.4 M PPD typically.

also, is there a way to be sure i'm working on COVID? there's no selectable options in the 'cause' dropdown
There is some info at page 1, posts #18 and #10.
 

Markfw

Moderator Emeritus, Elite Member
May 16, 2002
25,478
14,432
136
Last 24 hr worse for the teams and me.

Team Anandtech 131 m
Toms Hardware 98 m

me 8m . Still having waiting going on.Got really bad overnight, down to 5 m ppd.

On 2080ti ppd, I have 5 cards trying to run, only one has work, and its doing 2.8m
 
Nov 18, 2016
42
39
91
Many of the current workunits have higher than usual PPD. With 1080 Ti limited to 180 W (instead of the default 250 W), I am getting 1.4 M PPD typically.


There is some info at page 1, posts #18 and #10.
thanks!

also, it seems the high numbers were actually a dud, my current WU is hovering around 1800k PPD, which is more in line with a 2080S, like it should. it was too good to be true... :tearsofjoy:

edit: uhm, it also seems that the dreaded "needs large WU" applies though. i'm checking my stats and i'm hovering around 90% utilization with a relatively low power draw, despite the GPU reaching max boost at 2010Mhz given it's well below any thermal or power limit. damn. i've got da powa, use it baby :/
 
Last edited:
  • Haha
Reactions: Assimilator1

sswingle

Diamond Member
Mar 2, 2000
7,183
45
91
Both my systems can't get work. Both still have a WU that won't upload and will expire on the 25th.
 

sswingle

Diamond Member
Mar 2, 2000
7,183
45
91
My logs are all saying No WUs for this configuration. So are the servers up but out of work?