Protein Folding: Your Chance To Help (and Whup AnandTech) Starting Dec 1

Page 3 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
Toggle sidebar Toggle sidebar
Status
Not open for further replies.
Petrusbroder posting. I have set up Anandtechs Annual Folding@Home Holiday Season Races now for 12 years ... yes, it i a dozen years this year. Therefore we are very glad to have a three way race this year.
There is however a problem: In the article above the starting time is set to 3pm December 1, 2016. That time is incorrect.
The correct starting time for the race is 18:00 UTC, December 1, 2016, which corresponds to 1 pm EST.
There are several reasons for this:
1. That has been the starting time since the beginning 12 years ago.
2. 18:00 UTC corresponds to 12:00 CST at which time the best stats-site for folding, http://folding.extremeoverclocking.com/, updates its numbers. My script os prepared for this site and I dont have the time to change it for an other stats-site.
Please pass the word:

The race starts december 1, 2016 at 18:00 UTC, which corresponds to 12:00 pm (=12 noon) CST and 1 pm (13:00) EST.
 
The article above states:
"The Tom’s Hardware and AnandTech community teams are almost always folding as it is. Tom’s Hardware’s Folding team has been around since 2004. But this race will officially kick off on December 1 at 3pm Eastern. It will end on December 7, at 3pm Eastern. And it's open to all of Tom's Hardware readers and community members. Certainly once this event ends, we hope you stick around and continue, joining our fantastic team."

This statement is wrong. The race starts december 1, 2016, 18:00 UTC, which corresponds to 1pm EST.
Please see here: https://forums.anandtech.com/threads/11th-annual-folding-home-holiday-season-3-way-race.2489602/
 
Sorry for the double post! I first did not see if the first had been published.
Thanks for the fast action from your side! Very much appreciated!
 
I plan to join, been waffling for a while, but all I can offer at this point is an Athlon m320 laptop since my main rig needs to be shut down when I'm not using it. will be adding with some light browser folding when studying though :)
 
I just want to put in a plug for GPU folding. Even if you can't run it 24/7, fold for however long you can, on the fastest GPU(s) you've got.

For reference, it seems my GPU (GTX 980 Ti) is estimated at about 20x the PPD of my Sandybridge i7 CPU.
 
...pretty disappointed with the Linux client. Not only does the FAH Control package have a bad dependency for Ubuntu, but it also seems to default GPU folding to 'false'. For anyone else in the same boat, you can install by using dpkg -i and just ignoring the error about python-shared being uninstallable. Then, when you run FAHControl , set PYTHONPATH=/usr/share/pyshare . I also made symlinks from /usr/lib/python2.6 -> /usr/lib/python2.7 and /usr/local/lib/python2.6 -> /usr/local/lib/python2.7 , but I'm not sure if it's really necessary.

...then, you can use FAHControl to edit Settings, and enable GPU. Actually, I had to stop the FAHClient and manually edit /etc/fahclient/config.xml and add the line:
Code: 
  <slot id='1' type='GPU'/>
Then, after starting the client, I finally see the GPU slot in FAHControl, and heard the GPU fans speed up. *sigh* This is really lacking some polish.

BTW, when will I show up on the team page? I've already completed at least one work unit, on my CPU.
 
With the GPU, I'm now getting a lot of "Bad state detected..." errors. I guess that's likely a hardware issue (bad GDDR5 memory?), but I just thought I'd ask if anyone else is seeing them.

I'm using (Linux) Driver Version: 367.57. FAH reports CUDA version is 5.2 and CUDA Driver is 8000.

Reported GPU temps are between 78 - 80 C. Running at stock clocks. Board is EVGA FTW edition (ACX 2.0 - I'm not aware of any issues with this version).

Anyway, keep an eye out for these. I had to find them in the logs. When it reaches "ERROR:Max Retries Reached", the work is restarted and you get no points.
: (

I'll probably try the Windows client, next.
 


I actually am running it on my webserver (light duty CentOS box) and I agree, ran into a ton of problems with the basic installation of the .rpms. So - I went and picked up the tarball version - unzipped it - ran a bit of configuring - and it's running fine. It lacks some of ease of GUI controls but seems to be a bit easier to work with IMHO. I don't have a GPU in the server (obviously no need for a basic webserver) so I'm not sure how well that runs but the CPU part of it is fine.
 


If you need assistance with Linux head over to the team thread and ask there. We have several folders that are familiar with the complexities of F@H on Linux but they might not see your questions in this thread.

http://www.tomshardware.com/forum/id-3229962/tom-hardware-team-thread.html
 
Yes, the CPU folding also worked for me. Without errors, even.

I bailed on Linux and now have the same hardware booted to Win7. My Nvidia driver is 372.90. GPU folding is still encountering errors.
: (

Is this driver version good? Maybe my hardware is just bad.
: (
 


The 37x.xx series of drivers seem to have issues with F@H, what card(s) are you using?
 
Anyone else having trouble getting WU from the server? I completed one WU last night at start up but didn't get a new unit after it. I restarted my PC and it picked up a new WU this morning. Will see if it kept going all day or stopped again after the first WU.
 


I had some problems getting units on my CentOS box - but it seems to still be running just fine. Just takes a few tries then finally connects and gets back to work. My windows boxes aren't having issues from what I can tell, although before the competition started I did a full uninstall/reinstall and didn't change the hardware around once I had the system running.
 
When I had both CPU and GPU folding enabled, I noticed that the CPU took a while to get work.

I wonder if the server is only handing out smaller work units to CPU devices. And maybe those are in shorter supply?

BTW, can anyone recommend a good utility for underclocking my GTX 980 Ti? It has a factory overclock. I want to see if I can reduce the errors by dropping the clock, a bit. Otherwise, I might just drop out altogether.
 


I have no idea as to the CPU WU theory, but I'd used MSI afterburner, unless its an EVGA acrd, then I'd use Precision (Afterburner works with all GPUs, Precision only fully with EVGA cards).
 
So, I rolled back the Windows drivers to 368.69. (Update: I still get errors, but the error rate seems lower. Perhaps it's unrelated. I don't yet have enough run time to say.)

Anyway, I didn't realize how easy it was to roll back the drivers. I only had to go back one step, so I don't know if it'd be difficult to roll back multiple upgrades.

http://www.nvidia.com/object/driver_rollback.html

BTW, does anyone know whether EVGA would accept a return under warranty for a card that gets errors during folding?
 
I think this is an important point. Everyone should run FAHControl and check the "Log" tab. Click the checkbox to see only Errors & Warnings and make sure you're not getting any "Bad state detected..." errors! If so, you could be losing credits (or, worse yet, possibly computing erroneous results).

If you are, try rolling back your driver or reducing any OC.

It appears that AnandTech is leading us in points per day, so we really have to make sure we're getting the most out of our resources. Also, try to recruit some friends, if you can.
 
Going to get a folding rig up with a P4 era athlon dual core and a 650 ti that I can actually leave on until the end of the year sometime this weekend hopefully and that should be able to get some serious crunching done. All my other systems i have to turn off or are too slow to complete the packages so I have just been running dual core CPU packets on my two personal rigs while I work.
 
I found a whole thread about this problem, in EVGA's forums. It's also interesting to see what configs some of their top team members are (or were) using.

http://forums.evga.com/Testing-Issues-with-the-GTX-980-Graphics-Cards-and-Core-21-Projects-m2415012.aspx


Apparently, it's known to be a problem, particularly with Nvidia 9xx series GPUs:

https://folding.stanford.edu/home/issues-with-fahcore21/


One troubling aspect is that it seems the errors are at least partially related to the specific work units. This makes it more difficult to troubleshoot client issues, as two consecutive work units might be from different projects.


I also found that there's a new beta client, but I haven't tried it:

https://folding.stanford.edu/home/new-fah-software-client-v7-4-15-in-open-beta-testing/


No idea whether it'll help the problem above.
 
We're getting beat dammit.

4WnqAQ.jpg


I need a bigger GPU. 🙁
 
Status
Not open for further replies.

TRENDING THREADS

Latest posts

Moderators online

Share this page

COMPANY
RESOURCES
FOLLOW
Top Bottom