F@H Client 7.6.9 released with ability to prioritize COVID-19 + bug fixes

C

COCOViper

Gawd
Joined
Dec 16, 2000
Messages
560
https://foldingathome.org/2020/04/1...h-the-option-to-prioritize-covid-19-projects/
Folding@home Client Changelog
=============================
v7.6.8

- Remove bug submission dialog, point to GitHub instead - Use PUT instead of GET to access web session ID. Probably more secure.
v7.6.7

- Make sure Web control gets opened when requested.
v7.6.6

- Fix slot-options command response header
v7.6.5

- Always allow 127.0.0.1 access to web server. - Update GPUs.txt even if GPUs are not currently enabled. - Instead of altering data dir in Windows, confirm removal. - Remove unsupported windows themes on install.
v7.6.4

- Make sure Windows data directory ends with \FAHClient. - Fix Linux service shutdown. - Avoid caching old Web interface. - Fix for Windows service install. - Fix Windows start menu.
v7.6.3

- Organize info blocks
v7.6.1

- Mask newer GLIBC calls for better Linux compatibility. - Add COVID-19 option. - Hardened command server and Web interface security.
v7.5.2

- Updated missing GPU slot messages to prompt user to install the driver. - Updated copyright.
 
CPUs seem to be running, but the GPUs are idle and the client isn't checking back for multiple hours. Not a big issue if I can monitor, but while at work I cannot.
 
You can blame AWS and NVIDIA for the lack of GPU WUs as they are donating GPU compute to F@H. Aws was pulling > 6B points per Day for a while, and NVIDIA was >1B points per day last I looked.

They have added around 20 extra work servers and are occasionally sending out >140,000 WUs an hour. The work server list keeps growing, but apparently it takes a little work after the hardware is available to use them. F@H is honestly doing pretty well with the load since they grew from 90 petaflops to 2.6 exaflops in around 20 days.
 
Gilthanis said:
I recommend just running BOINC along side FAH to keep the rigs busy. Multiple projects now working on COVID-19. https://hardforum.com/threads/combined-covid-19-thread-for-dc-updates.1994228/
Click to expand...

Yeah, my days off are coming up so I'll probably take my machine and move it over to one of the other projects working on COVID and leave the dedicated Folding PC running Stanford. It will have taken approximately 1.5 months to double the number of "points" I had built up with Folding when I restarted running for COVID that originally took years to build. Of course that will barely scratch the actual number of units completed.

My CPU finished one early this AM that was taking almost 8 min per frame, and that's on a 7700 at 4.2 gig.

Code: 
 CPU: Intel(R) Core(TM) i7-7700K CPU @ 4.20GHz
 CPU ID: GenuineIntel Family 6 Model 158 Stepping 9
10:54:53:WU00:FS00:0xa7:Completed 1200000 out of 1250000 steps (96%)
11:02:52:WU00:FS00:0xa7:Completed 1212500 out of 1250000 steps (97%)
11:10:51:WU00:FS00:0xa7:Completed 1225000 out of 1250000 steps (98%)
11:18:51:WU00:FS00:0xa7:Completed 1237500 out of 1250000 steps (99%)
11:26:52:WU00:FS00:0xa7:Folding@home Core Shutdown: FINISHED_UNIT
11:26:53:WU00:FS00:FahCore returned: FINISHED_UNIT (100 = 0x64)
11:26:53:WU00:FS00:Sending unit results: id:00 state:SEND error:NO_ERROR project:14190 run:19 clone:425 gen:3 core:0xa7
11:27:03:WU00:FS00:Upload complete
11:27:03:WU00:FS00:Server responded WORK_ACK (400)
11:27:03:WU00:FS00:Final credit estimate, 28442.00 points

Was kinda shocked looking in the log to see GROMACS still being used.
 
CPUs use Gromacs, GPUs OpenMM (for f@h). Gromacs is still extensively used in research and actively updated, so while the f@h version is a little out of date, Gromacs itself is still cutting edge code.
 
COCOViper said:
Can you remote in to your machine via VNC for a quick pause/fold command? Usually my GPUs are only idle for a few hours at night but otherwise I've been able to keep em cranking for at least the last two weeks : https://folding.extremeoverclocking.com/user_summary.php?s=&u=842201
Click to expand...

I've thought about dropping a VNC client on the main Folding rig, but my time at work would be limited to try and log in and pause/unpause as I pretty much stay busy except for my one break. However it's a fair idea and worth trying. Thanks. :)
 
Got BOINC set up and running Rosetta on the main box with the CPU. GPU still on Stanford Folding as well as the dedicated Folding rig. Thanks to all for the information. 👍
 
The_Heretic said:
Got BOINC set up and running Rosetta on the main box with the CPU. GPU still on Stanford Folding as well as the dedicated Folding rig. Thanks to all for the information. 👍
Click to expand...
Glad to help. Keep in mind that Rosetta work units are heavy on L3 cache. My memory is flaky but I believe each work unit wants like 4MB L3 cache. So, if you don't have enough L3 cache you will lose a lot of potential output as work units wait their turn in line. The same goes for WCG's MIP work units since they use Rosetta's software. It would be best to run a mixture of work to get best overall performance. I know not everyone is worried about points, but wanted to point that out as others move over. I do not know what WCG's Open Pandemics app is going to look like yet. It might be a good mix to throw in with the Rosetta work once it is released.

Edit: Also, if you have any data limits on internet, I recommend going into the Rosetta preferences and change your work unit target CPU runtime to 24 hours - https://boinc.bakerlab.org/rosetta/...7741751&ttok=51520e980c5cfd5841215540344e035f Points per hour should still be the same or pretty close. However, you won't have as many transfers going back and forth but still do the same amount of overall work.
 
You must log in or register to reply here.
Back
Top