Page 1 of 2
Fahcore_23 still broken on Windows 7 systems
Posted: Wed Nov 29, 2023 7:23 pm
by Gary480six
The Fahcore_23 work units are back out as public work units. And once again, they do not work correctly on Windows 7 Folding systems.
As reported here
viewtopic.php?t=40534
The work units start and run fine. But they get to 100% completed - and just stall. Which means that the next work unit cannot start.
Code: Select all
09:20:15:WU01:FS01:0x23:Completed 2100000 out of 2500000 steps (84%)
09:20:15:WU01:FS01:0x23:Checkpoint completed at step 2100000
09:21:35:WU01:FS01:0x23:Completed 2125000 out of 2500000 steps (85%)
09:22:55:WU01:FS01:0x23:Completed 2150000 out of 2500000 steps (86%)
09:22:56:WU01:FS01:0x23:Checkpoint completed at step 2150000
09:24:16:WU01:FS01:0x23:Completed 2175000 out of 2500000 steps (87%)
09:25:36:WU01:FS01:0x23:Completed 2200000 out of 2500000 steps (88%)
09:25:36:WU01:FS01:0x23:Checkpoint completed at step 2200000
09:26:56:WU01:FS01:0x23:Completed 2225000 out of 2500000 steps (89%)
09:28:16:WU01:FS01:0x23:Completed 2250000 out of 2500000 steps (90%)
09:28:17:WU01:FS01:0x23:Checkpoint completed at step 2250000
09:29:36:WU01:FS01:0x23:Completed 2275000 out of 2500000 steps (91%)
09:30:57:WU01:FS01:0x23:Completed 2300000 out of 2500000 steps (92%)
09:30:57:WU01:FS01:0x23:Checkpoint completed at step 2300000
09:32:17:WU01:FS01:0x23:Completed 2325000 out of 2500000 steps (93%)
09:33:37:WU01:FS01:0x23:Completed 2350000 out of 2500000 steps (94%)
09:33:37:WU01:FS01:0x23:Checkpoint completed at step 2350000
09:34:57:WU01:FS01:0x23:Completed 2375000 out of 2500000 steps (95%)
09:36:17:WU01:FS01:0x23:Completed 2400000 out of 2500000 steps (96%)
09:36:18:WU01:FS01:0x23:Checkpoint completed at step 2400000
09:37:38:WU01:FS01:0x23:Completed 2425000 out of 2500000 steps (97%)
09:38:58:WU01:FS01:0x23:Completed 2450000 out of 2500000 steps (98%)
09:38:58:WU01:FS01:0x23:Checkpoint completed at step 2450000
09:40:18:WU01:FS01:0x23:Completed 2475000 out of 2500000 steps (99%)
09:40:19:WU00:FS01:Connecting to 13.59.134.176:8080
09:40:19:WU00:FS01:Assigned to work server 129.32.209.206
09:40:19:WU00:FS01:Requesting new work unit for slot 01: RUNNING gpu:0:GP104 [GeForce GTX 1070] 6463 from 129.32.209.206
09:40:19:WU00:FS01:Connecting to 129.32.209.206:8080
09:40:20:WU00:FS01:Downloading 22.98MiB
09:40:26:WU00:FS01:Download 49.22%
09:40:32:WU00:FS01:Download 89.46%
09:40:33:WU00:FS01:Download complete
09:40:33:WU00:FS01:Received Unit: id:00 state:DOWNLOAD error:NO_ERROR project:12432 run:7 clone:86 gen:590 core:0x22 unit:0x4e020000560000000700000090300000
09:41:38:WU01:FS01:0x23:Completed 2500000 out of 2500000 steps (100%)
09:41:38:WU01:FS01:0x23:Average performance: 108 ns/day
09:41:39:WU01:FS01:0x23:Checkpoint completed at step 2500000
09:41:43:WU01:FS01:0x23:Saving result file ..\logfile_01.txt
09:41:43:WU01:FS01:0x23:Saving result file checkpointIntegrator.xml
09:41:43:WU01:FS01:0x23:Saving result file checkpointState.xml.bz2
09:41:43:WU01:FS01:0x23:Saving result file positions.xtc
09:41:43:WU01:FS01:0x23:Saving result file science.log
09:41:43:WU01:FS01:0x23:Saving result file xtcAtoms.csv.bz2
09:41:43:WU01:FS01:0x23:Folding@home Core Shutdown: FINISHED_UNIT
11:50:56:FS01:Shutting core down
11:50:56:Clean exit
You'll notice the log says it took TWO HOURS to shut down the core. (and if I recall from back in August, the Core was still seen running in task Manager)
The bottom of the log is how I found the PC the next morning. With the newly downloaded next work unit in READY mode. It had been sitting that way for Hours.
This is what a finishing Fahcore_22 work unit looks like when it's done. The core shuts down in seconds.
Code: Select all
14:58:13:WU00:FS01:0x22:Completed 5000000 out of 5000000 steps (100%)
14:58:13:WU00:FS01:0x22:Average performance: 77.8378 ns/day
14:58:13:WU00:FS01:0x22:Checkpoint completed at step 5000000
14:58:14:WU00:FS01:0x22:Saving result file ..\logfile_01.txt
14:58:14:WU00:FS01:0x22:Saving result file checkpointIntegrator.xml
14:58:14:WU00:FS01:0x22:Saving result file checkpointState.xml
14:58:15:WU00:FS01:0x22:Saving result file positions.xtc
14:58:15:WU00:FS01:0x22:Saving result file science.log
14:58:15:WU00:FS01:0x22:Saving result file xtcAtoms.csv.bz2
14:58:15:WU00:FS01:0x22:Folding@home Core Shutdown: FINISHED_UNIT
14:58:15:WU00:FS01:FahCore returned: FINISHED_UNIT (100 = 0x64)
14:58:15:WU00:FS01:Sending unit results: id:00 state:SEND error:NO_ERROR project:12432 run:7 clone:86 gen:590 core:0x22
If you have a stuck Windows 7 Folding system, the only solution I have found so far, is to shut off Folding, Reboot the computer, and then Delete the Folding work folder that has the finished but stalled work.
That may only apply to Finished and stalled work units. I'm not sure what happens if you do that before the work unit reaches 100%
As was discussed back in August... well, but why support Folding on Windows 7 systems? Windows 7 now only represents about 5% of the worldwide OS landscape. Except that there are about Two Billion computers worldwide. So 5% is still about One Hundred Million potential Folding systems. Even if it's really only Half that.. I think it's worth trying to find out why the new Fahcore_23 is not working correctly on Windows 7 systems.
And it does seem odd that the Fahcore_23 is broken on Windows 7. All my other Windows 7 era software migrated just fine to my Windows 10 PCs.
Re: Fahcore_23 still broken on Windows 7 systems
Posted: Wed Nov 29, 2023 8:33 pm
by Joe_H
The answer is not going to be any different, F@h is no longer supporting Windows 7 and has not for years. There will be issues with developing the code and having the appropriate DLLs and other supporting software such as drivers. I would be extremely surprised if they even have a Windows 7 system to test any new builds of software.
There was 5% usage of Windows 7 back in February, I doubt if it has stayed that high over 9 months later. Even if there still are about a "One Hundred Million" or "Fifty Million" computers running Windows 7, most of those are probably in industrial or similar settings controlling some machinery and there is no suitable software update to run on a later version of Windows. Or in other words, they are not candidate for F@h.
More and more projects will be releasing with Core_23 as a requirement. Newer projects need features in the OpenMM code base that were not available when Core_22 was created. Eventually the projects using Core_22 will finish, and it will drop out of use. Options are to upgrade to a later version of Windows, switch to Linux, or stop folding on that machine.
Re: Fahcore_23 still broken on Windows 7 systems
Posted: Wed Nov 29, 2023 10:10 pm
by calxalot
The client software v7 and v8 is built to support Win 7.
If core 23 has a compatibility problem, the assignment servers should not assign work requiring core 23.
Re: Fahcore_23 still broken on Windows 7 systems
Posted: Wed Nov 29, 2023 10:13 pm
by calxalot
Is it certain that this wasn’t a Tuesday windows update restart?
Re: Fahcore_23 still broken on Windows 7 systems
Posted: Wed Nov 29, 2023 10:19 pm
by Joe_H
calxalot wrote: ↑Wed Nov 29, 2023 10:10 pm
The client software v7 and v8 is built to support Win 7.
If core 23 has a compatibility problem, the assignment servers should not assign work requiring core 23.
The last I knew servers do not check OS versions, just whether it is Windows, Linux, or macOS.
Is it certain that this wasn’t a Tuesday windows update restart?
Same problem was reported by another user back in August. Core_23 would process from 0-100% and then hang in the post-processing and never package the WU up for return. Common issue was use of Windows 7, and there have been no new updates for that in nearly a year.
Re: Fahcore_23 still broken on Windows 7 systems
Posted: Wed Nov 29, 2023 11:09 pm
by calxalot
Emailed Joseph.
Re: Fahcore_23 still broken on Windows 7 systems
Posted: Thu Nov 30, 2023 2:24 am
by Lazvon
Windows 7?!? Isn’t that dangerous to be running?
Re: Fahcore_23 still broken on Windows 7 systems
Posted: Thu Nov 30, 2023 7:42 am
by Librarian
Might be a silly question, but is there any flag in the config we can prevent the client from utilizing core23?
I just hate my rigs just burning electricity going nowhere and they churn out core22 projects 24/7 without issues.
Re: Fahcore_23 still broken on Windows 7 systems
Posted: Fri Dec 01, 2023 12:36 am
by calxalot
There are no flags I’m aware of.
The developers have been alerted.
For now, I think your workarounds are disable GPU slots, upgrade Windows, or use Linux for folding.
Re: Fahcore_23 still broken on Windows 7 systems
Posted: Fri Dec 01, 2023 8:47 am
by ElectricVehicle
When Fahcore_23 hangs on Windows 7, the process is just hung ...forever... using 0% CPU, but not finishing or exiting so it disables your folding computer until you manually dump the WU. I just had one that was hung for a day until I had to intervene. It is a waste of resources to try and fold Fahcore_23 on Windows 7. I am displeased at being forced to change the OS on that computer for this issue and not on my own time frame.
So the solution for me is to spend some time to update the computer to Ubuntu and run that. Windows is not an option for me as with Windows 11 and on they FORCE you to link to a microsoft account and Windows 11 also requires TPM hardware. I'm not interested in spending the money for a new motherboard, which will force me to get a new processor, RAM and heatsink and likely the power supply. I can save that money for other purposes, including an upgraded GPU, and upgrade the motherboard etc. at a much later date. I have no desire to use Windows 10 and have to fool it into letting me bypass the microsoft account, only to face the issue of going to Windows 11 or whatever in some years to come.
So I'm annoyed, but I'll solve it by moving to Ubuntu. One upside of moving to Linux / Ubuntu is that the PPD for the computer / GPU likely will significantly increase, possibly by 10 to 25% - hard to be exact as it varies with WUs and aggregate data like at
LARS can be difficult to apply as small sample sizes, host HW and network connections produce results that may not be directly comparable.
The era of using Windows 7 for GPU folding is coming to a close.....
Re: Fahcore_23 still broken on Windows 7 systems
Posted: Fri Dec 01, 2023 9:05 am
by ElectricVehicle
Interesting thing about this thread - looks like many of us have been folding back when there was only CPU and PlayStation folding! Then SMP CPU and GPU folding emerged, CUDA support...
Re: Fahcore_23 still broken on Windows 7 systems
Posted: Fri Dec 01, 2023 11:45 am
by wdanwatts
My Fedora 39 has the same problem with my GPU attempting to run core 23.
Re: Fahcore_23 still broken on Windows 7 systems
Posted: Fri Dec 01, 2023 3:17 pm
by Joe_H
wdanwatts wrote: ↑Fri Dec 01, 2023 11:45 am
My Fedora 39 has the same problem with my GPU attempting to run core 23.
No, you do not have the same problem. Under Windows 7 processing of the WU appears to run to completion with Core_23, it the does not go through the end sequence where the finished WU is packed up and returned. What you reported is the WU processing never starting and that is from Core_23 not finding a library that is there on your Fedora 39 system.
Re: Fahcore_23 still broken on Windows 7 systems
Posted: Mon Dec 04, 2023 10:26 am
by ElectricVehicle
Solved my Windows 7 issues by flipping my folding host over to Ubuntu 22.04.03 .
Booting the PC from USB to install Ubuntu from a USB drive was a little tricky - I set the USB emulation to harddisk instead of auto, reset the computer - reset - not power cycle, then the USB drive appeared as a boot selection in the boot order, and I selected that as the first boot priority, reset again and GRUB appears for a moment, I think you press enter to that which is easy to do without even thinking then the whole screen changes to just a blinking cursor from many minutes. When it takes over two minutes I figure something is hung and reset or try something else. Don't! It just takes forever, less than 15 minutes I think of just solitary, lonely blinking cursor..... And then Ubuntu install dialogs - Yay!
What ultimately worked is to use the Ubuntu Ubuntu Software app to install Folding@Home.
Ubuntu Software > Category > Science > Folding@Home
Folded the first Fahcore_23 WU successfully on this host! I expect my points have gone up 20+% but too early to know for sure.
I setup remote access to the host to monitor using FAHControl on another host. Add the IP and password for the monitoring host. Then restart FAHClient. I ended up pausing FAHClient in FAHControl, then killing the FAH process, and finally restarting the host to get FAHClient running again. Some time later I'll find a more elegant way...
The FAHClient stop / start procedures in this thread did not work for me -
service FAHClient start
or
/etc/init.d/FAHClient start
Remote computer connection
This is some of the thrash I went through before succeeding:
Using the official FAH Linux and instructions did not work for me. FAHControl failed to install, apparently it requires Python 2 which is obsolete and has been removed from Ubuntu 22.04.03 .
@cdberkstresser has taken the open source code for FAHControl and created an updated version for
Testing Builds for Python 3 Pre-release . This fixes the Python 2 issue but has a new issue with "MutableMapping".
This thread covers the "MutableMapping" issue and more
Ubuntu 22.04 LTS FAHcontrol installation .
Re: Fahcore_23 still broken on Windows 7 systems
Posted: Wed Dec 13, 2023 1:38 am
by smurfcorpse
I have a Windows 7 PC that I just can't upgrade at this point. It's worse now that Micro$oft has dropped the free OS updates. Updating to Ubuntu is a huge undertaking and even if I did, it sounds like there are still issues with Ubuntu and Core23 anyway. I would like to see this core updated/fixed for Windows 7. The client and cores are still 32 bit so I'm not sure what the actual issue is.I know we like to quote % of users = the support given but that % still translate to users and points. At the very least can we stop core23 from being deployed to Windows 7 PCs so they can continue to fold the available WUs. I now have a GTX1080 just sitting idle because it keeps downloading and running Core23.