Faulty Project 2665 WUs around

[ThunderRd]

I've had several failed SMP WUs this week in the 2665 project:

[16:16:04] Project: 2665 (Run 1, Clone 888, Gen 117)
[16:16:04]
[16:16:05] Assembly optimizations on if available.
[16:16:05] Entering M.D.
[16:16:11] Rejecting checkpoint
[16:16:12] NaN detected: x[7904][0]=4.82564 v[7904][0]=NaN
[16:16:12] utdown: BAD_CORE_FILES
[16:16:12] Finalizing output
[16:18:12] NaN detected: x[12294][0]=-3159739.00000 v[12294][0]=NaN
[16:18:12]
[16:18:12] Folding@home Core Shutdown: BAD_CORE_FILES
[16:18:12] Finalizing output
[16:20:17] CoreStatus = 1 (1)
[16:20:17] Sending work to server
[16:20:17] Project: 2665 (Run 1, Clone 888, Gen 117)
[16:20:17] - Error: Could not get length of results file work/wuresults_03.dat
[16:20:17] - Error: Could not read unit 03 file. Removing from queue.
[16:20:17] Trying to send all finished work units
[16:20:17] + No unsent completed units remaining.
[16:20:17] - Preparing to get new work unit...

These WUs as you see immediately NaN and fail most spectacularly

At first I thought it was a problem with my setup and struggled with it for a while before checking the forum. I should have checked there first, others are reporting it as well...so if you see it it is most likely NOT YOUR EQUIPMENT.

Your client will continue to download the same unit several times. To get around this, delete the work directory, the queue file, and the client.cfg file. When you start it up again, you will have to enter the team number and your username [ThunderRd, of course ] After that, the client should reach out for a new WU. If it doesn't, repeat the process until it does.

[Gizmo]

Oh! Maybe this has been my problem all along. I've been seeing this for about a month now, but not just on the 2665s.

Also, are they still recommending only running the SMP client on quad-core and up machines, or is it safe to run on dual-core machines?

[PorPorMe]

I'm running SMP on both a duel and a quad along with GPU clients and both work fine for me.

[ThunderRd]

@Gizmo
Dual boxes are fine, I run the SMP client on more than 25 duals. The client is designed with quads and multi-CPU rigs in mind, hence the 4 threads, but it works fine on duals also.

Check this page to see if any other WUs are faulty, or to report your problems. This page is monitored by the Pande staff:
Folding Forum • View forum - Issues with a specific WU

[Gizmo]

I just got a 2665 that wouldn't start up; it just hangs

[17:22:31] Loaded queue successfully.
[17:22:31] 
[17:22:31] + Processing work unit
[17:22:31] - Autosending finished units...
[17:22:31] Trying to send all finished work units
[17:22:31] Core required: FahCore_a1.exe
[17:22:31] + No unsent completed units remaining.
[17:22:31] Core found.
[17:22:31] - Autosend completed
[17:22:31] Working on Unit 07 [January 7 17:22:31]
[17:22:31] + Working ...
[17:22:31] - Calling './mpiexec -np 4 -host 127.0.0.1 ./FahCore_a1.exe -dir work/ -suffix 07 -checkpoint 15 -verbose -lifeline 3632 -version 602'

[17:22:31] 
[17:22:31] *------------------------------*
[17:22:31] Folding@Home Gromacs SMP Core
[17:22:31] Version 1.74 (November 27, 2006)
[17:22:31] 
[17:22:31] Preparing to commence simulation
[17:22:31] - Ensuring status. Please wait.
[17:22:32] - Starting from initial work packet
[17:22:32] 
[17:22:32] Project: 2665 (Run 2, Clone 930, Gen 169)
[17:22:32] 
[17:22:32] Assembly optimizations on if available.
[17:22:32] Entering M.D.
[17:22:49] pressed 514.1 percent)
[17:22:49] - Starting from initial work packet
[17:22:49] 
[17:22:49] Project: 2665 (Run 2, Clone 930, Gen 169)
[17:22:49] 
[17:22:49] Entering M.D.

That's all it does. Stopped the client, deleted the work directory and the queue.dat file, started the client, and now I'm working on a 2677.

[ThunderRd]

Yeah, I've seen that happen sometimes but not always on 2665. Your solution is the right one, get another WU. If you continue to get the same R/C/G numbers, you have to delete the client settings, too. Then reconfigure and you'll get a new WU.