koala

My system crashed (not caused by F@H) at 92% while running the SMP multi-core client. When I rebooted it started again from 0% but the deadline remained the same.

I usually finish SMP work units in about 50 hours, well within the 'preferred' deadline limit, but now it's due to finish 16 hours past the 'final' deadline.

Does this mean I won't get any points for the work unit and they won't be able to use the work done?

Snoopdogie187

Well you can try to see if it will finish before. I had my system crash a total of 3 times now and always had the time to make it through all.

I a not sure if any of us turned in a late WU yet or not though and i tend to think you will still get some points but i am not sure if hey can still use it.

kodi

You should get points for unfinished WU's
Just check John the Pilot and my results

Snoopdogie187

I do not think its the same situation, i believe the WU still goes even if its already due as long as it has not been closed, and reopened. And what Koala had happen doesnt send in the unfinished results, just restarts it. If you read the log file I think it mentions bad checksum or something simular to that.

mattlock

I can tell you first hand that you will not get any points for the late WU with the SMP Client.

I noticed today that I had only posted 1 WU per day for a few days (My AM2 5600+ turns a WU in about 28 hrs). Sure enough, I checked the other rig and found that it had been unable to connect to the server to upload the results. (it was however able to download new WUs) So I had 2 completed WUs sitting in the queue waiting to be uploaded and over and 57% complete on a 3rd. (S939 4400+ turns a WU in 36-44hrs depending what the kid is doing)

I stopped folding, rebooted my VM and started up F@H. The 1st thing I seen in the new log was:
WU 2 has past the deadline.
Unable to delete all of WU 2
Sending WU 3

It wouldn't even upload the file because it had past the deadline.

On another note, I've noticed that other folding members have completely lost thier work due to system crash when using the Windows client. I've had VMware lockup on me a few on different rigs and had to kill the program and processes manually and reboot to get it back.

The 4400+ was having an overheating issue just after moving it to a new case for the kid (case was fan was installed backwards from the factory) causing lockups. It was fine when it was just folding but when she started editing video and such it froze or shut down.

Anyway, getting to my point, with the Linux client I've never lost any work due to a crash, lockup, unexpected reboot or shutdown. Maybe another advantage to running the VMware/Ubuntu setup.

Sorry about the novel. It's late, I'm tired, and rambling on.

BTW, my new E8400 is up and running, OCed to 4ghz. As soon as the torture test is done I'll be rolling the dice on the Windows SMP Client. Just so I can have a little insight when helping others with thier issues.

Snoopdogie187

Koala when you say crash do you mean the BSOD?

And i get the blue screen a lot randomly, and not all the time does it cause F@H to restart, i think it just has to happen while F@H is saving or something like that.

I do wonder how long it takes me to do a WU but i never give this laptop a chance to do it all at once as I use it too much for school and games

koala

Thanks for the replies, everyone.

It was a media player BSOD caused by testing overclocked RAM/CPU while multitasking. My own fault really for not shutting down the client first.

I still have 22 hours to go before this WU reaches 100%. Hopefully the next one will give me the full amount of time to complete it.

The log says it was a bad checksum, but I have the checkpoint set to 3 minutes, so I thought the worst that could happen in a crash is that I'd lose the last 3 minutes of work, not the whole thing.

[05:27:51] - Checksums don't match (work/wudata_06CP.arc)
[05:27:51] - Attempting to revert to backup...
[05:27:51] - No backup
[05:27:51] - Could not revert
[05:27:52] - Failed to delete work/wudata_06- FaiWarning: check for stray files
[05:27:52] - Starting from initial work packet

kodi, I'm not sure but I think it's different for the SMP client. It's still in beta so the deadlines are much shorter and more strict than for the standard clients while they're testing, so I suspect I won't get any points for late results.

Snoop, next time you get a blue screen, make a note of the Stop code. It will help you find the cause of the crash.

Matt, it looks like you have worse luck than me with computers

Glaswegian

Interesting. My current WU is due by 4.00pm today - but it's only reached 93% (not sure why) so I'll be keeping a close eye on my stats and see what happens.

mattlock

WU Deadlines
All work units are time sensitive and have deadlines. There are 2 deadlines attached to most WUs:

Preferred deadline
Final deadline

A few projects have no final deadline, aka deadlineless or timeless.


When a client does not return a completed WU within its Preferred deadline, the WU is assumed lost. The WU will be reissued to another client to make sure the results will be returned. Every WU is important to each project even though a WU is only a small part of a single project, and part of the Folding@home Project.


A client that exceeds the Preferred deadline of a WU still has time to complete the WU before its Final deadline. Full credit will be awarded when a WU is returned before the Final deadline. The scientific value of the work is slightly reduced because it had to be duplicated when the client passed the Preferred deadline. If the client were to exceed the Final deadline of a WU, the client will discard the partially completed WU and download a new WU. No credit will be awarded.

Copied from F@Hwiki

So it seems that it will not continue folding the WU once it passes the final deadline. If it is still working once the final deadline has past, stop folding and then start folding again. It should see that the deadline has past, delete the WU, and get a new.

Keep us posted Koala.

@ Glas, what time is there now? it's 10:30 am here in the states. If running 24/7 your SPU should complete a WU in between 24 & 29hrs (if I remember right) or 1% every 17-18 minutes at the most. So by that estimation you would need just over 2hr to complete the WU. If you OC you can get you done in around 19hrs. This info is taken from Oddballs folding results with the same CPU.

koala

That seems fair enough. My final deadline is now in 2 hours, and I'm at 62% with 21 hours remaining. I'll shut down the client after the deadline has passed to see if it downloads a new WU or continues with this redundant one.

I wish! At 95% CPU usage on a 4200+, mine takes 30-35 minutes, depending on what other CPU-intensive apps are running at the same time.

Glaswegian

Matt it's 4.00pm.

I don't run the PC 24/7 - but it's on for about 9-10 hours per day, and more at weekends.

The message from F@H is

Unit's deadline has passed. Going to interrupt core and move on to next unit...

95% completed as well...

I'm about the same as koala - takes about 35 minutes to complete 1%, although looking at PC Wizard, the loads on each core are

Core 1 76%
Core 2 63%

The highest I've seen Core 1 is around 88%. Should I be making some adjustments to my config?

I've been meaning to have a go at OC, but it's not something I've done before so I'm still reading up on how best to do this.

koala

Did you get this message without having to do anything like restarting the client?

In my Task Manager under the Performance tab, the graphs show both cores running at max, and under the Processes tab there are 4 FahCore_a1.exe files running at 34%, 24%, 23%, 14% (total 95%). These values fluctuate and are usually around the 25% mark each. I'd say that you need to tweak the config file to get higher CPU %.

mattlock

Oddball and myself are both running the VMware/Ubuntu setup, so I'm not exactly sure what you guys should be seeing on the Windows client. I do know that the E6750 out performs the 4200+ on the SMP client by a landslide. In fact the E6600 with no OC beats out the 6000+. It seems the Gromac core which the SMP runs is handled better by Intel chips but the Tinker and Amber are owned by AMD.

Glas how do you have your CPU usage set? The SMP will only use about 90% of the total CPU (both cores combined) I agree you need to adjust your configuration.

Glaswegian

Yes - it appeared in the DOS window showing F@H.

A new WU has downloaded and started - looking at Task Manager, I have 4 FahCore_a1 running, 2 around 15% and the other 2 around 10%. In my config file, I set cpu at 50% - I understood (although that may now be mistaken) that 50% was correct for a dual core (and 25% would be applicable for a quad core).

Think I'll do some config editing now...

mattlock

THe client devides the work into 4 units whether dual core or quad. I have it running in VMware at 100% and windows task manager shows my CPU Usage at 90-95%

Glaswegian

Thanks Matt - I'm adjusting the cpu usage bit by bit just now - I'm letting things settle each time before adjusting again and so on. Hopefully this will get me back on track.

koala

I've had a new X38 motherboard, E6850 CPU and 8800GT sitting here for nearly a month now waiting to be installed. If Intels are quicker at F@H than AMDs then that's another incentive to pull my finger out and do the upgrade

Glaswegian

Absolutely!

Last 1% took 20 minutes - much better.

koala

Everything's back to normal now. I shut down the client a few minutes after the final deadline had passed, restarted it and it's now working on the next unit.

I just hope I don't lose any more if I do the upgrade mid-WU. Does anyone know if it will carry on with the same unit after switching mobo and CPU or will it have to download a new unit?


[17:35:25] Unit 6's deadline (January 26 17:34) has passed.
[17:37:45] - Preparing to get new work unit...
[17:37:45] + Attempting to get work packet
[17:37:45] - Connecting to assignment server
[17:37:47] - Successful: assigned to (***.***.***.***)
[17:37:47] + News From Folding@Home: Welcome to Folding@Home
[17:37:47] Loaded queue successfully.
[17:37:50] + Could not connect to Work Server
[17:37:50] - Error: Attempt #1 to get work failed, and no other work to do.
Waiting before retry.
[17:38:06] + Attempting to get work packet
[17:38:06] - Connecting to assignment server
[17:38:08] - Successful: assigned to (***.***.***.***)
[17:38:08] + News From Folding@Home: Welcome to Folding@Home
[17:38:08] Loaded queue successfully.
[17:38:49] + Closed connections
[17:38:49]
[17:38:49] + Processing work unit
[17:38:49] Core required: FahCore_a1.exe
[17:38:49] Core found.
[17:38:49] Working on Unit 07 [January 26 17:38:49]

koala

Snoop, I've moved your BSOD problem to a new thread in Vista: http://www.techsupportforum.com/micr...computing.html