• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

WU's keep disappearing!

H

Heliocentrical

2[H]4U
Joined
Mar 28, 2005
Messages
2,852
Something wierd is up with my P3 850. My system has dumped its third WU this week for no apparent reason. It was about halfway (250/500) though a gromac a few days ago when it suddenly reset and downloaded a new one. It worked about a quarter (100/400) of the way through that one then it dumped it too and downloaded a tinker. It's now at 26/400 of the tinker. I recieved no credits for the lost WU's. Has anyone encountered this? It's upsetting to lose these valuable points.

 
Does the fahlog.txt file have this: EARLY_UNIT_END, text in it?

If so, are you running -advmethods or big wu's? Also, is it overclocked at all?

Did it restart the unit it dumped the first time? There's been a little bit of that going around with some of the Gromacs WU's here in the last few weeks.

 
What Phantom said, plus I’d be checking my CPU temps, PSU voltages and for sure I’d be running memtest real soon.

When was the last time you blew all the dust out of your system?

Stop folding for a bit and try Prime95 stress tests.

It is rare to lose a WU on a stable system, even some of the “Problem” gromacs.

 
And one other thing to check, maybe try stopping F@H and deleting the fahcore_##.exe files in the F@H directory, sometimes those can get corrupted and cause units to bomb out.

 
The same thing happened to me...14 minutes ago. Working on a 270-some-odd point GROMACS, and it got about 10% done before it puked. This was on a laptop (Pentium-M 1.4 GHz), so no overclock, it's running just fine and cool. I'm not complaining, since it got a 600-point GROMACS in its place, which will take less time to finish than the 270-point WU would have taken.
 
You guys might consider posting at www.folding-community.org and see if those guys have any sort of clear concise answer as to why this is happening.

I know that Gromacs unlike most WU's don't follow the normal rules of checkpointing. So it only backs itself up at every finished frame instead of whatever value you give it.

However this has been happening with an increasing frequency here lately with no apparent explanation, also along with some random lockups.

You might also want to see if any parts of these units are being sent back in the logs. The MO I've been noticing is that the WU's just stop, and you get a new one, and the old one cannot be verified and no credit is given nor is the data sent back, it just gets deleted.

 
My WU *did* send back its results. And I got 34 points for it! Yeehaw!
 
It's happened to me a couple times also on a stable system, haven't had time to work on it yet but I'll be taking a look this weekend. If anyone finds anything out be sure to post it.

 
those old P3 sometimes have a problem with the -forceasm flag. If you have that, remove the flag and see what happens.
 
p[H]ant0m said:
Does the fahlog.txt file have this: EARLY_UNIT_END, text in it?

If so, are you running -advmethods or big wu's? Also, is it overclocked at all?

Did it restart the unit it dumped the first time? There's been a little bit of that going around with some of the Gromacs WU's here in the last few weeks.

Click to expand...

I have to check on the early_ unit_ end text.
I do have large WU's enabled.
It is not overclocked.
It did not restart the dropped unit, it downloaded a new one. and then dumped that one too.

BillR said:
What Phantom said, plus I’d be checking my CPU temps, PSU voltages and for sure I’d be running memtest real soon.

When was the last time you blew all the dust out of your system?

Stop folding for a bit and try Prime95 stress tests.

It is rare to lose a WU on a stable system, even some of the “Problem” gromacs.

Click to expand...

I clean the inside regularly since the case has easy access. There is no dust. I will run mem test and Prime95 to see what they say. How can you check cpu temps?

magnusvir said:
those old P3 sometimes have a problem with the -forceasm flag. If you have that, remove the flag and see what happens.
Click to expand...

What is the forceasm flag?
 
You can check temps either with a temperature probe, or you can use software such as Motherboard Monitor or Speedfan. I suggest the latter as it does a wonderful job of autoconfiguring the monitors and sensors.

Depending on the vintage of the board it might not have hardware sensors that these programs can read from however.

-forceasm is a flag that runs in conjunction with the client which force SSE boost to be re-enabled when the client is restarted after it's not powered down cleanly.

One thing none of us asked, are you running the graphical or console? If you're running the console client as a service, you have to edit the registry in order to enable this switch.

 
To ensure the quality of your workunits, as well as making sure you
never have an early unit end, or a work unit that is "lost" and never
credited, please take the time to run with the config only flag,
make sure your team number is set to 33, and name= KodiakStar

Thank you for your kind attention to this matter.

Fold On.
 
p[H]ant0m said:
You can check temps either with a temperature probe, or you can use software such as Motherboard Monitor or Speedfan. I suggest the latter as it does a wonderful job of autoconfiguring the monitors and sensors.

Depending on the vintage of the board it might not have hardware sensors that these programs can read from however.

-forceasm is a flag that runs in conjunction with the client which force SSE boost to be re-enabled when the client is restarted after it's not powered down cleanly.

One thing none of us asked, are you running the graphical or console? If you're running the console client as a service, you have to edit the registry in order to enable this switch.

Click to expand...

I am running the graphical version on this one. The pc is a Compaq Proliant ML330 server. I'm not sure what mobo it has or whether or not it has temp sensors. Do I need to this info in order for the temp software to work or does it auto detect?

KodiakStar said:
To ensure the quality of your workunits, as well as making sure you
never have an early unit end, or a work unit that is "lost" and never
credited, please take the time to run with the config only flag,
make sure your team number is set to 33, and name= KodiakStar

Thank you for your kind attention to this matter.

Fold On.
Click to expand...

Wow! I followed these instructions and they seem to have solved my problem.My boxen instantly dumped thier current WU's and dowloaded 600 pointers instead. :p
 
Aratech said:
I am running the graphical version on this one. The pc is a Compaq Proliant ML330 server. I'm not sure what mobo it has or whether or not it has temp sensors. Do need to this info in order for the temp software to work or does it auto detect?
Click to expand...

Speedfan should automatically detect any sensors it can read from. It the case that it's not cut and dried as to which sensor is which, just pause folding and watch which one goes down the most :D

The -forceasm switch doesn't apply to the graphical client persay, although you can (but don't want to) disable the SSE instructions. It's the "Disable highly optimized assembly code" toggle in the Advanced tab.

If the temps, prime and memtest don't provide anything conclusive, try deleting the core files as I suggested earlier, and possibly try shutting off big WU's.


 
I have a borged 2.8 P4 that has download 7 wu's in a week and half. Working on 1, completed 2 and lost 4. Seems to me that there is a problem with somework units getting corrupted or ??? The machine is a daily use unit and runs fine otherwise.
 
I have had a couple of machines lately that have dropped some WU's, also. These machines are rock solid and have primed and memtested just fine. One of them is at stock speeds and the other one is overclocked but has been Primed and memtested a lot. It seems about the only WU's I have had a problem with are 600 pointers. Prior to the past couple of weeks, I had no problem with the 600 pointers except when they first came out and were a bit buggy.

Unfortunately, I don't have ready access to all my machines and so it can be difficult to check which proteins I've had problems with. There's no guarantee I can still find the info in the logfile. From what I can remember, I think they may be some newer 600 pointers. I don't really keep up on when new proteins come out but I do generally recognize ones that I haven't seen before. The WU's that have died on me seem to be ones I don't recognize.

I didn't previously make a big deal about it since I haven't had that many, but there have been more than what is usual. Also, it's irrelevant on whether any other activity is going on with the machine or not. I've had one die on me while I was using the machine doing some light stuff while most times it's usually when the machine is idle other than folding.

 
You must log in or register to reply here.
Back
Top