SPCR Folds Team Blog

aristide1 · Post by **aristide1** » Sun Apr 06, 2008 8:45 am

As if these strange WUs were not enough of an annoyance now I finish WUs around 10PM and I don't see credit for them until 6AM the next morning.

aristide1 · Post by **aristide1** » Sun Apr 06, 2008 6:04 pm

And now even more crap. I recently shut down twice shortly after a WU completed. Once about a week ago to catch up on long overdue updates, and today to install a DVD writer. In both instances the current WU promptly error-ed and died, and started a fresh one.

aristide1 · Post by **aristide1** » Mon Apr 07, 2008 5:22 pm

More nonsense.

I finished a WU last night after 9PM, I don't see any credit for it.

Dutchmm · Post by **Dutchmm** » Tue Apr 08, 2008 5:48 am

There is a marked delay in seeing credits on the extreme overclockers site. You finished after 9pm, so the next stats in which you would be included would be at 12pm - assuming you are some 0 or multiple of 3 hours different from PST. At the moment, the Extreme site is working one hour after the Folding list, from the 3 hour updates. So, following this example, you should see the stats at extreme only at 1am. But, if you are one or 4, 7, etc hours different, then it could be that there is a six hour gap. Can't do the math, fumles with shoe laces. It is often 6 hours before I see the credit on Extreme ....

OTOH, if you meant the folding site, I don't know. They seemed to have put the latest missing credits incident behind them.

aristide1 · Post by **aristide1** » Tue Apr 08, 2008 5:28 pm

I only use Extreme.

Also - Talk about taking a hit a lot of the WU's are fewer points and they take longer as well. Oddly though, my one folder is pulling less out of the walk sock with those, ummm lets get it right, tethered visicles.

xan_user · Post by **xan_user** » Tue Apr 08, 2008 7:51 pm

Quote from EXTREME Stats here; http://folding.extremeoverclocking.com/

..."On a different note, I've gotten some complaints from people about a long delay in their points showing up in the stats. This is not a problem on my end, this has something to do with Stanford's back-end. I can only process the numbers which are provided in the files they generate for us. I emailed Vijay this morning and will let you know what he says as soon as I get a reply.

UPDATE: I got a reply from Vijay, "There was an unusual situation this weekend, where 2 servers weren't being credited over the last few days." The issue should be resolved, I'm not sure how long it will take for the points to all show up (if they haven't already)."...

aristide1 · Post by **aristide1** » Tue Apr 08, 2008 7:55 pm

The only unusual problems over at Stanford are the ones that happen daily. Otherwise they should be classified BAU Business As Usual.

aristide1 · Post by **aristide1** » Wed Apr 09, 2008 4:15 am

Oh brother, my C2D finishes a WU every 24 hours, now the stats say I didn't do anything yesterday.

Total b/s

aristide1 · Post by **aristide1** » Thu Apr 10, 2008 4:58 am

I would have to say I've been short changed over 10,000 points at this time.

I may just delete FAH for that pc, reinstall, and hope I end up on a server that still works correctly.

aristide1 · Post by **aristide1** » Thu Apr 10, 2008 8:06 pm

And once again, having finished a WU at 10:00 PM, 27 hours ago, I have no completed WU's since the one prior to that one.

Yeah, right.

aristide1 · Post by **aristide1** » Sat Apr 12, 2008 7:38 pm

Well, I checked my log and found several abends, so I reran Prime 95, and I had to up the voltage 1 tick. After 24 hours I decided I wanted to try to end up on a new server as well, so I deleted everything and reinstalled ver 5.91.

Dang, ended up on the same server.

FYI - 5.92 beta version is out, install process had changed, more security options to set.

KansaKilla · Post by **KansaKilla** » Sun Apr 13, 2008 4:48 pm

when will they come out with the unified client? i'm hoping that will make it a little easier.

Post by **NeilBlanchard** » Sun Apr 13, 2008 7:33 pm

Hello,

Well, I lost a SMP unit at 90% today, because I had to stop the client to change over to SLi -- it got corrupted and would not restart...

pixel_pimp · Post by **pixel_pimp** » Wed Apr 16, 2008 1:21 am

Hello top 100

Wibla · Post by **Wibla** » Wed Apr 16, 2008 2:24 am

Im trying to get F@H SMP running on my two E6600 boxen, but its proving difficult, and im not bothering with it atm..

I have a Xeon X3050 (E6400 ?) in colo with 32bit linux, but as the linux SMP client is 64bit only that wont bring me any joy...

Also finished building the HTPC rig with an E2140 @ 2.2GHz, i'm gonna get that one up and running with SMP F@H in linux when im not home

Wibla · Post by **Wibla** » Wed Apr 16, 2008 6:56 am

And there I have an SMP box running in linux, found out that our NYC hosted box is running 64bit SMP kernel, on an AMD X2 4200+

VanWaGuy · Post by **VanWaGuy** » Wed Apr 16, 2008 12:40 pm

Wibla,

would it cause you any trouble to change to 64-bit Linux?

Wibla · Post by **Wibla** » Fri Apr 18, 2008 6:38 am

Im not sure, but as my .nl box is running mission critical systems, im not gonna mess with it, I cant afford downtime..

First SMP F@H unit delivered from the NY box, woot.

Edit:I got both the Workstation at home and the one at work/school up and running with SMP again, seems the SMP client I had on those were out of date, that means another two E6600 cpus at work for SPCR!

aristide1 · Post by **aristide1** » Fri Apr 18, 2008 7:24 pm

My X2 boxes are for crap.

My Black Edition goes blank screen now. Stays on but responds to nothing, must be rebooted. Its running Ubuntu. I think I'll lower it from 3.2 to 3.0 and try to lower the voltage as well.

My 4200+ at 2.95 GHz went W2K BSOD twice now, so I dropped the FSB 5% and am returning to Prime 95 * 2. Now at least the memory is no longer OC'd.

There's a reason AMD never built a 3.0 or faster 65nm X2.

Better slower than not running.

Odds of adding a 4th rig? Pretty low right now.

spookmineer · Post by **spookmineer** » Fri Apr 18, 2008 8:01 pm

Although it hasn't happened a lot here, I now have lost an SMP work unit which was over 86%. Had to stop it because of a clan match, restarted and it was lost.

...I have reset the PC clock to January a few times because I'm still using the old Beta. Maybe it's time to finally install the newer one, or maybe it won't make a difference and these things happen every once in a while.

Wibla · Post by **Wibla** » Sat Apr 19, 2008 5:45 am

The newest beta with the new MPI is supposed to be alot more stable, but time will show.. im running the regular one for now.

There's a reason the betas expire, so just upgrade it

AZBrandon · Post by **AZBrandon** » Sat Apr 19, 2008 8:45 am

So after completing my first successful WU in years, my 2nd WU failed at 27% last night..

[12:49:53] Completed 1300000 out of 5000000 steps (26 percent)
[13:12:42] Writing local files
[13:12:42] Completed 1350000 out of 5000000 steps (27 percent)
[13:30:56] Warning: long 1-4 interactions
[13:30:56] Gromacs cannot continue further.
[13:30:56] Going to send back what have done.
[13:30:56] logfile size: 43476
[13:30:56] - Writing 44012 bytes of core data to disk...
[13:30:56] ... Done.
[13:30:56] - Failed to delete work/wudata_02.xtc
[13:30:56] - Failed to delete work/wudata_02.chk
[13:30:56] - Failed to delete work/wudata_02.xvg
[13:30:56] Warning: check for stray files
[13:30:56]
[13:30:56] Folding@home Core Shutdown: EARLY_UNIT_END
[13:30:56]
[13:30:56] Folding@home Core Shutdown: EARLY_UNIT_END
[13:30:59] CoreStatus = 7B (123)
[13:30:59] Client-core communications error: ERROR 0x7b
[13:30:59] Deleting current work unit & continuing...

I raised my Vcore from 1.25 to 1.275 in hopes maybe it will be more stable with more voltage, although the first WU ran for some 18 hours or so and completed without issue at 1.25v. So am I even correct that the above error is an indication that the CPU was unstable and it halted for that reason, or could it be something else, like SMP instability? This is on Windows XP 32-bit with an Opteron 185 at stock clockspeed, stock DRAM, just undervolted from 1.35 to 1.275v.

aristide1 · Post by **aristide1** » Sat Apr 19, 2008 4:45 pm

Folding machines need to be stable, ie 24 hours of Prime 95. Rounding errors can occur in folding that will not abend but make the results not legitimate.

spookmineer · Post by **spookmineer** » Sat Apr 19, 2008 9:21 pm

Upgraded to 5.91 and added the zip install of 5.92.

Ran into problems, not being able to receive new work: Address 0,0,0,0.

I have a router since last week and thought there might be a problem with it.
Opened up port 8080 and now FaH works fine.

Weird though... with the older version I didn't have to manually open up port 8080 in order to make it work. Maybe this has something to do with the new DeinoMPI?
Hope it will be stable, I lost a few units here and there.

aristide1 · Post by **aristide1** » Sun Apr 20, 2008 4:12 pm

My install of DeinoMPI went awry when I didn't accept the default directory.

And it's another zero point day for me. The X2 Black Edition went AWOL again. It got a voltage bump, but this is it, one more stop into DUH-land and it's going to see a major drop in speed.

The C2D is working on some friggin' WU with 25 minute checkpoints, meaning it won't finish today. The other X2 is also working on some other WU taking a full 35 minutes per checkpoint, IE over 2 days to complete a WU.

I wish the blasted servers had a better sense of what each processor can do.

aristide1 · Post by **aristide1** » Mon Apr 21, 2008 6:51 pm

Adios X2 at 3.2 GHz!

I upped the voltage to 1.45 volts and it fell on its face again. So Iganu style I dropped it to 2.8GHz (even multiples to keep memory at the proper speed) and 1.3 volts. And there's never been any signs of heat!

I'm like the only person on NewEgg that does not love his Black Edition.

I went over to the AMD web site to see what voltage they list for this and they spec it as a 90nm part when its a 65nm part.

Idiots.

AZBrandon · Post by **AZBrandon** » Mon Apr 21, 2008 7:08 pm

Does memory speed really matter much for folding? I thought it only came into play when you had a WU that well exceeded the available L2 cache, which shouldn't be an issue on your processor, should it?

Wibla · Post by **Wibla** » Tue Apr 22, 2008 2:17 am

There's usually increased overhead and a slight performance hit when you do not have 1:1 FSB:RAM, but this isnt much of an issue with AMD rigs, because of the integrated ram controller.

Faster ram is always better, but the performance difference is small at best.

On a sidenote, my production figures in F@H are really rising, and im still one SMP box short

aristide1 · Post by **aristide1** » Tue Apr 22, 2008 2:55 pm

Wibla wrote:There's usually increased overhead and a slight performance hit when you do not have 1:1 FSB:RAM, but this isnt much of an issue with AMD rigs, because of the integrated ram controller.

Faster ram is always better, but the performance difference is small at best.

Or even multiples like 1:2, AMD FSB 200 and DDR2-800. But if you look at the graphs on Tom Harware the differences are small and noted when Intel started releasing higher FSB/lower mutliplier processors with almost no performance increase. But it is a good trick to limit OCing by offering a much lower multiplier.

OT - Oh man, almost 3 days for 2144 points. Big deal.

Wibla · Post by **Wibla** » Wed Apr 23, 2008 4:39 am

Haha, yeah, I hate those WU's, they go on and on and on....

I have one 3064 unit at 54% in WS2, running at 80% cpu and 31min/1%, else the damn thing near overheats,
a 2605 at 88% in my colo box with 26min/1% (X2 4200+),
and a 3062 at 50% in my home WS, with about 20min/1% (running at 2.4GHz, 100% cpu)
+ three other single core folders, a Xeon X3050 in 32bit linux and my home fileserver...

... Im eager to get my HTPC / E2140 @ <insert overclock here> running full time in linux, it should be a tad faster than me workstation...

There will be some massive upgrades this summer;

E2140 HTPC -> E8500
E6600 WS -> Q9450
New fileserver with E8400 or E7200

... That should give me a nice boost

SPCR Folds Team Blog

No credit on the folding site, or on the extreme stats site?

work credit delay...