SPCR Folds Team Blog

aristide1 · Post by **aristide1** » Thu Sep 27, 2007 5:28 am

ARGH!

I was 28 checkpoints into a WU. I installed BitDefender A/V product which promptly put up a firewall. I started FAH, which got a communication error instantly, so it turned around and tossed the WU into the garbage can.

Man, give me a chance to click ALLOW why don'tcha?

KansaKilla · Post by **KansaKilla** » Thu Sep 27, 2007 7:02 pm

man, that stinks. sorry to hear about it. i've given up on doing much except web surfing when folding, so i'm not doing much now with any of my boxes. lost a wu two nights ago when i tried to oc my latest processor, 38 wu's into it. i feel your pain.

aristide1 · Post by **aristide1** » Sat Sep 29, 2007 11:42 am

Yeah, 9 1/2 hours, POOF!, down the drain.

It was bad enough when I had to install Yahoo IM for my current college class, and I ended up with a flippin' Yahoo toolbar on every app and it's mother. I removed it and it asked me why? Bad move on their arrogant butts part.

Now I need to go flash some new BIOS on my Biostar T-Force 6100 so I can OC it some more.

Post by **NeilBlanchard** » Sat Sep 29, 2007 4:53 pm

Hello,

Well, I now have a C2D 6750 (at work) running the Windows SMP client, and it is flying... Here at home, I finally have the Windows SMP running, and it completed one WU, but then on the very beginning of the next one, it gets an I/O error -- and this machine runs the Linux SMP client much faster, anyway. So, back to Ubuntu!

regal196 · Post by **regal196** » Sun Sep 30, 2007 3:11 pm

I just started folding... computer is always on, mind as well put it to good use.

KansaKilla · Post by **KansaKilla** » Tue Oct 02, 2007 7:52 am

Welcome to the red, Neil.

Flandry · Post by **Flandry** » Tue Oct 02, 2007 9:29 am

I've put my SMP dually Tualatin 1.26GHz rig onto it, too. It's an easy choice with these old systems that don't do any dynamic clock adjustments -- doesn't really change the power draw according to my Kill-A-Watt.

Wibla · Post by **Wibla** » Wed Oct 03, 2007 3:57 am

How many minutes per 1% on the dual tually?

Flandry · Post by **Flandry** » Wed Oct 03, 2007 6:33 am

Wibla wrote:How many minutes per 1% on the dual tually?

Well, i dunno. I checked the system this morning and the console window said 7% completed. However, I remember that it was at 4% yesterday morning, so i checked the CPU load and it was 0%.

Weird.

I pushed "esc" in the console window and it closed. Then i clicked on FAH and it started up, said that the core shut down incorrectly, resumed from 7%...

If FAH is going to crap out with no notice every day, i'll just turn off the system when i'm not using it like i used to. Has anyone ever seen this happen before?

Wibla · Post by **Wibla** » Wed Oct 03, 2007 6:48 am

Well, youre gonna get problems hitting deadlines with a dual tualatin even without it crashing, so if you turn it off its gonna miss deadlines, then there's no real point in having it running at all...

aristide1 · Post by **aristide1** » Wed Oct 03, 2007 7:01 am

Wow!

In just 4 months Vg30et made it to position 8 on the SPRC list and donated over 1 million points.

http://folding.extremeoverclocking.com/ ... =&u=251911

Flandry · Post by **Flandry** » Wed Oct 03, 2007 7:24 am

Wibla wrote:Well, youre gonna get problems hitting deadlines with a dual tualatin even without it crashing, so if you turn it off its gonna miss deadlines, then there's no real point in having it running at all...

That was sort of my point: If this mature software product that lets me give the developers something can't handle itself without coddling, it has no place on my computer. What i wanted to know is if the crashing is unusual, or a documented problem with a known (easy) workaround.

If it's not something easily resolved, i'll just uninstall next time it happens.

vanhelmont · Post by **vanhelmont** » Wed Oct 03, 2007 5:49 pm

Today I had to shut down to do some wiring, and when I restartedit said the client was expired, so I downloaded the new one. I didn't know it now requires an option -smp if you want to run it as smp (sounds silly, but they are merging it into the regular client, so you can run it either regular or smp). It didn't like my partially done smp core, so it deleted it. A little later I ran ps, and only one fah process was doing anything (although there were 4 processes there). I went back and read the faq, then stopped my client and ran it with -smp
then it deleted the regular core and downloaded an smp core.

Be forewarned, use the -smp option when you start the new client, and don't lose the work you have done.

Plissken · Post by **Plissken** » Wed Oct 03, 2007 6:03 pm

Is this with windows?
I installed the new SMP client without losing any progress, and without using any new switches.
Where is some info on the -smp switch? I couldn't find it on the Stanford site.

Flandry · Post by **Flandry** » Wed Oct 03, 2007 6:11 pm

A little follow up on my flaky client: it died again at 14%. Apparently every 7% it feels the need to bail out.

I will restart it each morning and night and see how things go. When that gets old, i guess i'll delete it.

Plissken · Post by **Plissken** » Wed Oct 03, 2007 6:42 pm

You might as well delete it now, because you aren't going to make the deadlines. I wouldn't necessarily blame the client itself, since a lot of people use it without any issues. More likely it's a problem with SMP installation or an unstable system. Make sure you choose "no" to using IE settings and "yes" for big packets. For switches, use -advmethods and -forceasm. Please don't give up!

aristide1 · Post by **aristide1** » Wed Oct 03, 2007 6:49 pm

I went through a brief stretch of bad ones a while back. Then it cleared up. Still a waste of time.

Expiration has come and gone and I am still running. Are betas good for 6 months and then every other time you need a new one?

Since it ain't broke I am not fixing it.

avi_dan · Post by **avi_dan** » Wed Oct 03, 2007 7:30 pm

It only checks for expiration when you start the client.

It will refuse to run next time you restart it.

You can install the new version over the old one (if it asks you if you want to uninstall the old one say no) and it will only take you a few minutes.

Cheers

ryboto · Post by **ryboto** » Thu Oct 04, 2007 5:20 am

aristide1 wrote:Wow!

In just 4 months Vg30et made it to position 8 on the SPRC list and donated over 1 million points.

http://folding.extremeoverclocking.com/ ... =&u=251911

without his contribution, we're going to drop in rank pretty dang fast.

Flandry · Post by **Flandry** » Thu Oct 04, 2007 8:39 am

Plissken wrote:You might as well delete it now, because you aren't going to make the deadlines. I wouldn't necessarily blame the client itself, since a lot of people use it without any issues. More likely it's a problem with SMP installation or an unstable system. Make sure you choose "no" to using IE settings and "yes" for big packets. For switches, use -advmethods and -forceasm. Please don't give up!

How do i see when the WU expires?

I don't remember what i chose for settings when i installed (just took the defaults offered). How do i change those?

As for stability -- I've had the system for almost five years now and it's been the most stable system i've ever had (including the P2B based system before it). I never had trouble with SETI or any of the benchmark/stability checking programs.

The program has made it past the next 7% without dying after a reboot, so hopefully it will be ok now. At the reboot i upped the FSB to 142 so that it has more chance of finishing units on time, for a CPU clock of ~1.35 GHz.

ryboto · Post by **ryboto** » Thu Oct 04, 2007 9:41 am

Flandry wrote:
Plissken wrote:
How do i see when the WU expires?

well, the client doesn't give you a date. The WU's expire after 4 days from starting them. So, if you start one, you've got 4 days to finish it. If you pause for too long, you might see "deadline for project XXX passed, deleting work"....that's what I've seen, although the SMP client never stops itself, and never deletes the work. So you have to do it manually. You have to delete both the work folder and the queue.dat file, then start the client again. Only the C2D's can get by with as much as 2 days interruption, but the X2's, and older need more time.

aristide1 · Post by **aristide1** » Thu Oct 04, 2007 10:00 am

The WU's expire after 4 days from starting them.

I've seen that for SMPs, but my Pentium III used to take a week+ to finish a WU and I never got such a message, and the points counted as well.

floffe · Post by **floffe** » Thu Oct 04, 2007 10:26 am

Yeah it's SMP units that have such tight deadlines. My WUs (currently mostly from 2415-2430) have 30 days for the preferred deadline, and another 50 or so until I don't get points anymore. I think after the preferred deadline they send out the WU again to someone else.

Wibla · Post by **Wibla** » Fri Oct 05, 2007 9:27 am

Got new ram for my P182 rig today, Crucial BallistiX PC8500, so now im cruising at 3224MHz and doing 12-13minutes/checkpoint

Gonna put the old Geil Ultra PC6400 ram in the dedicated F@H box ([email protected]) and see how high I can push it, also gonna fit my old E6600 stock cooler on it, as it has a copper base (the cheaper boxed coolers are all-alu)

Flandry · Post by **Flandry** » Fri Oct 05, 2007 4:27 pm

floffe wrote:Yeah it's SMP units that have such tight deadlines. My WUs (currently mostly from 2415-2430) have 30 days for the preferred deadline, and another 50 or so until I don't get points anymore. I think after the preferred deadline they send out the WU again to someone else.

Well, that sucks. How well would it work to run two copies of the normal client?

Plissken · Post by **Plissken** » Fri Oct 05, 2007 5:41 pm

Flandry wrote:Well, that sucks. How well would it work to run two copies of the normal client?

If you can run SMP, you'll get about 2X (or more) points per day than 2 normal clients combined. So it's worth it to get the SMP client running if you have multicore CPU. To answer your earlier question about changing settings, add the -config switch in the command line IIRC.

Flandry · Post by **Flandry** » Sun Oct 07, 2007 4:02 pm

Thanks for all your help. As i was warned, the WU expired before it was nearly finished. I did find out that there is a text file in the directory that gives the expiration date.

Anyway, i uninstalled and deleted the directory to get it to stop working on the already expired WU and reinstalled. I got it going on a new WU and it just stopped doing anything when it got to 6% this time.

It's gone, now. SETI doesn't expect the user to jump through hoops or throw away days of crunching because of the developers' ineptitude -- they'll get my cycles.

Good crunching to you.

aristide1 · Post by **aristide1** » Wed Oct 17, 2007 6:04 am

Wibla wrote:Got new ram for my P182 rig today, Crucial BallistiX PC8500, so now im cruising at 3224MHz and doing 12-13minutes/checkpoint

Gonna put the old Geil Ultra PC6400 ram in the dedicated F@H box ([email protected]) and see how high I can push it, also gonna fit my old E6600 stock cooler on it, as it has a copper base (the cheaper boxed coolers are all-alu)

Do you increase your northbridge cooling?

Wibla · Post by **Wibla** » Thu Oct 18, 2007 6:55 pm

No, not yet.. seems i have some small stability issues at 3.24GHz, but gonna see what I can do about it.. dont really care tho, might just put it back at 3.1GHz as then i know its stable even at near stock voltages...

aristide1 · Post by **aristide1** » Sat Oct 20, 2007 2:35 pm

ARGH! Again!

If my some reason you find yourself doing updates or hardware changes with requires you to unplug your pc may I strongly suggest that you find a way to remind yourself not to restart FAH if your LAN or DSL connection is not reconnected. FAH, with it's great programming, will instantly flush your WU down the drain, and then tell you it can't retrieve another one.
Another 60% of a 1760 pointer down the drain.

DUH!

Don't worry Wibla, I am not going to make a big deal about it over at Stanford.edu. Though I did find Voodoo rather humerous, because my objections came not as a programmer, but as a user of FAH who happens to have a programming background. And even then, my suggestions where based on what I could see was already being done in their code.

Meanwhile the losses will continue to mount.

Take a look at this guy's chart:
http://folding.extremeoverclocking.com/ ... =&u=186535
Here's a very serious contributor with an 8 day outage. Should we ask why?

SPCR Folds Team Blog

New SMP linux client: