600 Point Gromacs!

unregistered · Post by **unregistered** » Mon Dec 27, 2004 10:10 am

600 Point Gromacs!

http://vspx27.stanford.edu/psummary.html

Has anybody gotten one of those WHOPPERS?

JanW · Post by **JanW** » Mon Dec 27, 2004 12:04 pm

Yup, both of my boxes have been folding these 24/7 since Dec 20/21 - recieved 6 in a row so far (3 on each box). They complete in about 2.5 days on my Athlon XP's (2000 and 2200 MHz). Under normal conditions (no big WU bonus) I'd say I can barely make 200ppd. After submitting four of these, I was GREEN (400+ ppd 24hr average)!! Yayyy!!! I was even close to making the top 20 contributors list with just my two Athlon XP's. Really weird - or scary, depending on how you think about it. Needless to say that those 1200 points I received in one day is something I won't top in a while.

These beasts even seem to do some good to the production of our team: three consecutive days of 40k+ points, hadn't happened since early December (or is this just the usual fluctuation)? Surely boosted my production... Maybe a Xmas gift from Stanford?

They still don't have a description up on what the project is actually about.

EDIT: typos corrected

PS: unregistered: I'm due to overtake you in 571 days (if Stanford keeps the supply of 600 point Gromacs up 'til then) !

Webfire · Post by **Webfire** » Mon Dec 27, 2004 12:42 pm

Yeah I got one a couple of days ago.

peteamer · Post by **peteamer** » Tue Dec 28, 2004 4:41 am

Had a few here too.

Pete

Tibors · Post by **Tibors** » Wed Dec 29, 2004 6:17 am

My two fast PCs (2.8GHz and 3.0GHz P4s) are each working on their third one. It takes them about 2 days, 21 hours to finish one. My slow PC (1GHz Via C3) just picked one up too. It is projected to finish in about 14 days, but that is excluding the time I will be playing Settlers IV

petemg · Post by **petemg** » Wed Dec 29, 2004 7:55 am

Yep, here too - I think all of my boxes got them on the same day, because I had 36 hours of panic (0 WUs reported) and then a massive blast of 3000 points as 5 of them reported at the same time. Christmas day, no less! And I didn't even ask Santa for a boatload of points

btw! I'm dark red! I wonder how long it will last... Now if only I can get 3200 more points before 12/31 at 11:59 pm ...

burcakb · Post by **burcakb** » Wed Dec 29, 2004 10:38 am

So that's why I've been getting only the tiny 57 pointers that slow down my Athlon64 to 100 ppd!!! You were all hogging the good ones

Well, at least I started getting a few 300+ pointers in between and I'm back to 200+ ppd on that machine.

teknerd · Post by **teknerd** » Wed Dec 29, 2004 1:04 pm

man all my machines must have been getting those, cause i have half the WU's but more points than the guy right behind me.

unregistered · Post by **unregistered** » Tue Jan 04, 2005 3:11 pm

I GOT ONE!!!!!

YEEEEEE HAAW

burcakb · Post by **burcakb** » Tue Jan 04, 2005 3:34 pm

Heheh, I got one too. Trodas is going to get a serious wakeup call sometime soon

JanW · Post by **JanW** » Wed Jan 05, 2005 6:16 pm

Hmm, I just lost what was going to be the 11th of these 600 point units : it got deleted after 65% of the work done

Anyone had this happen, or any idea of what might be going on? I'm new to folding and this is the first WU I lose. Here's the interesting parts from the log:

Code: Select all

[07:30:24] - Calling './FahCore_78.exe -dir work/ -suffix 00 -checkpoint 15 -forceasm -verbose -lifeline 17175 -version 502'
[07:30:24]
[07:30:24] *------------------------------*
[07:30:24] Folding@Home Gromacs Core
[07:30:24] Version 1.70 (October 24, 2004)
[07:30:24]
[07:30:24] Preparing to commence simulation
[07:30:24] - Assembly optimizations manually forced on.
[07:30:24] - Not checking prior termination.
[07:30:25] - Expanded 2910718 -> 16166533 (decompressed 555.4 percent)
[07:30:26] - Starting from initial work packet
[07:30:26]
[07:30:26] Project: 1135 (Run 8, Clone 26, Gen 0)
[07:30:26]
[07:30:26] Assembly optimizations on if available.
[07:30:26] Entering M.D.
[07:30:33] Protein: p1135_RIBO_FSpeptide_HEL
[07:30:33]
[07:30:33] Writing local files
[07:30:35] Extra SSE boost OK.
[07:30:35] Writing local files
[07:30:35] Completed 0 out of 250000 steps  (0%)
[07:45:36] Timered checkpoint triggered.
[08:00:36] Timered checkpoint triggered.
[08:03:56] Writing local files
[08:03:56] Completed 2500 out of 250000 steps  (1%)

... (a day and a half later) ...

[22:24:01] Completed 162500 out of 250000 steps  (65%)
[22:31:28] CoreStatus = 0 (0)
[22:31:28] Client-core communications error: ERROR 0x0
[22:31:28] Deleting current work unit & continuing...
[22:32:21] Trying to send all finished work units
[22:32:21] + No unsent completed units remaining.
[22:32:21] - Preparing to get new work unit...

Maybe I should start backing up my folding directory regularly...
Nothing particular happened at 22:31:28, that I'm aware of.

peteamer · Post by **peteamer** » Thu Jan 06, 2005 11:03 am

Jan, there's a fair bit of info over at the F@H forums ref Error Codes 0x0 & 0x1, both seem to be only on linux and variants (incl. mac) and appear to be to do with 'incorrect shutting down sequence of the core'.

However I managed it this afternoon on XP by killing the wrong process in task manager

. 50% through a 600 point WU that was then deleted

... and that was after an act of stupidity earlier in the week that cost me a full 600 points.

Pete

unregistered · Post by **unregistered** » Thu Jan 06, 2005 2:18 pm

That happens, especially on fairly new cores like the 600 pointers. It is rare for most, but for peteamer

these things are common. He suffers from VOODOO folder stress SINdrom

.

burcakb · Post by **burcakb** » Thu Jan 06, 2005 3:03 pm

Yeee haaaa. Got my second one. On my pally unfortunately = 100+ hours

JanW · Post by **JanW** » Fri Jan 07, 2005 12:02 am

Thanks peteamer and unregistered for your responses. I'll just wait for now and see if it happens again --- did install a cron job, though, that backs up my folding directory every three hours.

Over at F@H forums, I saw that Error 0x0 means that "the client doesn't really know what happened, except that it's serious"

There seemed to be a lot of talk about glibc problems on Linux, but that was a while ago. It could also be simply instability, for example processor overheating. As the only thing I know about my XP temperature is that it is above 62C (placed a thermal probe right next to the core), that might be the problem... Had been stable for a month, though, running 24/7.

Ah, I guess the 100% big WU bonus we get on these is there for a reason: even after losing an occasional unit, I still want more of them

Which brings me on-topic again: folding number 11 and 12 right now (should be done tomorrow). So far 6000 of my 8740 points come from these biggies

Just thought I'd share for those of you who wonder why you didn't get any

unregistered · Post by **unregistered** » Fri Jan 07, 2005 1:42 am

62º is a little toasty. I have a rig that is similar to yours, except I have an XP2800(@2.09) and a Thermalright SP97/Panaflo92mmL1a(@~9v) combo(thermalrights are cool). According to Mbm5 CPU is 47º. I just vacumed the heatsinks with a crevice attachment, was 54º. Try MBM5, even though support was discontinued, it works on that board(mine same). Speedfan is also an option. I have it on my HTPC, it has more features, but I have had zero time to play with these things lately. I'm not sure if they have Linux versions. I have the same case without the intake fan, stock rear fan and PSU(loud), just one SATA 160 Spinpoint.

JanW · Post by **JanW** » Fri Jan 07, 2005 6:01 am

Sorry for drifting offtopic here...

Yes, unregistered, I found 62C outside the die to be on the hot side, too - and thanks for sharing your experience with that setup. I don't have a Windows license and have given up on reading the on-die diode under Linux for my board, so this is the only info I have.

Having tried around a little, I don't see how to improve things right now. Turning up the voltage on the CPU fan doesn't gain me much in temperature (for some odd reason), and the sacrifice in noise is just not worth it. Before someone suggests re-seating the HS, I'd add that I did that about a dozen times

(no exaggeration!) and while I won't guarantee that it's seated perfectly, it's the best I can do.
My guess was that case air flow is suboptimal --- but your temperatures with a very similar setup seem to prove that wrong. Nevertheless, the case is due for the standard front-bezel mod, and I also have plans for a Bluefront-like PSU relocation (using a SuperTornado 300W), but I won't have time for that for a month or so.

Meanwhile I'm back to 62% on a 600 point WU on that rig. As long as WU losses don't become more frequent, I won't do anything about temperatures for now. It's just that, had these WU's been reported to be flaky, I could have blamed them

But that doesn't seem to be the case

unregistered · Post by **unregistered** » Fri Jan 07, 2005 6:31 am

Check your temp in your bios,during a re-boot, that will give you a fairly accurate reading. It should not cool off much during re-boot.

I just sent a completed 600 pointer to Stanford and have another going early afternoon. It's going to be my best day ever.

wooglin · Post by **wooglin** » Fri Jan 07, 2005 8:24 am

you could always clock down the FAH cpu utilization to 70% or so, see if the temps decrease a bit... sure the WU will take a bit longer, but at least you won't lose them

JanW · Post by **JanW** » Fri Jan 07, 2005 9:36 am

Thanks wooglin, I'll keep this one in mind --- just one stupid question: how do I specify a maximum CPU utilization for FAH under Linux?

wooglin · Post by **wooglin** » Fri Jan 07, 2005 9:52 am

errrr... sorry, I was thinking about the GUI. I don't know the flags for the text-console I assume you're using under Linux

Putz · Post by **Putz** » Fri Jan 07, 2005 1:06 pm

Use the "-config" flag when starting it to set processor utilization. Say "yes" to advanced options when prompted.

Example:

Code: Select all

./FAH502-Linux.exe -config <other options, such as verbosity, advmethods, etc.>

The "-configonly" option also allows configuration of the client, but exits after configuration is complete instead of proceeding normally.

JanW · Post by **JanW** » Fri Jan 07, 2005 3:02 pm

Thank you Putz and sorry again to everyone for hijacking this thread!

x1m · Post by **x1m** » Tue Jan 11, 2005 2:58 pm

Mine seems a little slow. Its taking me an average of 41mins/frame on a 3200+. That sound right?

tomcat · Post by **tomcat** » Tue Jan 11, 2005 3:10 pm

>Mine seems a little slow. Its taking me an average of 41mins/frame on a >3200+. That sound right?

seems a bit slower than normal. i get 35:16 per frame on a 2600+ overclocked and 34:08 on the overclocked XP-M 2500+ (2.3Ghz).

(the A64 3200+ at 2.4Ghz does them at 29:26 per frame)

Markus

unregistered · Post by **unregistered** » Tue Jan 11, 2005 3:55 pm

How much RAM do you have in that box? I think these peptide gromacs unlike typical WUs, are RAM demanding, it also sends a rather large results file.

That is slower than what I have been seeing on my 2800 or [email protected].
Ive noticed the WUs are not very consistent time/frames. Do you have something else runing in the background?

nici · Post by **nici** » Wed Jan 12, 2005 6:35 am

Mine is also at 29 minutes per frame.. athlon 2600+, 1Gig ram. Oh, and i think you are right about these using a lot of memory, FahCore_78.exe is using 104720K of memory..

I assume this isnt common, i have not checked mem usage during other WUs so im not shure but it seems really high.

edit:typos..

Tibors · Post by **Tibors** » Wed Jan 12, 2005 10:17 am

Well these WU's aren't nicknamed BigWU's for nothing. And the 240MB minimum physical memory required to get them is for the same reason.

On my Hush: 104728K memory used by FahCore_78.exe.
On my P4 2.8: 104816K memory used by FahCore_78.exe.
On my P4 3.0: 103392K memory used by FahCore_78.exe.

This is why you get double the points for working on these WU's.

peteamer · Post by **peteamer** » Wed Jan 12, 2005 10:53 am

For the record...

55mins a frame on an XP2400+

1:05 - 1:10 on a XP1600+

Though these times vary a bit, depending on how ... 'chewy'... the mouthfulls are.

Pete

herosformula · Post by **herosformula** » Thu Jan 20, 2005 12:04 pm

I built a dual Opteron system after the holidays, all it appears to get are these 600 point units.