SPCR Folds Team Blog

A forum just for SPCR's folding team... by request.

Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee

colm
Posts: 409
Joined: Tue Jan 31, 2006 8:22 am
Location: maine

Re: great

Post by colm » Tue Apr 27, 2010 12:14 am

cordis wrote:Glad people have been having good results, finally. In other news, looks like they're going to be updating the GPU client soon, so in the spirit of being prepared, here's an fahaddict post about it.

http://en.fah-addict.net/news/news-0-21 ... o-fold.php

I'd urge you all to get as caught up as you can. I'm a little worried about how easy it will be to update my linux machines, but let us all know if you have any problems, we may be able to help.
I also read one post below that one about older clients being useful.
I clearly remember cranking out alotin a short time a few years ago... I did not take forever getting to 33,000. The sudden demise of the p4 doing well kept me away for a couple of years. I still get these crazy WU with 100 point offerings for 27 hours running. that is just ridiculous. the vid card died too...that is only 511 a day...back in 2008.
Seeing classic clients tells me market stampede is not involved with folding integrity.. I really want to see this..it has been my number one complaint for a long freakin time. I have a machine willing, and clients not up to par. That is thier problem..not the user with older stuff. :roll:

NeilBlanchard
Moderator
Posts: 7681
Joined: Mon Dec 09, 2002 7:11 pm
Location: Maynard, MA, Eaarth
Contact:

Post by NeilBlanchard » Tue Apr 27, 2010 4:15 am

Well, I'm not out of the woods -- that one work unit was completed last evening. BUT, then I had SIX WU's fail to get anywhere. Is this because my machine is NG, or is it because the WU's are really flaky?

aristide1
*Lifetime Patron*
Posts: 4284
Joined: Fri Apr 04, 2003 6:21 pm
Location: Undisclosed but sober in US

Post by aristide1 » Tue Apr 27, 2010 9:38 am

http://www.xbitlabs.com/articles/cpu/di ... html#sect0

Ah, here we see the AMD hex core performing worse than an i7-860 but using more power. Not much of a surprise is it?

dereksbelanger
Posts: 123
Joined: Thu Feb 26, 2009 5:59 am
Location: Worcester, MA, USA

Post by dereksbelanger » Tue Apr 27, 2010 9:53 am

Hey Neil - sorry to hear aobut your folding issues. I dont know anything about what is going on with your setup - but if it helps I can tell you that the clients I have running arnt having problems folding WUs.

Windows XP sp3
UNI 6.23 [projects: a bit of everything]
SMP 6.29 <passkey> -advmethods [projects 602x - 605x]

cordis
Posts: 1082
Joined: Thu Jan 15, 2009 10:56 pm
Location: San Jose

bad WU

Post by cordis » Tue Apr 27, 2010 1:03 pm

Hey Neil, well, I think we're at the limit of what my platitudes can do for you, at this point you should probably check this bad WU forum:

http://foldingforum.org/viewforum.php?f=19

And see if your WUs are among those that people are already complaining about. If not, then you should probably take your issue to this forum:

http://foldingforum.org/viewforum.php?f=58

and post a message about your issue. Be sure to paste in your log file, those guys over there can usually pick up error messages that aren't common knowledge. Good luck!

colm
Posts: 409
Joined: Tue Jan 31, 2006 8:22 am
Location: maine

Post by colm » Tue Apr 27, 2010 5:04 pm

aristide1 wrote:http://www.xbitlabs.com/articles/cpu/di ... html#sect0

Ah, here we see the AMD hex core performing worse than an i7-860 but using more power. Not much of a surprise is it?
I was checking up on upgrades.. I could vomit. The six core email I got yesterday showing me it is for sale now at new egg.

L2 is 256x4 for the extreme intel 1366 socket.. with the monster L3..
it is clearly the smarter from the six core, and 130w is not possible. When are they gonna realize this?
approaching a real 90w cpu consumption even attacks SECC steel triple urethaned and pops a mosfet... there is a hoax in our midst.

I stay with prescott until I see 512 L2 for each and every core, or even more than 512. that business is in a shrinking relapse so help my truth allmighty. They don't even know what a fat singular channel is anymore. foliding uses little tiny stabs and benefits all newer multicore, and that is about it. :idea:

and for 550 dollars, I would go buy a junk subaru and drive it home...on a real highway.

for now, I am staying cool at a 500ppd or so donation. Windows 7 finally fixed some serious issues for my prescott workstation. Vista and Xp was stuck as some kinda midget mode internally .. anyhoo patience is a virtue.

cordis
Posts: 1082
Joined: Thu Jan 15, 2009 10:56 pm
Location: San Jose

Post by cordis » Tue Apr 27, 2010 6:30 pm

aristide1 wrote:http://www.xbitlabs.com/articles/cpu/di ... html#sect0

Ah, here we see the AMD hex core performing worse than an i7-860 but using more power. Not much of a surprise is it?
Yeah, I saw that newegg is having a sale on 3.2GHz 6 core Phenom for $310, so if you picked up two of those with a motherboard like this:

http://www.newegg.com/Product/Product.a ... -_-Product

you'd have a 12 core system running about as fast as an i7 980x, but for less than the cost of the intel cpu alone! But you're right, it would use about double the power, so if you ran this for folding, the power bills would wipe out any cost benefit fairly soon. It'd be interesting to see if dual 6 core AMD chips would fold faster than the i7-980x, where half the cores are hyperthreaded.

NeilBlanchard
Moderator
Posts: 7681
Joined: Mon Dec 09, 2002 7:11 pm
Location: Maynard, MA, Eaarth
Contact:

I posted this in the Folding Forum, too

Post by NeilBlanchard » Tue Apr 27, 2010 8:10 pm

Hello,

I'm Folding on an iMac OSX 10.6 with the latest SMP client and the a3 core. It has completed 3 WU's successfully, but *dozens* of the 6012, 6014, and 6015 (mostly the former two) have failed to get past 0%, or 1% or maybe 2% or 3%. After too many failures, it is deemed unstable and goes into a 24 hour waiting period. No WU's from any of these three projects have worked.

The three WU's that have been successfully completed are all 6013. This machine has been Folding with earlier clients for over 2 years with virtually no problems.

KansaKilla
Friend of SPCR
Posts: 381
Joined: Fri Aug 01, 2003 12:13 pm
Location: Rochester, MN

Post by KansaKilla » Wed Apr 28, 2010 6:14 am

Neil,

Did you run a ram test like memtest?

Is your heatsink clear? Could be overheating with higher workloads on some WUs.

NeilBlanchard
Moderator
Posts: 7681
Joined: Mon Dec 09, 2002 7:11 pm
Location: Maynard, MA, Eaarth
Contact:

Post by NeilBlanchard » Wed Apr 28, 2010 6:42 am

I got an answer on the Folding Forum that it may be the combination of OSX 10.6.3 and the FahCore_a3.exe core, that is the problem.

The projects that have problems have never worked, and the 6013 project has never failed. The machine has Folded for over two years, and I did clean it out (when I had the case open to replace the HD), and I've got fan control software (called smcFanControl) that actually makes the fan run pretty fast once or twice a day. The CPU runs hot in these machines -- right now its at 83-86C, and the CPU blower fan is a little less than 1500 RPM -- it is totally inaudible. I'll check next time the fan cranks up (when it gets just a *little* bit hotter than that!) -- I think it gets up close to 3600RPM and you can hear a low pitched hollow-tube-air-flow sound that is definitely audible, but not loud. It then slowly ramps back down in a few minutes -- the temp drops to ~72-74C for the duration.

BTW, the smcFanControl program has full control of all three fans in the machine; and since the stock settings are absolute minimums, you can increase the fans (like the one for the HD) and you still do not hear the fans. These are "squirrel cage" blower fans, and they are very effective and very quiet.

dereksbelanger
Posts: 123
Joined: Thu Feb 26, 2009 5:59 am
Location: Worcester, MA, USA

RAM, Folding & You

Post by dereksbelanger » Wed Apr 28, 2010 11:59 am

How much RAM is needed for a higher-end dedicated SMP folding computer?

Does high performance RAM contribute to higher PPD? Would running a single channel be a handicap?

Does -bigadv have a noticeable impact on RAM useage, and what is its influence on PPD?

cordis
Posts: 1082
Joined: Thu Jan 15, 2009 10:56 pm
Location: San Jose

memory requirements

Post by cordis » Wed Apr 28, 2010 1:08 pm

For the new A3, memory will scale with core count. I think I read someplace that for each SMP A3 thread takes about 300-500MB, but that may be on the high side. I have an i7 860 running A3 jobs in 4GB, not sure exactly how much memory it takes. I had a little trouble with the i7 980x because the bigadv jobs wound up using around 5.5GB out of the 6GB I had installed, so I brought it up to 12GB just to make sure there was enough headroom. I think the bigadv job on the other i7 uses around 4GB, so it seems to be fine with 6GB.

As far as single vs. double or triple channel goes, I think there would be some speed drop, but probably not a big one, maybe at most 10%. Of course, the hit would be bigger for jobs that use more memory, so bigadv jobs would probably be hardest hit.

aristide1
*Lifetime Patron*
Posts: 4284
Joined: Fri Apr 04, 2003 6:21 pm
Location: Undisclosed but sober in US

Post by aristide1 » Wed Apr 28, 2010 2:38 pm

Bah, I've never seen any research or had any personal experience that would even remotely suggest that AMD processors would be part of any efficient pc that emphasizes PPW (or even PPD for that matter.) As far as I'm concerned they play a supporting role to a motherboard with multiple GPUs and not much else.

Of personal preference - I wish some company had the brains to put the GTX480 cooler on the GTX470 to lower temps. Any manufacturers out there paying attention? I doubt it.

cordis
Posts: 1082
Joined: Thu Jan 15, 2009 10:56 pm
Location: San Jose

bummer

Post by cordis » Wed Apr 28, 2010 3:29 pm

NeilBlanchard wrote:I got an answer on the Folding Forum that it may be the combination of OSX 10.6.3 and the FahCore_a3.exe core, that is the problem.

The projects that have problems have never worked, and the 6013 project has never failed. The machine has Folded for over two years, and I did clean it out (when I had the case open to replace the HD), and I've got fan control software (called smcFanControl) that actually makes the fan run pretty fast once or twice a day. The CPU runs hot in these machines -- right now its at 83-86C, and the CPU blower fan is a little less than 1500 RPM -- it is totally inaudible. I'll check next time the fan cranks up (when it gets just a *little* bit hotter than that!) -- I think it gets up close to 3600RPM and you can hear a low pitched hollow-tube-air-flow sound that is definitely audible, but not loud. It then slowly ramps back down in a few minutes -- the temp drops to ~72-74C for the duration.

BTW, the smcFanControl program has full control of all three fans in the machine; and since the stock settings are absolute minimums, you can increase the fans (like the one for the HD) and you still do not hear the fans. These are "squirrel cage" blower fans, and they are very effective and very quiet.
Yeah, now that I look around there are some posts about it, here's one:

http://foldingforum.org/viewtopic.php?f=58&t=14387

That points to another thread. Pretty annoying, I hope they get an update to that out sometime. In the meantime, it sounds like the alternatives are to roll back your OS, or turn off advmethods and hope that there are still some A1 and A2 wus around. Not great alternatives. That is strange about the 6013 WUs working, not sure why that would happen. Maybe they have new code in those that they're testing to fix this. But at this point, even if you left it on to pick up the odd 6013, you'd get so many fails on the other wus that you won't be getting bonuses for a long time. Screwed either way, really, that sucks, dude. :(

aristide1
*Lifetime Patron*
Posts: 4284
Joined: Fri Apr 04, 2003 6:21 pm
Location: Undisclosed but sober in US

Post by aristide1 » Wed Apr 28, 2010 7:58 pm

Boy oh boy, you gotta love that zoom capability at NewEgg when displaying photos of their products. Nothing like zooming in on $200+ 1156 motherboards and seeing the word FOXCONN printed on the cpu socket. Of course, sometimes you need to flip the board over because it's printed on the back.

And in case we haven't noticed we're approaching a milestone on this thread. :mrgreen:

aristide1
*Lifetime Patron*
Posts: 4284
Joined: Fri Apr 04, 2003 6:21 pm
Location: Undisclosed but sober in US

Post by aristide1 » Wed Apr 28, 2010 8:42 pm

Hey ya go:

http://vr-zone.com/articles/a-look-into ... 877-1.html

Socket 1156 and QPI are already old news. I think the echo lasted longer than they did.

And over here we have anti-progress or progress in reverse, as NVidia figures out all the way it can cause dain brammage to the new kid in town and sill sell it for a profit.

http://vr-zone.com/forums/645208/-rumou ... -5850.html

They must been following AMDs marketing strategy.

colm
Posts: 409
Joined: Tue Jan 31, 2006 8:22 am
Location: maine

Post by colm » Wed Apr 28, 2010 9:10 pm

cordis wrote:
aristide1 wrote:http://www.xbitlabs.com/articles/cpu/di ... html#sect0

Ah, here we see the AMD hex core performing worse than an i7-860 but using more power. Not much of a surprise is it?
Yeah, I saw that newegg is having a sale on 3.2GHz 6 core Phenom for $310, so if you picked up two of those with a motherboard like this:

http://www.newegg.com/Product/Product.a ... -_-Product

you'd have a 12 core system running about as fast as an i7 980x, but for less than the cost of the intel cpu alone! But you're right, it would use about double the power, so if you ran this for folding, the power bills would wipe out any cost benefit fairly soon. It'd be interesting to see if dual 6 core AMD chips would fold faster than the i7-980x, where half the cores are hyperthreaded.
I remember falling for the amd thing when 1ghz was on an amd athlon ceramic'd cpu. I stayed intel ever since. bad bad experience... Lately, as the 478/604/939 they are getting old, and already slow to noobs.. the amd is looking better. I have stuck with the vid cards since a rage lt pro...

altho, I would grab a tiger 7501 with the flash to dual core and a couple of 2.8ghz as a first dual cpu expereince in a regular box, and regualr atx 12v..for less than 200. I even got some ECC ddr1.. :wink:

meanwhile, I am keeping old sir prescott like the screenshot. vid card folding only...
Image
with cpu, it goes from 600to800 someting a day..been there for a couple of years.

dereksbelanger
Posts: 123
Joined: Thu Feb 26, 2009 5:59 am
Location: Worcester, MA, USA

Re: memory requirements

Post by dereksbelanger » Thu Apr 29, 2010 4:31 am

cordis wrote:For the new A3, memory will scale with core count.

Huh, this is what I found on my clients

P4 3Ghz(2 cores) + A3 Core = RAM 69, VM 100
Xeon 3.2Ghz(4 cores) + A3 Core = RAM 80, VM 125

RAM useage isnt anything near what you are reporting on i7 processors. I dont get it?

cordis
Posts: 1082
Joined: Thu Jan 15, 2009 10:56 pm
Location: San Jose

sorry

Post by cordis » Thu Apr 29, 2010 10:49 am

Yeah, I think I'm confused with bigadv memory, that's a lot bigger, just A3 looks pretty reasonable, apparently. So don't worry about that too much, I guess.

NeilBlanchard
Moderator
Posts: 7681
Joined: Mon Dec 09, 2002 7:11 pm
Location: Maynard, MA, Eaarth
Contact:

Post by NeilBlanchard » Thu Apr 29, 2010 4:50 pm

Yea -- I got a Project 6060, and it seems to be stable! :D

[Edit: Aaaarrrrrhhhggg! We just had two power outages here, and the 6060 WU got to 20% -- and my machine is unstable... :x ]

dereksbelanger
Posts: 123
Joined: Thu Feb 26, 2009 5:59 am
Location: Worcester, MA, USA

Post by dereksbelanger » Fri Apr 30, 2010 3:54 am

haha - you must have bad folding carma or something :p

NeilBlanchard
Moderator
Posts: 7681
Joined: Mon Dec 09, 2002 7:11 pm
Location: Maynard, MA, Eaarth
Contact:

Post by NeilBlanchard » Fri Apr 30, 2010 4:12 am

Now running a Project 6062 at 40% -- I'm crossing my fingers...

aristide1
*Lifetime Patron*
Posts: 4284
Joined: Fri Apr 04, 2003 6:21 pm
Location: Undisclosed but sober in US

Post by aristide1 » Fri Apr 30, 2010 5:23 pm

Jeez Neil, I didn't have that many problems when I took an AMD X2 4200 to 2.8GHz.

Anywhere here's a shot of a GTX480 with a water block. Yeah it's expensive but look, it's a single slot!

http://www.evga.com/products/moreinfo.a ... P3-1489-AR

One day I will live in a small log cabin and use a couple of these to heat the place.

NeilBlanchard
Moderator
Posts: 7681
Joined: Mon Dec 09, 2002 7:11 pm
Location: Maynard, MA, Eaarth
Contact:

Post by NeilBlanchard » Fri Apr 30, 2010 6:31 pm

Woohoo! It completed the WU! Now, it's waiting for work...

aristide1
*Lifetime Patron*
Posts: 4284
Joined: Fri Apr 04, 2003 6:21 pm
Location: Undisclosed but sober in US

Post by aristide1 » Fri Apr 30, 2010 6:38 pm


aristide1
*Lifetime Patron*
Posts: 4284
Joined: Fri Apr 04, 2003 6:21 pm
Location: Undisclosed but sober in US

Post by aristide1 » Fri Apr 30, 2010 7:09 pm

colm wrote: I was checking up on upgrades.. I could vomit.
Which is why I wrote that frustration post recently.

Cordis:

http://www.newegg.com/Product/ProductLi ... 27%3A56958

Here we see 8 cores for $299 and even some 12 cores, but they are G34 server socket, and taking that route you may as well get a dual socket board.

The thing about the 6 core AMD that would concern me is its TDP. All some foolhardy person need do is apply 1.4 volts on it and crank it up to 4GHz. The processor would probably run, but the board may not survive the amount of power it would need to supply it.

cordis
Posts: 1082
Joined: Thu Jan 15, 2009 10:56 pm
Location: San Jose

scary!

Post by cordis » Fri Apr 30, 2010 10:03 pm

Now that's a strange coincidence, I just recently found that site while looking for parts for the case mod I'm working on. I may still order a 5 1/4" drive cage, I got a $10 case from Weirdstuff Warehouse that I'm trying to pull one out of, but damn those rivets are annoying. My case won't be quite that big, but I did consider letting those guys build me a custom one. Too many parts lying around though, and I don't get a chance to get crafty too often. Looks like these guys do decent work, though.

colm
Posts: 409
Joined: Tue Jan 31, 2006 8:22 am
Location: maine

Post by colm » Sun May 02, 2010 2:35 am

aristide1 wrote:
colm wrote: I was checking up on upgrades.. I could vomit.
Which is why I wrote that frustration post recently.


The thing about the 6 core AMD that would concern me is its TDP. All some foolhardy person need do is apply 1.4 volts on it and crank it up to 4GHz. The processor would probably run, but the board may not survive the amount of power it would need to supply it.
I have killed a board with that problem. just one.. it is a rare problem. Cpu survived perfectly, as did everything else. I wonder about those coolers touching mosfets at all. if the board is proper, the coolers are not needed.

the 875p with a full set of evenly paired mosfets are uncooled ..and no problems whatsoever. biggest cpu it can take, and 4gb ecc ram, 72bit path, 725/400 vid card..and the real proof is allowing onboard lan and sound (used to be impossible with heat to run all that onboard with big cpu)

I really am going dual sockets next... old school type or not, not just for folding. It is a learning. I see the dual cores in 604, last version of them are going to realistic prices (used with warranty) I may tackle one this summer.

dereksbelanger
Posts: 123
Joined: Thu Feb 26, 2009 5:59 am
Location: Worcester, MA, USA

Post by dereksbelanger » Mon May 03, 2010 6:33 am

The new six core AMD chips are finally getting some benchmarks.

http://en.fah-addict.net/news/news-0-21 ... g-home.php

aristide1
*Lifetime Patron*
Posts: 4284
Joined: Fri Apr 04, 2003 6:21 pm
Location: Undisclosed but sober in US

Post by aristide1 » Mon May 03, 2010 7:33 am

dereksbelanger wrote:The new six core AMD chips are finally getting some benchmarks.

http://en.fah-addict.net/news/news-0-21 ... g-home.php
AZBrandon still hangs out here but he doesn't fold for us.

Post Reply