160 point Gromacs

CoolGav · Post by **CoolGav** » Fri Mar 19, 2004 1:10 am

Got one of these bigguns now, having never seen one before. Says its a p859_p53dimer859 with 100 frames for a total of 160 points

LogStats is claiming my Barton (currently running with 205FSB ~2.25GHz) is getting 100PPD from this one, even although its taken about 8 hours for 20% completion.

Anyone else seen big point proteins like this? Am I the "king of the bigguns"?

haysdb · Post by **haysdb** » Fri Mar 19, 2004 1:25 am

I have TEN of these! My Linux server and every one of my 7 blades got one of these, plus two of my WinXP machines.

The PPD are perfectly normal, but even my fastest machine will take over 34 hours to finish one, up to 56 hours for my slowest.

My points are going to take a BEATING today, because these 10 cpu's will not submit any completed work. Saturday should be good though, as I will turn in 7 of these.

David

dukla2000 · Post by **dukla2000** » Fri Mar 19, 2004 1:34 am

haysdb wrote:My points are going to take a BEATING today, ...

ditto

My (nearly) XP3200 Barton is slugged for 42 hours on a P858, with a XP2100 slugged for 52 hours by a P859! In the next hour my last box will come in for a pit-stop: will be interesting if it also gets slugged!

On reflection, my mind is boggling at how long it took the Stanford benchmark 500MHz Celeron to calibrate these things!

geordie · Post by **geordie** » Fri Mar 19, 2004 4:41 am

I want one! I want one!

Actually 7 would be very nice. My offline dual Xeon keeps getting little 18 pointers so it's a bugger to keep it topped up with work, especially over weekends. A few of these would make it much easier.

Trip · Post by **Trip** » Fri Mar 19, 2004 5:36 am

I've got a 400 WU, is that big?

mas92264 · Post by **mas92264** » Fri Mar 19, 2004 6:32 am

Trip wrote:I've got a 400 WU, is that big?

Did you just get started? Don't see you in the stats.

Oh. Re: 400 Wu. Not necessarily. That's just the total, umm, pieces, I guess. Higher point units take longer per frame to fold (a frame is usually or always 1/100 of the wu.) Take a peek in your log file and it will tell you which wu it is and show your time per frame.

Welcome to folding, Trip!

M

mas92264 · Post by **mas92264** » Fri Mar 19, 2004 6:40 am

Yep, snagged 2 of the big uns here too. My p4 has got the 859/858 combo - 160 pts each.

Yikes! 31:52 and 31:39 per frame running 2 at a time with ht. Won't be done 'til Monday.

M

dasman · Post by **dasman** » Fri Mar 19, 2004 7:11 am

I have 7 of them -- taking between 25 & 45 minutes per frame -- none of these boxes will finish today and some won't finish until Sunday!

I am getting pretty good PPW numbers out of them tho

Dave

EDIT -- I guess here is MAC's chance to make up ground since it looks like alot of us are going to almost zero out today

CoolGav · Post by **CoolGav** » Fri Mar 19, 2004 7:36 am

I only have the one, and it should only take another 24 hours before its done... I should be #34 by then!

Wrah · Post by **Wrah** » Fri Mar 19, 2004 7:57 am

dasman wrote:I guess here is MAC's chance to make up ground since it looks like alot of us are going to almost zero out today

It's already showing...

Last update points on Arachnids:
SilentPCReview Folding@Home 1,223.1
Team MacOS X 2,445.3

They may be gaining ground now but it's gonna be 2 steps back again in a day or two.

(I hope

)

alcimedes · Post by **alcimedes** » Fri Mar 19, 2004 8:57 am

of course, us Mac folks are getting the same gromacs you all are. it's not our fault the g5's chew through them like nothing.

Macaholic · Post by **Macaholic** » Fri Mar 19, 2004 9:04 am

Special ops in progress! Sorry we did'nt let you know ahead of time! But, if we told you, then we would have to terminate you!

How do you spell relief? F-O-L-D!

sthayashi · Post by **sthayashi** » Fri Mar 19, 2004 9:13 am

Damn, I must have forgotten to put on my Troll-B-Gone this morning. Sorry guys, my fault.

Douglas Bailey · Post by **Douglas Bailey** » Fri Mar 19, 2004 9:37 am

Those 160 point, Chew For Several Days WU's seem to be universal. I'm slogged with them too.

Trip · Post by **Trip** » Fri Mar 19, 2004 10:48 am

mas92264 wrote:
Trip wrote:I've got a 400 WU, is that big?
Did you just get started? Don't see you in the stats.

Oh. Re: 400 Wu. Not necessarily. That's just the total, umm, pieces, I guess. Higher point units take longer per frame to fold (a frame is usually or always 1/100 of the wu.) Take a peek in your log file and it will tell you which wu it is and show your time per frame.

Welcome to folding, Trip!

M

I'm folding for Mac OS

nah j/k, yea I just got started. The comp. is about half way with the folding but I'm in the middle of replacing the North Bridge HS... So depending on whether it survives the upgrade or not I'll be an SPCR folding comrade.

haysdb · Post by **haysdb** » Fri Mar 19, 2004 1:49 pm

alcimedes wrote:of course, us Mac folks are getting the same gromacs you all are. it's not our fault the g5's chew through them like nothing.

OK, I'll bite. How long for a G5 to "chew through" one of these 160 point proteins? Less than 34 hours?

David

Zyzzyx · Post by **Zyzzyx** » Fri Mar 19, 2004 2:37 pm

Seems like these suckers run a bit hotter as well. One of my users in the office mentioned her system sounded like a musclecar revving its engine. Its a basic Dell Dimension system, and the cpu/case fan was speeding up and slowing down. It hasn't done this before, and I noticed her system picked up one of the 160 pt WUs just a couple hours earlier. No luck yet getting it to pick up a different WU, I'll try again before the weekend starts. Only one other system here has a 160pt'er, and no ill reports from that one. Yet.

haysdb · Post by **haysdb** » Fri Mar 19, 2004 2:43 pm

One of my blades crashed last night while working one of these, but only 1 of 7. I am not able to monitor temps on my blades, so I don't know if they are running hotter or not.

David

TRC-13 · Post by **TRC-13** » Fri Mar 19, 2004 3:03 pm

I currently have 11 of these beasts on my systems and temps are exactly the same for me as they were 3 days ago.

Maybe it’s a good thing the graphs are not finished yet over at EOC this will bring havoc to a lot of peoples daily scores.

mas92264 · Post by **mas92264** » Fri Mar 19, 2004 3:30 pm

Yeah, but we'll get a big blast of points when they get finished.

M

NoahJ · Post by **NoahJ** » Fri Mar 19, 2004 3:48 pm

Yeah, well we get these units too, and we are still gaining. Keep that in mind.

alcimedes · Post by **alcimedes** » Fri Mar 19, 2004 4:46 pm

OK, I'll bite. How long for a G5 to "chew through" one of these 160 point proteins? Less than 34 hours?

now now, we already have ease of use, UI, industrial design, and security on you poor guys. last thing i want to do is tell you how bloody fast they are and take away price/performance to boot.

in reality though, we're going to catch up real quick for about a day and a half to two days, then i bet all of the sudden you guys get a HUGE spike. i'll have to time my dual G5 processing one of these beasts.

haysdb · Post by **haysdb** » Fri Mar 19, 2004 5:18 pm

alcimedes wrote:last thing i want to do is tell you how bloody fast they are and take away price/performance to boot.

Let's run that "thought experiment." If a $200 Athlon "blade" takes 38 hours to complete one of these proteins, how quickly would a G5 need to complete one to achieve a higher price/performance ratio?

If ANY G5 can complete one in less than the 34 hours 36 minutes of even a P4 2.66B, I will be VERY impressed.

David

haysdb · Post by **haysdb** » Fri Mar 19, 2004 5:23 pm

NoahJ wrote:Yeah, well we get these units too, and we are still gaining. Keep that in mind.

Interesting that these WU's are going out to Windows, Linux, and OSX.

Are you getting them in the same numbers we are? I'm guessing not, simply because you get your points from a greater number of slower machines, so the performance fractions will typically not be as high, and therefore many of your machines will not be assigned these huge proteins. This is just an educated guess, but I am pretty confident it's true.

It doesn't really matter, except that it will make it look like you are gaining on us faster than you really are. I will be shocked and subdued if you post more points than us on both Saturday and Sunday. If you do, that will be "proof" that there is little that we can do to stop you. In the meantime, I have to believe we are not beaten yet, and we can yet pull a wabbit outa da hat.

David

ColdFlame · Post by **ColdFlame** » Fri Mar 19, 2004 5:28 pm

alcimedes wrote:
OK, I'll bite. How long for a G5 to "chew through" one of these 160 point proteins? Less than 34 hours?
now now, we already have ease of use, UI, industrial design, and security on you poor guys. last thing i want to do is tell you how bloody fast they are and take away price/performance to boot.

Security? Since when Macs had security? If no one writes trojans for Macs is because nobody cares not because they are secure.

FYI most of the PC security breaches come from 2 sources: 1) buffer overruns 2) stupid users that run attached EXEs. I don't see how Macs are better in this regard. Do they sport the NX-bit thingie like AMD 64-bit CPUs do?

ColdFlame · Post by **ColdFlame** » Fri Mar 19, 2004 9:19 pm

I got 1 of these on my champ 2.41 Ghz Barton and it is producing 563 PPW at this moment. Since it's been producing as much as 896 PPW lately I'd say that the way Stanford "values" their WUs is screwed, at least from my personal prospective. I hope that in near future there are going to be more distributed computing projects so that I will have a choice where to put my PCs to work.

Macaholic · Post by **Macaholic** » Fri Mar 19, 2004 9:49 pm

Here you go big fella. Plenty of projects to choose from. Enjoy!

http://www.aspenleaf.com/distributed/di ... jects.html

http://distcomp.rynok.org/dcprojects.htm

haysdb · Post by **haysdb** » Fri Mar 19, 2004 9:56 pm

ColdFlame wrote:I got 1 of these on my champ 2.41 Ghz Barton and it is producing 563 PPW at this moment. Since it's been producing as much as 896 PPW lately I'd say that the way Stanford "values" their WUs is screwed, at least from my personal prospective. I hope that in near future there are going to be more distributed computing projects so that I will have a choice where to put my PCs to work.

You must have got one of those WU's where something interesting is happening, so the computations take longer. My 2.2GHz Bartons are showing

707.37 PPW
703.77
697.68
694.18

And even my 2600+ and 2400+ T'breds running at stock speed are doing

626.33
598.07

A P4 2.66B is doing 874.15

David

haysdb · Post by **haysdb** » Fri Mar 19, 2004 10:17 pm

Guys, we are getting THRASHED by these things today. My daily average is 941 and I have posted a whopping 219 points today. A LOT of people have taken a hit today. As a team we have only 14K points today, with just two hours to go. Team MacOSX has also taken a hit, but not to nearly the degree we have. They have gained almost 5K point on us today. At that rate they'd catch us in 4 days. On the bright side, we should put up some big numbers on Saturday and Sunday.

David