Personal computing discussed

Moderators: renee, farmpuma, just brew it!

 
PRIME1
Darth Gerbil
Topic Author
Posts: 7562
Joined: Mon Apr 22, 2002 5:07 pm
Location: , location
Contact:

Hardware List

Thu Apr 25, 2002 11:06 pm

I have noticed in various places people listing what kind of hardware they are running F@H on and I thought why not have a central area for people to post their hardware list and other stats.

For the record I am running:
PIII 800 - WinXP
PIII 800 - Win2K
PIII 800 - Mandrake 8.2
(Dual Pentium Pro 200 - Mandrake 8.2) <i>not online yet</i>

My 2 microsoft machines have been going through the protien A series in less than 2 days a WU and my linux box has been processing the 700 series peptides in around 20 hours. I hope this info helps anyone trying to benchmark.
 
Coldfirex
Graphmaster Gerbil
Posts: 1122
Joined: Wed Dec 26, 2001 7:00 pm
Location: College Station, TX
Contact:

Thu Apr 25, 2002 11:16 pm

My as well ask here. What is the folding client more sensitive to? Raw speed, fpu power, cache, bandwidth? I remember hearing Seti used to be pretty heavy on cache so I was curious.
Your bargaining posture is highly dubious.
 
absurdity
Gerbil Elite
Posts: 890
Joined: Sat Mar 02, 2002 7:00 pm
Location: VT

Fri Apr 26, 2002 12:43 am

I've been wondering the same thing, Cold. Hopefully someone can put us in the know.
 
danny e.
Maximum Gerbil
Posts: 4444
Joined: Thu Apr 25, 2002 3:09 pm
Location: Indonesia/Nebraska/Wisconsin

Fri Apr 26, 2002 8:03 am

it has been taking me 19 or 20 hours for the 5-point jobs on my machine, but i also am on my machine about 6 hours a day... so during that time folding is not as fast naturally... especially when i am playing comanche 4 or something.


takes 11.67 minutes per frame for the 5 point jobs on my machine. that includes time when i am using my puter also though.
You don't have to feel safe to feel unafraid.
 
Steel
Global Moderator
Posts: 2330
Joined: Wed Dec 26, 2001 7:00 pm

Fri Apr 26, 2002 9:44 am

FPU power seems to be the biggest factor. While it was going, the 2.2GHz P4 was cranking out a frame every 15 minutes on the 5 point jobs so the fact that a 1.4GHz Tbird can embarrass it that badly must mean something...
 
Steel
Global Moderator
Posts: 2330
Joined: Wed Dec 26, 2001 7:00 pm

Fri Apr 26, 2002 9:46 am

Oh yeah, here's what I've got going currently:

4x P!!! 1GHz
1x P!!! 850MHz
2x P!!! 800MHz
1x P!!! 450MHz
1x Dual PII 400MHz
1x PII 400MHz
2x P!!! 900MHz Laptops
2x P!!! 650MHz Laptops
And of course the unstoppable power of a Mac G3 450 8).
 
dan@imprezion
Gerbil
Posts: 66
Joined: Mon Mar 04, 2002 7:00 pm
Location: Delaware
Contact:

Fri Apr 26, 2002 9:57 am

So.... Steel.... Intel fan?
 
Steel
Global Moderator
Posts: 2330
Joined: Wed Dec 26, 2001 7:00 pm

Fri Apr 26, 2002 10:25 am

Not especially, it's just what I've got available; aside from the P3 850, all the machines are at work. Trouble free operation is far more important than most bang for the buck and the vendor that we've been dealing with is Intel only (and no, it's not Dell ;)). Although now that there's some viable alternatives to VIA it would be nice to see AMD making more inroads to the corporate environment.
 
IntelMole
Grand Gerbil Poohbah
Posts: 3506
Joined: Sat Dec 29, 2001 7:00 pm
Location: The nearest pub
Contact:

Sat Apr 27, 2002 7:44 pm

As I'm studying the A-Level in Biology and have done all the stuff about why proteins fold and hydrogen bonding, here's the stuff about what the program is all about

Get your water molecule, plain old two hydrogen, one oxygen stuff, not heavy water or anything, the stuff that makes up 99.76% of the oceans...

Right, hyrdrogen has an affinity for electrons of one, whilst oxygen has an affinity of two...

This imbalance means that charge is unevenly distributed between the hydrogens and oxygens, with the end result that the molecules all attract each other, and this is responsible for the follwing things:

    High boiling point of water, for such a small molecule
    High heat capacity
    High Latent heat of fission
    High surface tension


Anyway, this is all well and good, but the reason this has to do with proteins is that the effect is the same other molecules with this affinity thing... e.g. nitrogen (3) and carbon (4)

Of course, this attraction is quite weak and is easily overcome, but the sheer number of hydrogen bonds means that this is a serious force, and is the thing that forms the shapes of proteins (e.g. enzymes get their shape from hydrogen bonding)

So, the hydrogen bonding is responsible for the 3D shape. This is the important bit. Things like enzymes work by having the exact correct shape for the thing they speed up, hence increasing surface area to work on. They do not work on other things, they are specific. Hence a single incorrect molecule could result in disaster for a whole organism... disorganising the bonds and changing the shape (this is the way inhibitors work btw)

These molecules make up amino acids, which make up proteins. Last time I heard, there was a billion billion atoms in a full stop. That's a whole lotta bonds to crunch through !!!

Oh, I forgot to mention things like covalent bonding because AFAIK they don't influence the 3D shape too much....

So the integer work will be fairly intensive. However, the fact that the hydrogen bonding occurs in 3D indicates a bit of a swing to the FPU...

Don't forget that this application is probably more optimised for processors like the P3 and Athlon...

Overall, I reckon the program is more optimised for "legacy" processors, but the Athlon's rather powerful FPU does it no harm whatsoever...

Equal clocked P3 vs. Athlon could be interesting...

Don't think that I'm claiming to be an expert in Biology, because I'm not, but thought you might like to know what its all about,
IntelMole
Living proof of John Gabriel's theorem
 
tarball
Gerbil
Posts: 70
Joined: Tue Jan 01, 2002 7:00 pm
Location: göteborg, sweden
Contact:

Sat Apr 27, 2002 8:01 pm

1 p120
 
Princess Die
Gerbil XP
Posts: 389
Joined: Wed Mar 06, 2002 7:00 pm
Location: Pittsburgh, PA
Contact:

Sun Apr 28, 2002 9:31 pm

I just got with the program a couple days ago on my 1.2GHz Tbird.

On the protienA series I've been turning work units about every 18 - 20 hours or so. I only average about an hour or so a day of real use on the box. :( So...now It's doing something useful.

I just added a Celeron 766 today. I plan on adding at least one P3-450 and an old K6-200 to the mix tomorrow.

EDIT: I'm currently trying to talk a buddy at work to do some farming on 21 P3-500's :lol:
 
DiMaestro
Gerbil Elite
Posts: 890
Joined: Wed Dec 26, 2001 7:00 pm
Location: North Dakota NoMoah!

Fri May 03, 2002 1:10 pm

Yea, those ProteinA's are a bugger. I've been finding on my cele 450 the stats sometimes don't show up on the proteinA's. I think there is a time limit on WU returns.
 
Princess Die
Gerbil XP
Posts: 389
Joined: Wed Mar 06, 2002 7:00 pm
Location: Pittsburgh, PA
Contact:

Fri May 03, 2002 7:43 pm

Yeah, I think the time limit for the ProteinA series is 9 days.

My TBird 1.2 is cranking about 12.2 minutes a frame,and my P3-450 at work is cranking 47 minutes a frame. My folks are on vacation in Florida, I'm watching the house...feeding the pets...my Mom doesn't know she's "helping the cause" with her Cele 766. :D

One of the proteinA's yesterday was taking 18 minutes a frame on the 1.2 , I think it must have been a little steeper on the # of computations.

Does anybody know if there is a difference between the Durons, 1+ GHz and equivalently clocked Athlons as far as times? I'm curious how the L2 cache-size difference affects Folding@Home...
 
Ttocs
Gerbil Elite
Posts: 979
Joined: Tue Apr 09, 2002 9:40 am
Location: Bellevue, WA
Contact:

Mon May 06, 2002 7:18 pm

A few quirks about the CPU's...

Yes, it is mostly FPU.

K6/K6-2/K6-3 (not +) vs Pentium/PPro/P2/P2 Xeon/Celeron: The Pentium/PII beats an equal clocked K6-*, hands down. I have a PII OverDrive (essentially a 512K L2 PII Xeon) that will match my K6-2 500@504 in linux. Seeing as you lose about 15-40% (havent had the exact figure yet), the PII OverDrive will most likely beat the K6-2 504 windows vs windows. A Celeron 300A beats the K6-2 504, and a Celeron 300A overclocked to 450 runs rings around it. I think you get the idea... F@H and G@H does not care much about the size of L2 cache, and its argued whether or not the speed of L2 matters.

K6-2+/3+: I haven't heard much about this, but I hear its similar to a Thunderbird in its FPU, and a K6-3 450 can supposively be taken to 600 easily. I haven't heard any 'official' testing on it, so take from that what you will.

Cyrix/VIA anything: Forget it, use the Cyrix as a housewarmer if you live in a very cold place, and use the C3 as a internet/DVD/MP3/whatever box so it doesn't take cycles away from stronger folders. An 800MHz C3 (in a recent benchmark done by our teammate neevo) will be destroyed by a PIII-550...it takes twice as long on the C3 as it does pn the PIII-550.

Athlon/Duron vs PIII: Athlon/Duron will slightly edge out an equally-clocked PIII in folding. This is partially due to the fact that the FPU in the Athlon is fully-pipelined, and the PIII's is not. For more information regarding this, I found this to be a good read. However, the Tualatins meet or beat equal-clocked athlons, thanks to the extra L2 cache. Durons are equally as strong at folding as the Athlon, since F@H doesn't seem to care much about the size of L2.

Pentium IV: Um... er.. I don't have exact numbers here... but both the P4 and the P4 xeon will lose to the Athlon/Duron. How bad, I'm not sure.. if I remember the numbers right, the P4 is about 1.5x longer per frame than the Athlon/Duron.

If you want some massive folding power, grab a few of these, along with a few of these, stick 2-3 of them in a case with a big power supply, and you have some strong folding power in one box!

Hope this helps.

.:Ttocs
Team [H]ardOCP
 
Ttocs
Gerbil Elite
Posts: 979
Joined: Tue Apr 09, 2002 9:40 am
Location: Bellevue, WA
Contact:

Mon May 06, 2002 8:55 pm

Oh, and.. for those that are interested, the page that lists the proteins currently in progress, the time limit on them, and how much points you'll get for completing one of those units is here:

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

Click on 'Description' on any one of the proteins will show you the descriptions for each of the proteins, if they have one available. The direct link is this:

http://folding.stanford.edu/cgi-bin/allprojects

Hope this helps!

.:Ttocs
Team [H]ardOCP
 
Kuhtarl
Gerbil XP
Posts: 409
Joined: Thu Apr 18, 2002 10:53 am
Contact:

Tue May 07, 2002 1:51 pm

Ok... where to start the tally???

My "fold" (these puns just keep coming, don't they?)

Intel:
PIV 1.8ghz
PIV 1.7ghz
Dual PIII 1.4ghz
Dual PIII 1.3ghz
PIII 1.3ghz
PIII 1ghz (8 of them)
PIII 850mhz
PIII 650mhz
PIII 550mhz (5 of them)
PII 350mhz

AMD:
Tbird 1.2ghz (2 of them)
Tbird 1ghz

I think I got all of the machines I'm folding on... if I missed any of them forgive me. :D
 
DiMaestro
Gerbil Elite
Posts: 890
Joined: Wed Dec 26, 2001 7:00 pm
Location: North Dakota NoMoah!

Tue May 07, 2002 2:45 pm

Wow, that list makes me feely kind of puny.

2 Celeron 300a's @ 450.
1 Athlon Classic @ 1 ghz.
 
Coldfirex
Graphmaster Gerbil
Posts: 1122
Joined: Wed Dec 26, 2001 7:00 pm
Location: College Station, TX
Contact:

Tue May 07, 2002 4:54 pm

Kuhtarl: Those from work?
Your bargaining posture is highly dubious.
 
Kuhtarl
Gerbil XP
Posts: 409
Joined: Thu Apr 18, 2002 10:53 am
Contact:

Tue May 07, 2002 6:39 pm

Coldfirex:
Kuhtarl: Those from work?


Nope. I converted my garage into a geekzone.
I've got the loft portion of the garage set up with 6 workstations and the rest are spread throughout the ground level of the garage.

I have them all networked together and I charge local highschool kids 20 bucks an hour to come play multiplayer starcraft and counterstrike.

I had to sell my firstborn on the blackmarket in order to purchase all of the equipment. My wife also left me because our electric bill meant that we couldn't afford clothes and she was tired of wearing paper bags and used carpet remnants. :(

Then my neighbors complain because of all the squishy 'zergy' noises that come out of the garage at odd hours of the night. I was arrested twice because of this. :x



Actually.... they ARE from work. I got formal approval to put F@H on work computers.
The first story was more interesting though. :D
 
Coldfirex
Graphmaster Gerbil
Posts: 1122
Joined: Wed Dec 26, 2001 7:00 pm
Location: College Station, TX
Contact:

Tue May 07, 2002 6:50 pm

Ya, your other story sounded more believable :)
Your bargaining posture is highly dubious.
 
Princess Die
Gerbil XP
Posts: 389
Joined: Wed Mar 06, 2002 7:00 pm
Location: Pittsburgh, PA
Contact:

Thu May 09, 2002 2:20 pm

Thanks, Ttocs,

That was my big question, how does L2 cache size affect the Duron vs Athlon, 64Kb vs 256...

Yep, I was aware of the raw FP advantage of the K7 core over the P6 units, and clock for clock, its even larger advantage over the P4 core in FPU, due to the fact that it can retire on average more than one FP op per clock.

I normally agree with the Honorable Mr. Stokes :) , but I would stick with the accurate assesment that the P6 FPU is fully pipelined, albeit limited by the lack of independence between its FP Add and FP multiply units....A problem not shared by the K7 FP Add and FP Mul units.

I wasn't aware that the larger L2 of the Tualatins would pull them ahead of the k7s... doh! I hope that doesn't carry over to the Durons after all. :(

I plan on adding a Duron 900 to the fold soon, for a little while anyways, and then upgrade it to an XP 2000+.

Right now, I'm running:

K7-1200
Celeron 766
P3-450
part of another P3-450

:-)

Right now I'm pulling 10-12 minutes per frame ( ProteinA ) on the 1200, depending on the particular set...

Anybody have frame times for 1Ghz+ Durons???
 
radix
Gerbil First Class
Posts: 103
Joined: Sun Jan 13, 2002 7:00 pm
Location: Mountain View, CA
Contact:

Some questions

Thu May 09, 2002 6:46 pm

My Tbird 1.4Ghz is taking about 11 min per frame on proteinAg29aNat. I´m using win98se and the console version of F@H. I have some questions though, maybe someone can help me:

How does the server assign a determined WU to a client? Does it benchmark the client first before deciding which WU to send? I think it works this way, because I´m only getting monster WU´s to process here, and on slower machines I got other types of WU, smaller ones. If it works this way, every type of machine helps, because slower ones won´t get monster WU´s, that would expire even if the computer worked 24/7 full load on them.

Which OS and client version runs faster (using the same machine)? win98?98se?XP?2000?NT? Or linux? The console version or the screen saver?
 
absurdity
Gerbil Elite
Posts: 890
Joined: Sat Mar 02, 2002 7:00 pm
Location: VT

Fri May 10, 2002 2:55 am

I've been running my celeron 800 and duron 650, and they both constantly get a mix of big and small ones...I'm not really seeing any pattern to it.
 
MadManOriginal
Gerbil Jedi
Posts: 1533
Joined: Wed Jan 30, 2002 7:00 pm
Location: In my head...

Fri May 10, 2002 3:15 am

My turn to post hardware:

(My computer) P4 1.8A @ 2.4

and...um, that's it for now :)

I have an ASUS TUSL2-C mobo, Cel 633, etc that is a near-complete system I was intending to sell. hmm...I am now thinking of buying a Tualatin CPU and putting it to work folding. Damn this crap is addicting, and you don't even do anything! lol

Oh, also what happens when a computer is turned off with the program running? does it pick up where it stopped on a protein or have to start all over again? I could run this on my parent's computer I built for them, an XP 1700+, no problem, but I am afraid that it would be worthless if it had to start over each time it was poered up. Anyone know the details on this?
 
radix
Gerbil First Class
Posts: 103
Joined: Sun Jan 13, 2002 7:00 pm
Location: Mountain View, CA
Contact:

Fri May 10, 2002 7:24 am

MadManOriginal wrote:
Oh, also what happens when a computer is turned off with the program running? does it pick up where it stopped on a protein or have to start all over again? I could run this on my parent's computer I built for them, an XP 1700+, no problem, but I am afraid that it would be worthless if it had to start over each time it was poered up. Anyone know the details on this?


Yes, it picks up where it stopped, don't worry about losing results.
 
radix
Gerbil First Class
Posts: 103
Joined: Sun Jan 13, 2002 7:00 pm
Location: Mountain View, CA
Contact:

Fri May 10, 2002 7:30 am

absurdity wrote:
I've been running my celeron 800 and duron 650, and they both constantly get a mix of big and small ones...I'm not really seeing any pattern to it.


Well, maybe I was misleded by the fact that I only got big WUs on my Athlon, and smaller ones on older Pentiums (couldn't send the results for the Pentiums because of a firewall). But, even though, a question still remains: would a Pentium 100 Mhz (or something older) get a ProteinA to process? Because a fast computer can process any WU (yours can), there are no performance restrictions, but a older one won't be able to get some WU's done in time.
 
Princess Die
Gerbil XP
Posts: 389
Joined: Wed Mar 06, 2002 7:00 pm
Location: Pittsburgh, PA
Contact:

Fri May 10, 2002 7:13 pm

Yeah, probably not.

The P5 isn't as fast, clock for clock, as the P6.

But, for the sake of argument, assume it is. My P3-450 is running about 47 minutes a frame, normalizing for the clock speed, the P-100 should turn about 212 Minutes a frame on ProtienA, or about 3.5 hours per frame. 15 days, give or take to run all 100 frames on ProteinA.

I had a similar vintage box, that I decided not to add to the fold due to the times :-(
 
Steel
Global Moderator
Posts: 2330
Joined: Wed Dec 26, 2001 7:00 pm

Sun May 12, 2002 6:42 pm

Steel wrote:
Oh yeah, here's what I've got going currently:

4x P!!! 1GHz
1x P!!! 850MHz
2x P!!! 800MHz
1x P!!! 450MHz
1x Dual PII 400MHz
1x PII 400MHz
2x P!!! 900MHz Laptops
2x P!!! 650MHz Laptops
And of course the unstoppable power of a Mac G3 450 8).

Just added and Athlon XP 1800+ which pulls ~8.5 minutes a frame on the protienA series. :)
 
Ttocs
Gerbil Elite
Posts: 979
Joined: Tue Apr 09, 2002 9:40 am
Location: Bellevue, WA
Contact:

Mon May 13, 2002 6:40 pm

Princess Die wrote:
The P5 isn't as fast, clock for clock, as the P6.


Unless Stanford took a step back in time, and used some OLD OLD FORTRAN compiler, and used 16-bit code. Then we're all in trouble. :o

.:Ttocs
 
PRIME1
Darth Gerbil
Topic Author
Posts: 7562
Joined: Mon Apr 22, 2002 5:07 pm
Location: , location
Contact:

Thu Feb 24, 2005 2:16 pm

Boy things have sure changed, It's been almost 3 years :o

I used to be #7, I used to be a contender :cry:

Now I am running....

XP 2800+
XP 1900+
P4 1.6
Dual Athlon 1.2 MP
P4 2.4
A64 3000+
Image
"Give me a scotch. I'm starving" ~ Tony Stark

Who is online

Users browsing this forum: No registered users and 1 guest
GZIP: On