Page 1 of 1

2499 frame time

Posted: Fri Aug 14, 2009 12:15 am
by mag00
Why this protein takes so long to process?

I am running it on a dual OPTERON 248, with 16GB of ram.

Code: Select all

[19:02:41] Core required: FahCore_78.exe
[19:02:41] - Autosend completed
[19:02:41] Core found.
[19:02:41] Working on queue slot 01 [August 12 19:02:41 UTC]
[19:02:41] + Working ...
[19:02:41] - Calling '.\FahCore_78.exe -dir work/ -suffix 01 -nocpulock -checkpoint 30 -verbose -lifeline 3876 -version 623'

[19:02:42] 
[19:02:42] *------------------------------*
[19:02:42] Folding@Home Gromacs Core
[19:02:42] Version 1.90 (March 8, 2006)
[19:02:42] 
[19:02:42] Preparing to commence simulation
[19:02:42] - Looking at optimizations...
[19:02:42] - Files status OK
[19:02:45] - Expanded 2966832 -> 15085241 (decompressed 508.4 percent)
[19:02:46] 
[19:02:46] Project: 2499 (Run 93, Clone 14, Gen 10)
[19:02:46] 
[19:02:50] Assembly optimizations on if available.
[19:02:50] Entering M.D.
[19:03:12] (Starting from checkpoint)
[19:03:12] Protein: Translocon_ALX2
[19:03:12] 
[19:03:15] Writing local files
[19:03:15] Completed 21622 out of 250000 steps  (9%)
[19:03:17] Extra SSE boost OK.
[19:20:34] Writing local files
[19:20:54] Completed 22500 out of 250000 steps  (9%)
[19:52:13] Timered checkpoint triggered.
[20:08:30] Writing local files
[20:08:50] Completed 25000 out of 250000 steps  (10%)
[20:40:10] Timered checkpoint triggered.
[20:57:41] Writing local files
[20:58:01] Completed 27500 out of 250000 steps  (11%)
[21:29:21] Timered checkpoint triggered.
[21:45:45] Writing local files
[21:46:06] Completed 30000 out of 250000 steps  (12%)
[22:17:27] Timered checkpoint triggered.
[22:33:47] Writing local files
[22:34:07] Completed 32500 out of 250000 steps  (13%)
[23:05:28] Timered checkpoint triggered.
[23:24:26] Writing local files
[23:24:46] Completed 35000 out of 250000 steps  (14%)
[23:55:07] Timered checkpoint triggered.
[00:15:22] Writing local files
[00:15:42] Completed 37500 out of 250000 steps  (15%)
[00:46:02] Timered checkpoint triggered.
[01:06:12] Writing local files
[01:06:34] Completed 40000 out of 250000 steps  (16%)
[01:36:54] Timered checkpoint triggered.
[01:57:28] Writing local files
[01:57:49] Completed 42500 out of 250000 steps  (17%)
[02:28:09] Timered checkpoint triggered.
[02:48:47] Writing local files
[02:49:07] Completed 45000 out of 250000 steps  (18%)
[03:19:27] Timered checkpoint triggered.
[03:39:42] Writing local files
[03:40:02] Completed 47500 out of 250000 steps  (19%)
[]'s

Re: 2499 frame time

Posted: Fri Aug 14, 2009 12:47 am
by P5-133XL
Well, I'll start out with that this is a uniprocessor WU, so it will only use once processor, regardless of the number of available processors: To more efficiently use the multiple processors you need to run multiple copies of the uniprocessor client. While 16GB is a large amount of ram, surplus ram is not used and does not increase the speed of processing. Then you've got an AMD processor built in 2003 that is equivalent to a high performance P4. So, what are you expecting it to do?

I think a 17.6 minute frame time for a 905 point WU giving a PPD of 740 is very reasonable for a P4 equivalent CPU considering that the 2.8GHz benchmark P4 is supposed to average around 110 PPD. Since you can run two clients your machine can get around 1500 PPD which is very good for that era computer.

Re: 2499 frame time

Posted: Fri Aug 14, 2009 12:57 am
by bollix47
@P5-133XL - the 17.6 minute time frame was for a partial percentage point. The actual frame times are ~ 48-51 minutes.

Re: 2499 frame time

Posted: Fri Aug 14, 2009 1:03 am
by mag00
AOhahoiheoihoiHOihoiEAHOEiA

bollix47 get the correct time info.

I didn't know that this processor was too trash. As I use this machine to run some advanced engineering programs for structural analysis I was thinking that this CPU was in the "same level" of a core2duo.

My boss always say that all of our CPUs are "professional and very powerful". :/

So, if I compare this processor to a xeon the difference of performance on folding can be around of 60% or more?

[]'s

Re: 2499 frame time

Posted: Fri Aug 14, 2009 1:07 am
by P5-133XL
I looked, and Bollix is correct! So with a 50 min. frame time the PPD is 260 which is still good for a processor that is equivalent to a P4. When comparing for reasonableness one should not forget that a 2.8GHz P4 is only producing 110 PPD with no SSE.

Re: 2499 frame time

Posted: Fri Aug 14, 2009 1:19 am
by P5-133XL
mag00 wrote:AOhahoiheoihoiHOihoiEAHOEiA

bollix47 get the correct time info.

I didn't know that this processor was too trash. As I use this machine to run some advanced engineering programs for structural analysis I was thinking that this CPU was in the "same level" of a core2duo.

My boss always say that all of our CPUs are "professional and very powerful". :/

So, if I compare this processor to a xeon the difference of performance on folding can be around of 60% or more?

[]'s
I wouldn't classify it as trash, it is just not up to todays performance standards. It's just a little long in the tooth. When it was new, it would have been a very high performance machine.

Comparing to a Xeon is very hard to do since that name covers a very long time period: From P II's to i7's. All Xeon stands for is Intel's branding of a server CPU regardless of performance. Opteron is AMD's server branding and it also covers a long time period. Rather than looking at the branding, look at the performance capabilities of the chip a 248. I did a google on that chip and here's a review for the 248. Try looking at the benchmarks for it ...

Re: 2499 frame time

Posted: Fri Aug 14, 2009 1:20 am
by bruce
P5-133XL wrote:I looked, and Bollix is correct! So with a 50 min. frame time the PPD is 260 which is still good for a processor that is equivalent to a P4. When comparing for reasonableness one should not forget that a 2.8GHz P4 is only producing 110 PPD with no SSE.
The P4 has SSE. It's SSE2 that has been disabled in the benchmark machine.

Re: 2499 frame time

Posted: Mon Aug 17, 2009 5:12 pm
by SantaFe
So far (According to FAHMon) I'm getting this (On a 3.0 GHz Pentium D, 1 Gig Ram, Ubuntu 9.04}

-- FAH2 --

Min. Time / Frame : 50mn 00s - 260.64 ppd
Avg. Time / Frame : 50mn 14s - 259.43 ppd
Cur. Time / Frame : 55mn 38s - 234.25 ppd
R3F. Time / Frame : 55mn 38s - 234.25 ppd
Eff. Time / Frame : 52mn 32s - 248.07 ppd


-- FAH1 --

Min. Time / Frame : 49mn 57s - 260.90 ppd
Avg. Time / Frame : 50mn 10s - 259.77 ppd
No Cur. Time / Frame
No R3F. Time / Frame
No Eff. Time / Frame

Now if ONLY either core would get ANYTHING other than P2498/P2499's! ;)

Re: 2499 frame time

Posted: Mon Aug 17, 2009 10:54 pm
by mag00
This is the same WU but in my home machine. Q6600 with 4GB RAM.

Code: Select all

[12:47:43] Writing local files
[12:47:43] Completed 71583 out of 250000 steps  (29%)
[12:47:46] Extra SSE boost OK.
[13:00:24] Writing local files
[13:00:26] Completed 72500 out of 250000 steps  (29%)
[13:30:27] Timered checkpoint triggered.
[13:34:50] Writing local files
[13:34:52] Completed 75000 out of 250000 steps  (30%)
[14:04:54] Timered checkpoint triggered.
[14:09:12] Writing local files
[14:09:14] Completed 77500 out of 250000 steps  (31%)
[]'s