project 1799 very slow.

[kiore]

Am running a Gromacs 33 project 1799 319 point unit, it is running ok but seems extraordinarily slow and getting slower.
Usually the gromacs 33 units are fast, is there an issue with this unit, or is fahmon reading the points wrong?
Currently running on a Phenom II 940 3GHz winXP32 at 134ppd.
SSE boost is on.

Code: Select all

 Project: 1799 (Run 0, Clone 72, Gen 1)
[09:19:49] 
[09:19:49] Assembly optimizations on if available.
[09:19:49] Entering M.D.
[09:19:55] FAH Init
[09:19:55] Checkpoint file: 
[09:19:59] (Starting from checkpoint)
[09:19:59] Read checkpoint
[09:19:59] Protein: 41361
[09:19:59] Writing local files
[09:20:00] Completed 80000 out of 500000 steps  (16 percent)
[09:20:01] Extra SSE boost OK.
[09:51:24] Writing local files
[09:51:24] Completed 85000 out of 500000 steps  (17 percent)
[10:28:15] Writing local files
[10:28:15] Completed 90000 out of 500000 steps  (18 percent)
[10:58:19] Writing local files
[10:58:19] Completed 95000 out of 500000 steps  (19 percent)
[11:28:22] Writing local files
[11:28:22] Completed 100000 out of 500000 steps  (20 percent)
[12:00:12] Writing local files
[12:00:12] Completed 105000 out of 500000 steps  (21 percent)
[12:33:28] Writing local files
[12:33:28] Completed 110000 out of 500000 steps  (22 percent)
[13:05:06] Writing local files
[13:05:06] Completed 115000 out of 500000 steps  (23 percent)
[13:36:01] Writing local files
[13:36:01] Completed 120000 out of 500000 steps  (24 percent)
[14:07:26] Writing local files
[14:07:26] Completed 125000 out of 500000 steps  (25 percent)
[14:36:50] Writing local files
[14:36:50] Completed 130000 out of 500000 steps  (26 percent)
[15:05:38] Writing local files
[15:05:38] Completed 135000 out of 500000 steps  (27 percent)
[15:19:48] + Working...
[15:37:49] Writing local files
[15:37:49] Completed 140000 out of 500000 steps  (28 percent)
[16:10:18] Writing local files
[16:10:19] Completed 145000 out of 500000 steps  (29 percent)
[16:44:27] Writing local files
[16:44:27] Completed 150000 out of 500000 steps  (30 percent)

NB project summary gives a different points total.

Addit, downloaded new projects again which adjusted the points up but still just 177ppd.

[P5-133XL]

Check your task manager to see if something else is using the CPU leaving little for folding. Folding is designed to always get out of the way rather than interfere with your own personal usage of the machine.

[kiore]

I have even tried pausing another client so the cpu at only 75% but it doesn't change.. My old single core laptop at 1.8GHz is running a 1798 at almost this speed and usually have no issues on the Phenom running 4x uniprocessors and 4 gpus so am thinking it is this unit that is slow for some reason, will let it finish if it will but half expecting it to die on me..

[Nathan_P]

I have even tried pausing another client so the cpu at only 75% but it doesn't change.. My old single core laptop at 1.8GHz is running a 1798 at almost this speed and usually have no issues on the Phenom running 4x uniprocessors and 4 gpus so am thinking it is this unit that is slow for some reason, will let it finish if it will but half expecting it to die on me..

these are monster units, my e6600 has been running one for 2 days now and is only at 50%!!!! it looks like i will get 300ppd but thats only a guess as they are not on psummary

[patonb]

My t5800 is chugging a 1798 slowly too... which sucks as its not 24/7.

[kiore]

Psummary does list them but at just over 400 points, but they move at the same rate as the 2k + ones that were coming out, that's why I wondered if the points were wrong .

[toTOW]

I already said that these WU (and p1798 which is similar) should earn the Big WU bonus (upload size is 18 MB) ... but I guess no one heard me

[John_Weatherman]

Running a p1799 (R37 C172 G0) WU on a P4 3.0 with HT and that's getting 142 PPD (according to FAHMON) - 42 min per %. Which makes it about right with the benchmark P4 2.8. Mind you it's only done 7% so I'll be keeping an eye on it.

[bruce]

Running a p1799 (R37 C172 G0) WU on a P4 3.0 with HT and that's getting 142 PPD (according to FAHMON) - 42 min per %. Which makes it about right with the benchmark P4 2.8. Mind you it's only done 7% so I'll be keeping an eye on it.

Does "with HT" mean you're running 2 clients, or just one?

The benchmark process assumes that there is one client running and nothing else. Under those conditions, you "should be" getting close to 118 PPD so 142 is good, and if there's anything else running, it's excellent.

I've got several 1799/1798's running here and they do take a long time.

[John_Weatherman]

There's only one client (7im would slap my wrists and give me a good ticking off, if I ran 2 clients on a HT machine ) and speed fan running (and whatever XP decides to do without my knowledge, of course). As i said it's only at the beginning so we'll see how it goes.
Edit - I see I've a Project: 1799 (Run 22, Clone 172, Gen 0) on another machine with just a 2.8 ghz P4, and doing nothing else it takes 52 mins per % - which is 114 PPD according to FAHMON.

[new08]

Project number: 1799 (14-140-0)
Work unit: p1799_AQUIFEX_300K
WU size: 911.0 KB
WU result:
Credit: 412.00
Frames: 100
Core: GROMACS33
Server IP: 134.139.127.32
PPH (points per hour): 3.62
PPD (points per day): 87
Avg time per step: 1:08:17
Bonus factor: 1.0000
Client.cfg: bigpackets=normal
Completed: 29%
FahSpy 2.0.1[/quote]

This unit does seem under priced
It's doing little better PPD than my old T21 700 @ a third the speed CPU.
Some posts on line show this has been around as a query a few years- but points increased from 300 ish sometime, maybe to compensate- but not enough as commented elsewhere. PS Normally @ about 95% or more cpu.

[kiore]

I reconfigged my console clients today and changed from big to normal size and stopped advanced . However I still get these units on 'normal' size despite them seeming to be big units.
Have now got 2 ( a 1798 & 1799) on my Phenom ii 955 @ 3.6 GHz and they are running at 260 ppd which is better than the 180ppd on my 940, but for comparison a p 1771 on the 955 is currently showing 903ppd.
Think these units at their current setting are going to cause problems if they are landing on machines configged for normal and not adv.

[new08]

Anything I can do in advanced config to improve things -or stop these units loading?
I can live with it -as the work does just need doing, but these units just seem to be struggling on some machines, like mine.
Sempron 2600+ clocking @ 1685 on 1GB -needing a leg up on occasion

[P5-133XL]

I don't have any uniprocessor clients so I have no experience as to what configuration these need. However, you can fiddle with -advmethods; big, or small WU's allowed, how much memory is allocated and maybe you can find a configuration that still gives out WU's but not these.

[bruce]

This unit does seem under priced
It's doing little better PPD than my old T21 700 @ a third the speed CPU.
Some posts on line show this has been around as a query a few years- but points increased from 300 ish sometime, maybe to compensate- but not enough as commented elsewhere. PS Normally @ about 95% or more cpu.

I'm not sure why you feel they're "under priced" The official standard is 110 PPD on a 2.8 GHz P4 and John says he's getting at least that much.

. . . on another machine with just a 2.8 ghz P4, and doing nothing else it takes 52 mins per % - which is 114 PPD according to FAHMON.

Perhaps it's a little bit like being a drug addict. It continually takes more to satisfy us.