proformance boost, maybe?

shadowscotland

shadowscotland

shadowscotland

Soldato
Joined
31 May 2006
Posts
7,564
Location
West London
Question to the more technically minged out their.

I've got Fah set up as a service on all the workstation here and as such it starts with out the need to log on.

Theirfore will the machine fold quicker if left at log in prompt? Due to windows not using up resources.

Running XP pro

Usually not an issue as user will be using each an ever PC but this week we've got a few empty seats. Just thought I'd ask
 
Ok done the tests and

1) non-loged in pc - steps average 12mins 40
(max 12:58, Min 12:31)

2) Loged on PC (no users) steps average 12:27
(max 13:02, Min 12:13)

3) Loged on PC (with user - word/excel ) steps average 13:20
(max 13:55, Min 12:23)

SSE boost active in all three cases

Observations
Max maximum fold time with user actively working (no prizes their)
Simular Maximum fold times with out user, with and without login (not quite expected)
But lower average times when loged in!
(with exception of 13:02 all other results below 12:30 aka Minimum time with 1)

So it look like windows fully running (no screen saver) will process this particular single WU faster than windows at login screen.
Ok it's only 2% quicker but still.



I'll do tests on multi thread tomorrow and report back
 
Last edited:
Anomalous results due to flaws in the method tbh
Click to expand...

20 plus folds used to calculate each average (mean)
The Mean calc'ed with out highest and lowest give result within 3 sec's but infavour of findings :p
Aka 1) 12:42 and 2) 12:24

Edit: 512MB Ram not ever half used with a 1700
 
Last edited:
found this on http://www.gromacs.org/external/search.html

Is there any way I can make GROMACS run faster?

That depends on your setup. If you are using x86 processors you should definitely make sure that you compile GROMACS with assemblt loops, and that your OS supports SSE instructions if you are using Pentium III/IV processors. If you compiled GROMACS with assembly loops there will be a line in the logfile telling you which loops we are using.
On alpha hardware you might want to play around with enabling/disabling the software invsqrt, and the inner loop vectorization. Modern alpha chips have a fairly fast hardware sqrt, but they also seem to benefit even more from vectorizing the innerloops and using the vectorized invsqrt provided in GROMACS.
If you are using IBM hardware you should locate or download the MASS libraries (mathematical accelerated subsystem). If you provide the location of this library in the LDFLAGS environment variable GROMACS will automatically use fast vectorized inner loops on IBM. (are these subsystem only active post login? just a thought)
On any system apart from Linux/x86 (where we use assembly innerloops) you should also try to use a fortran compiler for better performance, and if you run Linux/alpha you should use the Compaq compilers instead of gcc.

Edit: intresting reading but a bit over my head :o
 
Last edited:
I hear what your all saying - and I agree, just thought I'd post my results as they were the opposite of what I expected.

We had a power cut this morning so twin thread results will be posted at lunchtime now. (see new thread for this morning folding problems :( )
 
duel threads

Ok duel thread results

1) non-loged in pc
Thread 1 - steps average 9mins 52 (max 10:00, Min 9:50)
Thread 2 - steps average 21mins 46 (max 21:56, Min 21:41)

2) loged in pc
Thread 1 - steps average 9mins 53 (max 9:54, Min 9:52)
Thread 2 - steps average 21mins 47 (max 21:50, Min 21:45)

no real differances except range of results more varied in non-loged period.

Different machine to single thread results.
 
You must log in or register to reply here.
Back
Top Bottom