SSE and no SSE folding times

[Planus]

I ran F@H on the same protien with the flag -forceSSE and without, at the times to complete one WU was exactly the same 12mins (on a gromacs core) why is this? I have a ahtlon2500+

[nikhsub1]

Originally posted by Planus
I ran F@H on the same protien with the flag -forceSSE and without, at the times to complete one WU was exactly the same 12mins (on a gromacs core) why is this? I have a ahtlon2500+
Cant be. With SSE your frametime (12mins would not be WU time but frame time) should be about 50% faster than with 3DNow! Rerun it and let it do a few frames each way and you will see.

[Planus]

I did. I ran it for one hour while I was in a physics seminar, did a screenshot, closed it down, took out the flags, reran for 2 hours while in a physics prac, came back, 12 mins BOTH times. Dunno what the hell is wrong

[Cowboy Shane]

Is this with the new 4.0 beta? Did the log say it was using SSE one time and 3DNow the next?

[nikhsub1]

Originally posted by Planus
I did. I ran it for one hour while I was in a physics seminar, did a screenshot, closed it down, took out the flags, reran for 2 hours while in a physics prac, came back, 12 mins BOTH times. Dunno what the hell is wrong
Post the logs of each way, SSE and 3DNow! It is not possible for the times to be the same. What client? What flags did you add? More info please!

[Planus]

Client is FAH4.00. Heres a lil piccy if it works (first time ive done this)

BLAH

http://img.villagephotos.com/p/2003-11/474447/untitled2.JPG

http://img.villagephotos.com/p/2003-11/474447/untitled1.JPG

worky worky mr images???!

*edit*
WOO

[nikhsub1]

For some reason you SSE optimizations are NOT running. You would see after the first frame with SSE that "Extra SSE boos OK". I do not see that in you logs. Look at mine, you will see...

Arguments: -service -advmethods -forceasm -forceSSE

Warning:
By using the -forceSSE flag, you are overriding program
safeguards that monitor the stability of SSE
instructions on your system. If you did not intend
to do this, please restart the program without
-forceSSE. If work units are not completing fully,
then please discontinue use of the flag.

Warning:
By using the -forceasm flag, you are overriding
safeguards in the program. If you did not intend to
do this, please restart the program without -forceasm.
If work units are not completing fully (and particularly
if your machine is overclocked), then please discontinue
use of the flag.

[21:00:15] - Ask before connecting: No
[21:00:15] - Use IE connection settings: Yes
[21:00:15] - User name: nikhsub1 (Team 32)
[21:00:15] - User ID = 2242EA9F40F04F8D
[21:00:15] - Machine ID: 1
[21:00:15]
[21:00:15] Loaded queue successfully.
[21:00:15] + Benchmarking ...
[21:00:18]
[21:00:18] + Processing work unit
[21:00:18] Core required: FahCore_78.exe
[21:00:18] Core found.
[21:00:18] Working on Unit 03 [November 4 21:00:18]
[21:00:18] + Working ...
[21:00:19]
[21:00:19] *------------------------------*
[21:00:19] Folding@home Gromacs Core
[21:00:19] Version 1.53 (October 2, 2003)
[21:00:19]
[21:00:19] Preparing to commence simulation
[21:00:19] - Assembly optimizations manually forced on.
[21:00:19] - Not checking prior termination.
[21:00:20] - Expanded 364169 -> 1767785 (decompressed 485.4 percent)
[21:00:20]
[21:00:20] Project: 344 (Run 51, Clone 54, Gen 13)
[21:00:20]
[21:00:20] Assembly optimizations on if available.
[21:00:20] Entering M.D.
[21:00:41] (Starting from checkpoint)
[21:00:41] Protein: p344_gnra_unf_1
[21:00:41]
[21:00:41] Writing local files
[21:01:25] Completed 230000 out of 500000 steps (46)
[21:01:25] Extra SSE boost OK. )<<<<See that?
[21:14:44] Writing local files
[21:14:46] Completed 235000 out of 500000 steps (47

[Planus]

errr, what do I do?

[Arkaine23]

It didn't even say "extra 3dnow boost ok".

What kind of CPU are you running, and what mothebroard?

post your client.cfg... maybe you chose to disable advanced assembly instructions.

[Planus]

XP2200+ Oc'ed to 2500+speeds on a KT4V mobo
[settings]
username=Planus
team=32
asknet=no
machineid=1
local=1

[http]
active=no
host=localhost
port=8080
usereg=yes

[clienttype]
type=1

[core]
priority=0
cpuusage=100
disableassembly=yes
checkpoint=5
ignoredeadlines=no

*edit*
ARRGG
I SEE

I misread the stupid thing when I configured it, OKAY, NOW lets see if I can get it to work

[Planus]

RIGHT! WOW, thanks! here goes....



[21:29:09] Loaded queue successfully.
[21:29:09] + Benchmarking ...
[21:29:11]
[21:29:11] + Processing work unit
[21:29:11] Core required: FahCore_78.exe
[21:29:11] Core found.
[21:29:11] Working on Unit 02 [November 6 21:29:11]
[21:29:11] + Working ...
[21:29:11]
[21:29:11] *------------------------------*
[21:29:11] Folding@home Gromacs Core
[21:29:11] Version 1.53 (October 2, 2003)
[21:29:11]
[21:29:11] Preparing to commence simulation
[21:29:11] - Assembly optimizations manually forced on.
[21:29:11] - Not checking prior termination.
[21:29:12] - Expanded 373714 -> 1861225 (decompressed 498.0 percent)
[21:29:12]
[21:29:12] Project: 801 (Run 5, Clone 1, Gen 12)
[21:29:12]
[21:29:12] Assembly optimizations on if available.
[21:29:12] Entering M.D.
[21:29:32] (Starting from checkpoint)
[21:29:32] Protein: p801_p53dimer801
[21:29:32]
[21:29:32] Writing local files
[21:29:32] Completed 266560 out of 500000 steps (53)
[21:29:32] Extra SSE boost OK.
[21:33:27] Writing local files
[21:33:29] Completed 270000 out of 500000 steps (54)

Frame time dropped from 12-13 mins to 4mins! Thats 3 times faster!!!!!

*edit* using -service -advmethods -forceasm -forceSSE flags

[nikhsub1]

Originally posted by Planus
Frame time dropped from 12-13 mins to 4mins! Thats 3 times faster!!!!!

*edit* using -service -advmethods -forceasm -forceSSE flags
Yes because before you were running with 'standard loops' which mean NO optimizations whatsoever... Will fold really slow that way. 3DNow! will be about 40 - 50% slower than SSE.

[Fast420A]

That looks MUCH better, get ready to see your PPW on that rig increase by a lot!! :D

[DManeKid]

O_o since when was there a v4??

just checked... interesting...

[altec]

I cant even get the -forceSSE flag to work with my client.....it says that it is a bad argument, so I just left -forceasm on......why?!?!

[dagamore]

altec, the -forcesse only works on the beta v4 client, but the new 2.5(?) client (the curent one for download at sanford) should keep your amd systems folding with sse with just the -forceasm hope this helps

[Planus]

Off topic, but I dont care....Mr. Planus jus got the time for a exam wrong....got in halfway through the stupid thing.....

*is not happy*

on topic again, im getting 12mins on a 2400+ per frame for this work unit, its 683, does that sound right? it sounds a bit much to me, but it might be a slow folder.

[altec]

Where is the v4 beta client available for download?


Edit: Nevermind, I found it.....:p