Anyone try v4.0 beta?

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

Looks like a new -forcesse flag is gonna be our new standard in addition to the ones we've been using.

Also, a new 1.53 gromacs core is out and started to be forced upon us.

Make sure to tell us how that goes, so we can get right to installing it if there is a nice improvement. Maybe we can catch up a bit before the other heavy hitters realize the new core is out and better that ever.

Matt

Seems more like feature changes than actual deep-down changes, know what I mean....?

I will be very busy installing the new client on all my machines, what a bore...

Arkaine23 said:

Seems more like feature changes than actual deep-down changes, know what I mean....?

Click to expand...

Yea, now I do. Heh. Thanks for clarifying.

Matt

Very Odd....CPU usage is 100% for like 15 seconds, then then it drops back to nearly nothing,

I can't wait to get home and start installing!


Maximum machine IDs raised to 8!

Good for those 8 processor servers we all have access too!

OK, ran it in the console for about 2 minutes, closed the console, restarted the service, looks good so far,

so mostly a facelift.. no real speed difference?

BIG heads up here! If you have an AMD CPU and install the new client and new core, you MUST run the -forceSSE flag!!!! Otherwise you will just get 3DNow! boost with the -forceasm flag!!! This is VERY BAD!!! We all must update the flags at once! Check my log file:

Arguments: -service -forceasm -advmethods

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.

[01:18:12] - Ask before connecting: No
[01:18:12] - User name: nikhsub1 (Team 32)
[01:18:12] - User ID = 2C1C27E208D25D3B
[01:18:12] - Machine ID: 1
[01:18:12]
[01:18:12] Loaded queue successfully.
[01:18:12] + Benchmarking ...
[01:18:15]
[01:18:15] + Processing work unit
[01:18:36] Core required: FahCore_78.exe
[01:18:36] Core found.
[01:18:36] Working on Unit 07 [November 4 01:18:36]
[01:18:36] + Working ...
[01:18:36]
[01:18:36] *------------------------------*
[01:18:36] Folding@home Gromacs Core
[01:18:36] Version 1.53 (October 2, 2003)
[01:18:36]
[01:18:36] Preparing to commence simulation
[01:18:36] - Assembly optimizations manually forced on.
[01:18:36] - Not checking prior termination.
[01:18:36] - Go method
[01:18:36] - Expanded 403980 -> 2483227 (decompressed 614.6 percent)
[01:18:36]
[01:18:36] Project: 549 (Run 27, Clone 42, Gen 3)
[01:18:36]
[01:18:36] Assembly optimizations on if available.
[01:18:36] Entering M.D.
[01:19:09] (Starting from checkpoint)
[01:19:09] Protein: p549_BBA5_pf
[01:19:09]
[01:19:09] Writing local files
[01:19:09] Completed 35000 out of 500000 steps (7)
[01:19:09] Extra 3DNow boost OK.

Correct flags are:

-advmethods -forceasm (those you know)

-forcesse (self-explanatory)

-verbosity 5 (5-9 if you want to see a message in the log it writes a checkpoint.)

You can now select how often the wu is check-pointed.

check pointed?

Arkaine23 said:

Correct flags are:

-advmethods -forceasm (those you know)

Click to expand...

Arkaine, forceasm does NOTHING now! It only forces 3DNow! but will not run SSE optimizations on AMD's! However, I have only the -forceasm flag set on my Pentium box and IT runs with SSE with the new core and client. Here is the new log file with the -forceSSE flag INSTEAD of the forceasm flag. I consider the forceasm flag useless now...

Arguments: -service -forceSSE -advmethods

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.

[01:37:24] - Ask before connecting: No
[01:37:24] - User name: nikhsub1 (Team 32)
[01:37:24] - User ID = 2C1C27E208D25D3B
[01:37:24] - Machine ID: 1
[01:37:24]
[01:37:24] Loaded queue successfully.
[01:37:24] + Benchmarking ...
[01:37:26]
[01:37:26] + Processing work unit
[01:37:26] Core required: FahCore_78.exe
[01:37:26] Core found.
[01:37:26] Working on Unit 07 [November 4 01:37:26]
[01:37:26] + Working ...
[01:37:27]
[01:37:27] *------------------------------*
[01:37:27] Folding@home Gromacs Core
[01:37:27] Version 1.53 (October 2, 2003)
[01:37:27]
[01:37:27] Preparing to commence simulation
[01:37:27] - Assembly optimizations manually forced on.
[01:37:27] - Not checking prior termination.
[01:37:27] - Go method
[01:37:27] - Expanded 403980 -> 2483227 (decompressed 614.6 percent)
[01:37:27]
[01:37:27] Project: 549 (Run 27, Clone 42, Gen 3)
[01:37:27]
[01:37:27] Assembly optimizations on if available.
[01:37:27] Entering M.D.
[01:38:00] (Starting from checkpoint)
[01:38:00] Protein: p549_BBA5_pf
[01:38:00]
[01:38:00] Writing local files
[01:38:00] Completed 40000 out of 500000 steps (8)
[01:38:00] Extra SSE boost OK.

It should still force no-checking of prior termination.

Ok, I see -forcesse does that too. Yup -forceasm now useless on AMD's that can do SSE.

I have a Athlon 2500+, what flags should I use? and where do I put them?

So is the forceasm flag useless on Intel CPU's too nikh? Here is my log file with an Intel Celeron Tualatin 1.2G:


Arguments: -local -service -advmethods -forceasm

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.

[01:26:46] - Ask before connecting: No
[01:26:46] - Use IE connection settings: Yes
[01:26:46] - User name: Penguin4x4 (Team 32)
[01:26:46] - User ID = 5DA07E8A69D5CFE3
[01:26:46] - Machine ID: 1
[01:26:46]
[01:26:46] Loaded queue successfully.
[01:26:46] + Benchmarking ...
[01:26:48]
[01:26:48] + Processing work unit
[01:26:48] Core required: FahCore_78.exe
[01:26:48] Core found.
[01:26:48] Working on Unit 08 [November 4 01:26:48]
[01:26:48] + Working ...
[01:26:48]
[01:26:48] *------------------------------*
[01:26:48] Folding@home Gromacs Core
[01:26:48] Version 1.51 (September 25, 2003)
[01:26:48]
[01:26:48] Preparing to commence simulation
[01:26:48] - Assembly optimizations manually forced on.
[01:26:48] - Not checking prior termination.
[01:26:48] - Go method
[01:26:49] - Expanded 405855 -> 2483227 (decompressed 611.8 percent)
[01:26:49]
[01:26:49] Project: 549 (Run 85, Clone 36, Gen 1)
[01:26:49]
[01:26:50] Assembly optimizations on if available.
[01:26:50] Entering M.D.
[01:27:44] (Starting from checkpoint)
[01:27:44] Protein: p549_BBA5_pf
[01:27:44]
[01:27:44] Writing local files
[01:27:44] Completed 325000 out of 500000 steps (65)
[01:27:44] Extra SSE boost OK.
[01:44:34] Writing local files
[01:44:34] Completed 330000 out of 500000 steps (66)
[01:58:18] Writing local files
[01:58:18] Completed 335000 out of 500000 steps (67)
[02:11:57] Writing local files
[02:11:57] Completed 340000 out of 500000 steps (68)

I downloaded and installed the new 4.00 beta client for linux. I replaced the -forceasm with -forceSSE. Do I need to delete the old cores?

Axle

Penguin4x4 said:

So is the forceasm flag useless on Intel CPU's too nikh? Here is my log file with an Intel Celeron Tualatin 1.2G:

Click to expand...

Obviously not, that is what I have discovered! With Pentiums and JUST the -forceasm flag you DO get SSE but not with AMD's. With AMD's you must now use the -forceSSE flag. The client or the core knows the difference between AMD's and Pentiums...

Must alert Wedo to fix the one-click.

axlecrusher said:

I downloaded and installed the new 4.00 beta client for linux. I replaced the -forceasm with -forceSSE. Do I need to delete the old cores?

Axle

Click to expand...

yes, delete the old core.

YES, UPDATE THE One-Click! Luckily all my borgs are Pentiums, just AMD's at home.