F@H V4 client released

[osage]

Beta no more, get it here

for more info on the new client see here

time to update your client if you have not done so yet

 

[BlackMountainCow]

Thx for the info!

 

[RustyNale]

I'm on it

 

[r0tt3n1]

Thanx for the heads up. Too bad I cant update my work machine til NEXT YEAR. I am on vacation til then, darn...

 

[r0tt3n1]

Having bad problem with the v4 Win CLI client on 1 machine. As can be seen in the snippets, using the v4 client slowed down frame times dramatically! Apparently using the default 3DNow optimizations are LAME! Using the -forceSSE flag with the v4 client returned an error about optimizations. hmm, hafta look into further later on. The Linux v4 client is running smoothly.

[09:49:30] Project: 1034 (Run 6, Clone 36, Gen 2)
[09:49:30]
[09:49:30] Assembly optimizations on if available.
[09:49:30] Entering M.D.
[09:49:50] (Starting from checkpoint)
[09:49:50] Protein: p1034_A21nat_305_99
[09:49:50]
[09:49:50] Writing local files
[09:49:50] Completed 261598 out of 2500000 steps (10)
[09:49:51] Extra 3DNow boost OK.
[10:06:45] Writing local files
[10:06:47] Completed 275000 out of 2500000 steps (11)
[10:38:17] Writing local files
[10:38:19] Completed 300000 out of 2500000 steps (12)
[11:09:54] Writing local files
[11:09:56] Completed 325000 out of 2500000 steps (13)

[12:55:10] Project: 1034 (Run 6, Clone 36, Gen 2)
[12:55:10]
[12:55:10] Assembly optimizations on if available.
[12:55:10] Entering M.D.
[12:55:30] (Starting from checkpoint)
[12:55:30] Protein: p1034_A21nat_305_99
[12:55:30]
[12:55:30] Writing local files
[12:55:30] Completed 386865 out of 2500000 steps (15)
[12:55:30] Extra SSE boost OK.
[13:02:45] Writing local files
[13:02:47] Completed 400000 out of 2500000 steps (16)
[13:16:36] Writing local files
[13:16:38] Completed 425000 out of 2500000 steps (17)
[13:30:25] Writing local files

 

[OrganizedChaos]

are there any performance reasons to upgrade? won't it always use whatever the latest core is anyway?

 

[osage]

any performance reasons to upgrade ?

not speed wise at least, stability perhaps, but it's mostly small refinements and more options.

check the release notes for more details

 

[r0tt3n1]

There is no real performance value to the client upgrade. But, in my case, as you can see in the post above, the way flags are handled can have a pretty big difference. Using the v4 client with the -forceasm flag turned on 3DNow optimizations which slowed that cruncher down a lot. By using the -forceSSE flag (which works now for whatever reason), brought the frame times back down to the same level as when I used the 3.25 client.
Usually, a new client is mostly to fix minor bugs in previous clients, and add a few useful functions. Such as the -configonly flag that stops the client and exits after configuration, useful when someone wants to set it up as a service.

 

[Zuke]

The client simply manages which core is used and the download/upload of work units. The cores are where all the actual "work" is done. Thus - no performance changes.

 

[BCinSC]

So with a Barton, what is best/fastest command? -forcesse, -forceasm, other?

 

[RustyNale]

forcesse

 

[BCinSC]

Does 3DNow suck, or is it a bug in code? What are other options?

 

[r0tt3n1]

Not sure if the 3DNow part of the code is buggy, but man does it SUK! As you can see above, it slowed my client down a heck of a lot! I make sure to use the -forcesse flag on my AthlonXP machines. You may want to use the -advmethods flag also. It lets Stanford know that you are willing to use new cores, and recieve new work units, such as Gromacs. Us stats junkies like Gromacs work units on faster machines because the point-to-time ratio for Gromacs is worth it.

 

[RustyNale]

I use forcesse and advmethods...