Folding Forum

vvoelz wrote:Just a heads up that new advmethods AMBER WUs: 4552, 4553, and 4554, to go online soon (later today).
(As always, see the description page for points: http://fah-web.stanford.edu/cgi-bin/fahproject?p=4552)
--Vince

Just had a few 4553 units..........on machines with advmethods turned off.

Pete

Pette Broad wrote:Just had a few 4553 units..........on machines with advmethods turned off.

I understand this may happen sometimes if the assignment server thinks there are no other suitable WUs to send out. I *just* put the WUs online when you posted, so maybe I just "surprised" it. I'll keep an eye on it -- hopefully the assignment server isn't low on work. --Vince

vvoelz wrote:I understand this may happen sometimes if the assignment server thinks there are no other suitable WUs to send out. I *just* put the WUs online when you posted, so maybe I just "surprised" it. I'll keep an eye on it -- hopefully the assignment server isn't low on work. --Vince

One core on my X2 is running a DOUBLE GROMACS WU (not nearly as fast as it does on either C2D) and it has the -advmethods flag.

sneakers55 wrote:One core on my X2 is running a DOUBLE GROMACS WU (not nearly as fast as it does on either C2D) and it has the -advmethods flag.

I don't see a problem with this. The Amber WUs have been added to the -advmethods group temporarily, but there are still other -advmethods projects. DGROMACS WUs earn nice PPD, too, for any hardware with SSE2 - - but the C2D did raise the performance bar and AMD is still trying to catch up.

"The Amber WUs have been added to the -advmethods group temporarily,..."
Ahh, does that explain why I seem to be picking up 2170s much more often?

anko1 wrote:"The Amber WUs have been added to the -advmethods group temporarily,..."
Ahh, does that explain why I seem to be picking up 2170s much more often?

Actually, no. P2170 is not one of those bonus -adv Amber projects. You picked up a 2170 because it was best suited for your client at the time it needed a WU. Either because the bonus 45xx Ambers were all out, or because 2170s needed to be done sooner, or for any of several other reasons. The work assignment logic is both fluid and complex.

Thanks for elucidating.

Another heads up for new -advmethods AMBER WUs: 4555, 4556, and 4557, to go online later today.
(see for points: http://fah-web.stanford.edu/cgi-bin/fahproject?p=4555)
--Vince

Just a comment... something has changed... concern?

Fahlog from Project: 4554 (Run 111, Clone 28, Gen 4)

[23:37:47] Completed 0 out of 2500000 steps (0%)
[23:46:43] Completed 25000 out of 2500000 steps (1%)
[23:55:37] Completed 50000 out of 2500000 steps (2%)

Fahlog from previous 45xx Ambers:

Project: 4553 (Run 50, Clone 15, Gen 6)
[16:57:31] Completed 0 out of 2500000 steps (0) <---- No %
[17:03:26] Completed 25000 out of 2500000 steps (1)
[17:09:26] Completed 50000 out of 2500000 steps (2)

Project: 4552 (Run 5, Clone 41, Gen 0)
[06:47:40] Completed 0 out of 2500000 steps (0) <---- No %
[07:02:21] Completed 25000 out of 2500000 steps (1)
[07:18:03] Completed 50000 out of 2500000 steps (2)

7im --
That's odd. I'll ask around to see if there have been any changes in the AMBER client that I wasn't aware of
--VInce

7im wrote:Just a comment... something has changed... concern?

Did the version of AMBER change between those assignments?

I've never gotten the "%" sign, looking back through my logs of the Ambers. 4552 54/11/3 seems to be an extreme outlier in its rate of fold though. Perhaps 2.5X the average? Assembly optimizations checked and on (forced manually, actually).

I've asked around in the lab, and there have been no recent changes to the AMBER core, for what it's worth.

I saw the % sign after upgrading to the last v6 beta client ... I didn't have it with older versions.

How do the v6 versions compare to the 5.04 version speedwise? I'm asking specifically with regard to the new Amber units but observations in general are welcome too.