Folding Forum

It's exactly the same as it doesn't change anything to the cores.

OK, but there's a radical difference in speed between SMP 5.91 and 5.92. Isn't this the same thing, or does the SMP variable change the equation?

There are more fluctuations in the SMP clients has they don't use the same MPI implementation, nor the same Fahcores.

ArVee wrote: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.

The client's don't do any folding. The Fahcores do all the folding. Upgrading a client typcially does not change folding speed. And with SMP v5.92, the client change didn't change the speed, but the new fahcore that came with the client version may have changed the speed.

Project 4555 WUs really wiz by - and 500 PPD too - whoopee!

John_Weatherman wrote:Project 4555 WUs really wiz by - and 500 PPD too - whoopee!

You must be using an AMD chip, on AMBER calculations my one AMD (which is the slowest of the three on GROMACS) takes off.

So does the CPU cooling fan. When only one core is folding, you can tell because there is a strange silence in the room.

Why can't we have these WU's when port 8080 is blocked and assignment server is assign2.stanford.edu:80 ?

At home, I have these WU's on my AMD 2000+ because assignment serveur is assign.stanford.edu:8080
But at work I only have p2170 on all clients (AMD 3000+, Celeron 2000,...) configured with -advmethods because port 8080 is blocked and assignment server is assign2.stanford.edu:80.

When I see all these WU's my AMD 2000+ calculate (minimum 2-3 per day) I could calculate more and more with clients at work.

sneakers55 wrote:

John_Weatherman wrote:Project 4555 WUs really wiz by - and 500 PPD too - whoopee!

You must be using an AMD chip, on AMBER calculations my one AMD (which is the slowest of the three on GROMACS) takes off.

So does the CPU cooling fan. When only one core is folding, you can tell because there is a strange silence in the room.

My old, tired, Athlon XP at 1600 MHz is doing 200 PPD! These Ambers love AMD, too bad that's my only AMD box. Runs rather warm at 60C, but it always has.

Even more -advmethods Amber WUs. Three more to be exact. 4558, 4559, and 4560 will go online later today.
(see http://fah-web.stanford.edu/cgi-bin/fahproject?p=4558 for points: )
--Vince
p.s. (There are three more coming too, in a bit.)

I've noticed (or at least it appears to me) that the projects seem to come out in groups of 119.30, 84.70 and 47.00 points. Is this coincidence or a function of different experiments for a larger project?

Are these projects using fewer runs or are they just getting finished sooner? So far the "furthest (?)," "oldest (?)" run I've seen on a 45xx is a run 17.

Also, I know within a group of WUs, each gen for a particular run and clone is dependent on the prior generation being finished, but is one clone dependent on a prior clone being finished?

Thanks,


Angela

anko1 wrote:I've noticed (or at least it appears to me) that the projects seem to come out in groups of 119.30, 84.70 and 47.00 points. Is this coincidence or a function of different experiments for a larger project?

These are different experiments for a larger project. We are investigating the role of local structuring in protein folding, simulating 8-, 12- and 16-residue fragments of a larger protein sequence.

anko1 wrote:Are these projects using fewer runs or are they just getting finished sooner? So far the "furthest (?)," "oldest (?)" run I've seen on a 45xx is a run 17.

We and our collaborators are participating in a "blind" assessment, in which a third party selects protein sequences, and we have a limited time to simulate them. This is why you will see projects coming online, and later going away.

anko1 wrote:Also, I know within a group of WUs, each gen for a particular run and clone is dependent on the prior generation being finished, but is one clone dependent on a prior clone being finished?

No. The clones are designed to be statistically independent simulations, the data from which can be combined when we analyze the results.

Vince

As promised, three more of these are coming online soon. Expect 4561, 4562, and 4563 by later today.
(again, see http://fah-web.stanford.edu/cgi-bin/fahproject?p=4561 for points)

Thanks for the answers to my questions, Vince.

Angela

Heads up for new AMBER -advmethods targets 4564, 4565, and 4566 going online.
(as always, see http://fah-web.stanford.edu/cgi-bin/fahproject?p=4564 for points)
--Vince

At the risk of sounding repetitive: heads up for new AMBER -advmethods targets 4567, 4568, and 4569 going online soon.
(as always, see http://fah-web.stanford.edu/cgi-bin/fahproject?p=4567 for points)
--Vince