Quick Question

Is it true that Stanford would prefer u not fold with two clients if u only have hyper-threading ? (vs actual dual core) If so why ?

I've never heard that. Most people here (if not all) run two instances...

actually.. stanford DOES prefer you run 1 instance per PHYSICAL CPU. In fact... here is the link to the thread on thier forums about it and i'm gonna copy-n-paste thier reasoning... http://forum.folding-community.org/viewtopic.php?t=10427

direct from pande Himself:

Pande said:

This is a direct quote from Dr. Vijay Pande:
1) If you care primarily about points, running 2 procs on HT is still the best bet. We are grateful for all contributions and if people choose to run 2 procs on HT, our approach is that all contributions are welcome.

2) If you care about the science foremost and are interested in our recommendations, then do not run 2 procs on HT, but please just run one process. That won't be best for points, but is best for the science.

3) If your machine cannot make the deadlines, then one should run the timeless WUs.

Click to expand...

With QMD's at least (none around at the moment ... since end of Jan) each wu is in a sequence ... so if you take longer to crunch 1 since your are doing 2 of them, you slow down the overall project.

QMD's may be a bit of a special case since they are very memory and cpu intensive, and so folding 2 at the same time on a p4 HT rig does not get significantly more done than folding 1 at a time.

On an intel DC (800D/900D) rig however, 2 instances are significantly faster than folding 1 at a time.

So for QMD's 2 instances were ok on an intel 800 or 900D but not on a p4 HT.

Since the other (non QMD) WU's do not botlleneck on the mem bandwidth to the same degree as QMD's, I have set all my p4 ht rigs to fold 2 instances. It is maybe 25% better throughput/elapsed time to cruch 2 at a time, than one after the other. Not perfect, but it is not a perfect world.

The purists would still say only crunch 1 at a time. But points do inspire many of us so overall a 25% increase in points would be deemed by many to be a decent tradeoff vs elapsed time and the serialised nature of the folding projects. A 5% increase in points (as with 2 qmds) would not be a good tradeoff.

Furthermore, for QMD's, I found that my p4 rigs ran cooler and could be overclocked higher with stability folding a single instance. So that also yielded the best ppd!

Now that they are back to 2 instances, i have had to back off the clocks a bit on a few rigs to remain stable when folding 2 DGromacs.

Others may have variations on my opinion, and different tradeoff points.
But this is the model that I have been operating under.

Thank u. That's exactly what i was looking for.

All the current WUs are sequential. The server generally issues only one of each run/clone/generation. When a WU is turned in, the server creates the next generation of a particular run/clone. So if you complete run 0, clone 0, gen 0 in 1 day folding one instance as opposed to 1.6 days folding two instances, you allow the creation of run 0, clone 0, generation 1 .6 days sooner. By folding one instance,the completion of all the generations of this run/clone can be completed sooner and analysis begin earlier.

Here's a dumb question for you...This DOES NOT apply to dc chips right?

not a dumb question. I would say it does not apply. There only beef with running two on HT is that the length of time gets increased substantially until the completion of a work unit. With dual core there is probably very little or no increase in completion time when running two compared to running one.

just a guess.

The original request was one instance per physical CPU but that came long before DC cpus. The dual cores are considered to have 2 physical cpus.

Good, because my dc rig is tearing through WUs like they're nothing.

But when I start folding on my xeon duallie(with ht) I should still only run 2 instances instead of 4 right?

I wouldn't feel too guilty folding 4 instances.

the only dumb question is one not asked.
ps it is ok to feel silly about asking questions
i have never met anyone who knows everthing,only people who claim to.

pps regarding this i feel silly all the time.