BE Folding@Home Team Info

[RotorHead]

That seems a little high for a P4 even with HT enabled.

Here is my protien data generated from EMIII, there are a couple of 220's in there to compare with.

http://members.cox.net/ebecquet/protein_data.htm

Can you shut down one of the clients and let the other one run for a couple of frames, then see what the estimated time is.

[NeoShowtime]

After going away for a while and coming back the times went to 27 and 28 hours. How can I check and make sure SSE is being used? And ard are there any other things I should be concerned about performance wise?

[RotorHead]

The log file in the folding directory will have a line entry if SSE is working correctly.

[09:20:33] Extra SSE boost OK

The following info will help a little with performance.

If you are starting the client from EMIII, open the options screen, put a check in the “use extra launch commands” box, then type in the following switches: -advmethods -forceasm -forceSSE. After that, just shut down the clients and EMIII, then restart EMIII and you’ll be set.

If you are starting the clients using the executable, then just create a short cut, right click it, go to properties, short cut tab, and in the target box add the same command line switches listed above. Then just use the short cut to start the client and your golden.

Hope that helps

[Duvie]

So how does the scoring go, and why am I puling all 400 frame WU's...At this rate it will take a wekk to get 10 WU's done....Still haven't ran in HT mode yet as I am finishing some testing and doing some other work....

A did a 220 protein and got a score of 30...What does that mean???

How would a a 393 protein score since it is also 400 frames but appears to be almost double the time per frame???


Just trying to understand the process a bit.....

[Duvie]

Originally posted by RotorHead
That seems a little high for a P4 even with HT enabled.

Here is my protien data generated from EMIII, there are a couple of 220's in there to compare with.

http://members.cox.net/ebecquet/protein_data.htm

Can you shut down one of the clients and let the other one run for a couple of frames, then see what the estimated time is.

Hey I saw the stats and it appears number of atoms and obvious frame count is important to completing work units....How come I am drawing all 400 frame work units and you have tons of 100 frame WU's....Is it luck of the draw???

Patterns are hard to draw here as 1000 atom units though still 100frames had quicker times per frame then say 202atoms...How is that??

Is it just not comparable across WU's???

[pointreyes]

Originally posted by Duvie
Is it just not comparable across WU's???

AFAIK, no.

However, it does make logical sense that the WU would not be comparable to another. One WU could be a completely different protein from another and hence requires more or less work to process it.

The reason why you are pulling up 400 frames WU is due to the hardware you have. The client software trys to locate the best WU possible to use based on the quality of the hardware being used. Imagine passing that 400 frames WU unto a p200 machine.

The quantity of WU's completed does not denote the points gained. The points are based on the level of complexity to complete the WU.

[SupermanCK]

yeah...
right now i am running protein 393 on the p3 cel666mhz and it said the remaining is a lil over 100hrs...lol

[Duvie]

I have the 393 protein running on my Athlon 2100+ machine w/ forcesse....

It looks like it is rolling 6:08 sec frame times....

My p4 is rolling the same 393 protein and getting more like 4:50 frame time...that is with 1 instance....

I have the second sarted but can't seem to get them to run at same time...I think I have it set up wrong...they are both in the EMIII setup as HT#a and HT#b but I can't get more the one FH.exe running and thus only one protein is running....


Fixing up my work PC right now at 1.4ghz to run some....

[SupermanCK]

393 on a 666cely take 24 minutes per frame...:mad:

[Duvie]

Hey how imprtant is the setting in the FH.exe about preference to FH or genome in getting WU's??? I hear some ppl don't want to genome ones...is it because they don't like the 400 frame or hug atom count work units???

[RotorHead]

Duvie, are you using the (-advmethods) switch? This switch will force the assignment server to get the work units from the servers that have newer proteins.

You can check which servers are passing out the new proteins here.

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

If the servers have (adv) or (bet) listed in the PRO column, then it’s handing out the good stuff.

A quick way to check if you have a newer work unit is, bring up EMIII and hover your mouse over the little computer icon next to the computer name. If it is blue you will be crunching a “Gromacs” which are faster. If it’s green you’re crunching a “Tinker” which are slower.

[Duvie]

I have seen the tinker command in the fah logs...So I think it is those ones....

I do indeed have the advmethods checked and should be running in EMIII if not the FH.exe shortcut where I placed it there as well....


Bad news!!! I decided not to run the work PC....I am more outin the field of recent and haven't beenpaying attention to this cpu but IMO I think this cpu is running way too hot now with a flaky case fan....I would be worried about leaving this athlon chip get too hot on an ECS mobo over the weekend...

Oh well...i need to get in here and spend some time cleaning it up and fixing the F-ups the other ppl who occasionally use this have created.


I will have 3 then...3.5ghz (sometimes with HT)...2.08ghz oc'd athlon....1.5ghz centrino which I think does well...Does anybody know this for sure...I think I did some testing back in May and it did the WU pretty fast even compared to my 2.4b running stock....Does it like the large 1mb l2 cache of the centrino???

[RotorHead]

I updated the main post with some configuration information.

Duvie, if you have all the switches set, then you should pick up a Gromacs once the current work unit is finished. Sometimes you'll still get a Tinker if the servers with the Gromacs are busy, down, or have run out of work units.

Also, to run two instances, you will need to install a new client in another directory, then during the configuration, answer Yes to the “Change advanced options (yes/no) [no]?" then use 2 for the “Machine ID (1-8) [1]?”. If it still gives you problems then include the –local switch with each instance of the client.

[RotorHead]

Oh ya…….almost forgot.

For the most part the F@H client responds best to pure Mhz, but you will get a performance boost from a larger cache. I can tell you that, back in the day, my PIII-s would stomp the crap out of the older P4a chips. At 1.26Ghz I could easily out fold the 1.6 and 1.8 P4a’s.

The AMD chips are very fast folders, the client really likes the strong FPU, though the newer client has narrowed the gap a little with SSE optimizations.

As far as your centrino goes, it should perform well, but I haven’t seen a lot of data on them.

[Duvie]

Early results is.....Gonna take forever!!!! 4th downloaded WU out of all my machines and yet again another 400 frame tinker core (still haven't got any gromacs) same exact wu as my HT#1 which does a frame in 5min this is seeming to do it in 10 min....My estimate will be 60 hours!!! Yikes...

I hope I get big scores for running these 400frame wu's otherwise I don't see how I catch guys running 100 and 200frames wu's....