Howdy all.
I just recently re-started folding. I have a few machines folding currently, and plan on borging more here at the office. Here at my office I have users who like to keep their machines on all night long, even after I "ask" them not too, might as well make use of the electricity for something! I say I "ask" them not too, because I'm the IT/IS Manager here, but try not to be the "IT Nazi" and force them to do things. So, I can borg if I want too, and not get into trouble for doing it to other's machines.
But, have a couple questions.
I have a hyperthreaded p4-2.4. I've done some initial testing and have observed a few things. First thing, boy, it took a while to get the -local to work. Machine ID 1 always received an "old core" and would just hang, never doing anything. MID 2, it worked right away. If I take off the -local and try to run only MID 1, it still only used 50% cpu according to Task-Manager. Is the F@H client CPU aware, even without the -local switch?
So, with -local added in, and both MID 1 and 2 set to use 100% of idle time, Task Manger shows each using 50%. Switching to the Performance tab does show that each "cpu" is working 100%, so I suppose that is working correctly.
But, I have a question about the HT units working.
Log section from MID 1:
[15:56:06] Folding@Home Gromacs Core
[15:56:06] Version 1.81 (April 6, 2005)
[15:56:06]
[15:56:06] Preparing to commence simulation
[15:56:06] - Looking at optimizations...
[15:56:07] - Created dyn
[15:56:07] - Files status OK
[15:56:07] - Expanded 301543 -> 1815257 (decompressed 601.9 percent)
[15:56:07] - Starting from initial work packet
[15:56:07]
[15:56:07] Project: 1416 (Run 77, Clone 54, Gen 0)
[15:56:07]
[15:56:07] Assembly optimizations on if available.
[15:56:07] Entering M.D.
[15:56:14] Protein: p1416_polyQ_beta-helix
Log Section from MID 2:
[15:30:04] Folding@Home Gromacs Core
[15:30:04] Version 1.81 (April 6, 2005)
[15:30:04]
[15:30:04] Preparing to commence simulation
[15:30:04] - Looking at optimizations...
[15:30:04] - Created dyn
[15:30:04] - Files status OK
[15:30:04] - Expanded 301826 -> 1815257 (decompressed 601.4 percent)
[15:30:04] - Starting from initial work packet
[15:30:04]
[15:30:04] Project: 1416 (Run 34, Clone 53, Gen 0)
[15:30:04]
[15:30:04] Assembly optimizations on if available.
[15:30:04] Entering M.D.
[15:30:11] Protein: p1416_polyQ_beta-helix
Now, both "cpu's" are working on the same project number, but diferent "Run" and "Clone". Have I been wasting cycles on the same fold? Or is this really diferent. I couldn't find any concice answers to this. Accoridng to EM-3 this is a 243 point protein, so it really could be a huge waste of time if it is in fact only going to count once. But, if it is going to count twice, that's good.
On to another question... Has anyone established a "database" of sorts, or at least an idea of how long some of these proteins should take to fold, on what size of CPU?
I ask this because here at my desk, I'm setting up a new machine, a Celeron 2.5G. Right now, it is working on a 48.0 point protein, and each "frame" is taking nearly 15 minutes (according to EM-3). BUT, the prevously mentioned P4-2.4 working on the 243 pointer, is taking close to an hour per "frame".
So, is there a place I can go to get an idea on how long certain point proteins should take on diferent speed machines? I know, there are lots of other variables, but still there has to be some general ideas as to how long some should take.
I think that is it for now.
I just recently re-started folding. I have a few machines folding currently, and plan on borging more here at the office. Here at my office I have users who like to keep their machines on all night long, even after I "ask" them not too, might as well make use of the electricity for something! I say I "ask" them not too, because I'm the IT/IS Manager here, but try not to be the "IT Nazi" and force them to do things. So, I can borg if I want too, and not get into trouble for doing it to other's machines.
But, have a couple questions.
I have a hyperthreaded p4-2.4. I've done some initial testing and have observed a few things. First thing, boy, it took a while to get the -local to work. Machine ID 1 always received an "old core" and would just hang, never doing anything. MID 2, it worked right away. If I take off the -local and try to run only MID 1, it still only used 50% cpu according to Task-Manager. Is the F@H client CPU aware, even without the -local switch?
So, with -local added in, and both MID 1 and 2 set to use 100% of idle time, Task Manger shows each using 50%. Switching to the Performance tab does show that each "cpu" is working 100%, so I suppose that is working correctly.
But, I have a question about the HT units working.
Log section from MID 1:
[15:56:06] Folding@Home Gromacs Core
[15:56:06] Version 1.81 (April 6, 2005)
[15:56:06]
[15:56:06] Preparing to commence simulation
[15:56:06] - Looking at optimizations...
[15:56:07] - Created dyn
[15:56:07] - Files status OK
[15:56:07] - Expanded 301543 -> 1815257 (decompressed 601.9 percent)
[15:56:07] - Starting from initial work packet
[15:56:07]
[15:56:07] Project: 1416 (Run 77, Clone 54, Gen 0)
[15:56:07]
[15:56:07] Assembly optimizations on if available.
[15:56:07] Entering M.D.
[15:56:14] Protein: p1416_polyQ_beta-helix
Log Section from MID 2:
[15:30:04] Folding@Home Gromacs Core
[15:30:04] Version 1.81 (April 6, 2005)
[15:30:04]
[15:30:04] Preparing to commence simulation
[15:30:04] - Looking at optimizations...
[15:30:04] - Created dyn
[15:30:04] - Files status OK
[15:30:04] - Expanded 301826 -> 1815257 (decompressed 601.4 percent)
[15:30:04] - Starting from initial work packet
[15:30:04]
[15:30:04] Project: 1416 (Run 34, Clone 53, Gen 0)
[15:30:04]
[15:30:04] Assembly optimizations on if available.
[15:30:04] Entering M.D.
[15:30:11] Protein: p1416_polyQ_beta-helix
Now, both "cpu's" are working on the same project number, but diferent "Run" and "Clone". Have I been wasting cycles on the same fold? Or is this really diferent. I couldn't find any concice answers to this. Accoridng to EM-3 this is a 243 point protein, so it really could be a huge waste of time if it is in fact only going to count once. But, if it is going to count twice, that's good.
On to another question... Has anyone established a "database" of sorts, or at least an idea of how long some of these proteins should take to fold, on what size of CPU?
I ask this because here at my desk, I'm setting up a new machine, a Celeron 2.5G. Right now, it is working on a 48.0 point protein, and each "frame" is taking nearly 15 minutes (according to EM-3). BUT, the prevously mentioned P4-2.4 working on the 243 pointer, is taking close to an hour per "frame".
So, is there a place I can go to get an idea on how long certain point proteins should take on diferent speed machines? I know, there are lots of other variables, but still there has to be some general ideas as to how long some should take.
I think that is it for now.





