Posted on 12/18/2005 10:03:55 AM PST by texas booster
It's difficult to say really. Some of the WUs are really big and depending on whether your processor supports SSE or not can really affect how long it takes to get a WU done. I know I went a couple days on my first few WUs on all the machines before I started completing faster. It's luck of the draw I suppose. One of my complaints that I posted on the main forum was WU completion time. It would be nice to have progress checkable faster, like RC5-64 or SETI units, etc. The very first WU on one of my two fastest machines (4000+ w2gb ram) took almost 4 days to complete. By the end of that first 7 days, it had done 12.
As to the point figures, I do know enabling large WUs will earn more points, but it's self defeating unless the machine has a pile a ram.
Of course, if they're on the same WUs and doing it slow, I don't know what to say, I've not seen that, just see fast and seen slow and REALLY slow WUs. One user said that something about aborted WUs(out of bounds, bad data, something) that can shut off SSE computation and make it much slower. A reboot and/or a restart of the client will fix it.
However, what do you mean by your "work has slowed to a crawl"? Are the machines running slower or do the WUs seem to be taking longer to complete?
In my experience so far, you get WUs that F@H thinks your machine can handle within the deadline, unless you've chosen to get WUs without a deadline. I suspect that Stanford's F@H project just doles out WUs as available. It matters most that the WU is completed within the deadline. The point weighting factor should consider your machine's capability plus the complexity of the WU that was processed. I pulled one WU down today that's going to take nearly 4 days to process; hopefully we'll get some good point from it.
Wow, all the latest folks have REALLY cranked up the pressure. I think QUT_HPC isn't much of a threat at all :). We're already above their daily output and as a few more folks come online, we're gonna leave 'em in the dust!
My first WU on 2 computers was 3750 frames. One computer finished it early this morning, about 36 hours after starting it. The next WU it downloaded was 500 frames. The 3750 frame units were finishing at about 28s/frame. The 500 frame unit is working at 1m30s/frame. Is the protein being "folded" larger, more complex to account for such a slower time?
Lookin' good!! I have 2 WUs finishing within an hour... with all the others that have signed up, the FR team should start overtaking at an even faster rate.
Free Republic Folders |
Date of last work unit | 2005-12-20 19:08:15 |
Active CPUs within 50 days | 84 |
Team Id | 36120 |
Grand Score | 145873 (certificate) |
Work Unit Count | 1115 (certificate) |
Team Ranking (incl. aggregate) | 1372 of 41779 |
Home Page | http://www.freerepublic.com |
Fast Teampage URL | http://fah-web.stanford.edu/teamstats/team36120.html |
Team members |
Don't take my word for it, but just realize this is what I've read. To fold ONE protein would require a single, new, fast computer 30 years. The units per depend on how large the team scientifically determines what makes scientific sense.
I dunno what that means. I do know that at this point, some are alot, some aren't so much. In the end, it may be one of us FREEPERS that solves alhzheimers. I think that's fantastic.
My first WU had 500 frames, and it took about 2 days to complete. The second one had 400 frames. The frames themselves took much faster, and it finished the project in a few hours. Now, I have a 5000 frame WU, and it's not expected to finish until April!
re. the WU that won't finish until April... had it just started processing when it gave that date? If so, I've found that the WU needs to crunch for a while before the time estimates become accurate. Once a sufficient number of frames have been processed, things start to converge and then you can start believing what it says.
Now that it's at 87/5000 (est.) it now looks like it's going to take about 6 days. It seems like it estimates 30 minutes per frame first. For my PIII running on the TINKER core, that is about correct, for my P4, that's a gross overestimate. I'm going to pull a K6-2 out of the attic later and see how that does on the timeless rounds. That one might have a WU that takes three months.
I tried a K6-2 (500MHz) last week... it was agonizingly slow, plus the graphics card is so old that the display from the FAH client was virtually unreadable. I ended up pausing the WU and transferred it to another machine.
Over 6,000 points in the last 24 hours!!
With luck we'll put them, ahem, safely behind us by 10 minutes after the hour when the stats update.
We have 90 computers folding for Team FreeRepublic!
Yeah, pretty cool! I'm holding reserve yet, when we hit a wall, I've got more in me yet :). That said, Freerepublic needs to get more folks on board. 90 of 100,000 users? Pathetic! C'mon all yee freepers, let's make FR soar!
It'd be nice if we talk to the top few that now have zero for 50 days...why did they stop?
-Mal
Ken, I've got a couple old GFX cards, Geforce 2mx, geforce 4400 and I think a TI4200 laying around. If you got an AGP slot, I'd send you a card if you want(no cost, any address you want). Shrug. It's junk to me, will hit the dumpster when I clean out my computer cart. Freepmail me if you're interested.
Disclaimer: Opinions posted on Free Republic are those of the individual posters and do not necessarily represent the opinion of Free Republic or its management. All materials posted herein are protected by copyright law and the exemption for fair use of copyrighted works.