Folding@Home: Interesting statistics

Following a link from John Childs brings you to this page:

http://folding.extremeoverclocking.com/team_overview.php?TeamID=13392

…which has quite a lot of funky info, such as predictions for when we’re going to overtake other teams. Team Milkdud is down for June 7th!

However it also appears that our work unit throughput has fallen off quite a bit just recently… what gives?

Phil

Well, this is probably just me, but incidences of random computer death have increased quite a bit recently, and stopping folding@home seems to help. Currently running but only using 70% of the cpu, and it seems to be working.

Drive-bay fans will be installed shortly…

John

Aha! So it is entirely your fault that the DynaDudes are shortly to overtake us. I should have known… <sneer>

Phil

They recently updated the core and started adding more projects that use the GROMACS core rather than the TINKER core. GROMACS is more optimized than the TINKER core. GROMACS makes aggressive use of the SSE, 3DNOW, and other special CPU features. The GROMACS code causes the CPU to run a little bit hotter than TINKER.

Make sure your CPU fan is working and make sure there is no dust in the heat sink. Check the CPU temp to see if you’re crashing due to an overheated CPU.

I haven’t had any problems with the new core. Everything has been solid. I have only had one crash in all the months I’ve been running the client. In that case it only crashed the client and the OS kept chuging along (Windows 2000). It’s been very solid for me. I have an Athlon 1200 and run at about 96% CPU utilization.

There have been some discussions on the folding community forums about stability problems with the new core. Some CPUs are having problems, specifically some of the new Athlons that support SSE2. Check the forums here: <http://forum.folding-community.org/&gt; in the bug reports secions.

What CPU are you running?

Its an Athlon 1600, I think. Not sure about the 1600 bit. It’s running a bit warm, hence the thought about getting an extra fan. All the current fans and heatsinks are all fine.

But what’s really annoying me about f@h is the way that I’ve done more WU’s than Phil, but he’s still 120 points ahead of me! I reckon its that bloomin project 632, but there’s nothing I can do about that. Grrr.

John

Something else to check is your power supply. I was getting some strange random shut downs. And occasionally when I powered on the computer the BIOS would complain about being unable to initialize the CPU and plop me at the BIOS setup screen – A hard reset would get the computer booting again.

I finally caught a clue and suspected the power supply was slowly dieing on me. I replaced the power supply and all those odd problems went away.

The high temperature over the past few days definitely doesn’t help!

I found a program called DTemp which monitors the temperature of new-ish hard drives, which is quite useful.

The second thing I noticed was that it’s not so much the computer pumping out loads of heat but the monitor… I could make my dinner on the top of the monitor at the back; a thermometer placed on top was just shy of 50 degrees C!

Ah well; when I am rich and famous I will get myself a shiny new flat panel monitor… :slight_smile:

Well, just look at that pesky John Childs character, he’s in front with even fewer units. Tch… :slight_smile:

Phil

Obviously my Athlon 1200 is superior to your lousy processor. :slight_smile:

I’m just getting lucky with getting assigned more high point work units than low point work units. I also have my client set to prefer Folding@Home work units over Genome@Home work units. Folding work units tend to be higher point values than the Genome units. I like the science in the Folding project better than the Genome project so that’s why I made that preference in the client.

But the point values for the individual work units shouldn’t matter too much in your final score. The point values for a work unit are based on your CPU time that you donated. Work units that finish faster are worth fewer points than work units that take longer. The work unit point values are based on benchmarking that work unit on a 500 MHz Celeron. A work unit that takes that Celeron 1 day to complete will be worth X points. A work unit that takes that Celeron 2 days to complete will be worth 2X points. It all should even out in the end. Here is a description of that calculation in the Folding@Home FAQ

But since the benchmark Celeron does not have SSE or 3DNOW support the points calculation formula does get a little off when you’re running on an Athlon or Pentium IV that supports SSE or 3DNOW. If you’re seeing a discrepancy between the point values for different work units it is likely due to some of the work units making better use of the SSE and 3DNOW optimizations than other work units and the benchmark Celeron doesn’t take that in to account.