Genome@Home + HTT

[blackhex]

I know some of you may be aware of this and some may already know how to fix it, but I'm not one of them With HTT on, only 50% of my CPU gets used for G@H... any way around this? Or should I turn HTT off?

 

[Mattman]

I think you're just seeing the OS interpreting the use of one half of the HT processor to 100% and everything's OK. I'm assuming the same thing holds true with Genome as it does with Folding on an HT processor...if you run a second instance (which will make your usage 100% incidentally) you will get a few percent more points than you would running just one instance.

 

[GotNoRice]

http://www.hardforum.com/showthread.php?s=&threadid=712785

You should find a lot of useful information in that thread.

 

[pduan87]

You might want to look into Folding@Home. Genome@home is basically a dead project, Stanford is currently supporting Folding only.

FOLD ON!!!

 

[gnewbury]

Originally posted by pduan87
You might want to look into Folding@Home. Genome@home is basically a dead project, Stanford is currently supporting Folding only.

FOLD ON!!!

Genome is not dead, but is sleeping peacefully. They are still collecting some data and if you go to the official board there is some chatter going on. Remember - if one does not enter one of the old (see sig) team numbers all genome work is gathered as project 799. Checking your stats - pduan87 - you will see that you have completed 9 of p799. I've turned in 63, and MikeMcHugh has done almost 600.

Except for the discrepancy in points per ghz per day compared to Gromacs it's a very valid project. There are some good data points gathered:

I wish I could confirm that all is equally scored on a Celeron 500, but it doesn't appear to be so here. I also have exactly that same hardware, with 128K DRAM. The only thing possibly different is that I'm running Linux. I've recorded the score rate in points/hour as reported by qd for roughly the last fifty units I've done. Here, in general terms, is what I've seen.
Gromacs projects 53x produce about 0.280 points/hour.

Gromacs projects 51x produce about 0.215 points/hour.

Gromacs projects 39x produce about 0.220 points/hour, but are variable.

Tinker projects 62x produce about 0.180 points/hour.

Tinker projects 20x produce about 0.120 points/hour.

SPG project 799 probably produces about 0.260 points/hour now. My latest data is from before when the formula was doubled on 6 February 2003.
So even on a machine without the accelerated instruction sets, Gromacs scores considerably better than Tinker. Some of the other variations might be due to cache performance differences with the different internal Gromacs methods.

Some of these project series, mainly the Gromacs ones, may be primarily beta projects, so it's possible that when corresponding work is released as new projects for the general community, not only will it run with fewer errors, but the scores will line up a little better too.
_________________
-Dick Howell (rph_iv)

over here.

Note I think that the actual genes produce a little less than tinkers.

But if your like me there are advantages to Genome - it is easily sneakernetted, you can download 8 units at a time, there is not time limit on turn in. I've a few duallies I don't see very often and the closest they get to the net is being hooked to a printer. They chew about 1 big gene every 2 days per CPU.
So I can either load a couple of random Gromacs or Tinker proteins on them, or 16 genes. The proteins may not get turned in before the due date. I recently got a batch of p547's and p548's - those rot after 5 days. Not good for science.
But 8 genes per CPU is makes use of the idle cycles and are like rice cakes. Chewy and tasteless, but they never go bad.
They are also good for my old Cely 300's and I like to throw one on my other duallies just so when they run out of real work they still contribute,