How many points should I be getting?
Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee
How many points should I be getting?
Using the config in my sig how many points should I be getting per week? Right now it's around 230 PPW. I was just reading a poster who was getting better numbers with a much slower pc. I was just wondering if these numbers were in line with other ppl's points? Thanks
You should probably be seeing ~5-600ppw with that.
My bet is that FAH is not using SSE code, which helps immensely. With the AMD you'll need to force FAH to use SSE by adding a config flag on the command line of the shortcut.
If you're using Version 3.x, the flag is -forceasm
If you're using Version 4.x, the flag is -forceSSE
My bet is that FAH is not using SSE code, which helps immensely. With the AMD you'll need to force FAH to use SSE by adding a config flag on the command line of the shortcut.
If you're using Version 3.x, the flag is -forceasm
If you're using Version 4.x, the flag is -forceSSE
I use console FAH client so I just specify the argument -forceSSE. I think you want the space between WinFah.exe and -forceSSE.
I'm not sure about your question about higher point WUs. WUs have different point value which is proportional to the "difficulty" of the protein and is proportional to the amount of time you need to spend to crunch it.
It has been found that SSE instruction set gives a significant speed advantage and Tinker WUs which can not support SSE get processed significantly slower than Gromacs WUs that support SSE.
I think if you specify -advmethods flag you will be getting almost exclusively Gromacs.
Between Gromacs WUs there seem to be 2 kinds, one kind gets done quickly and another kind gets done about 30% slower on the same machine (point wise). There is no way to select which kind you want out of these two. The positive thought is that everyone is subject to the same distribution in the long run so it is fair to everyone.
Hope this helps.
I'm not sure about your question about higher point WUs. WUs have different point value which is proportional to the "difficulty" of the protein and is proportional to the amount of time you need to spend to crunch it.
It has been found that SSE instruction set gives a significant speed advantage and Tinker WUs which can not support SSE get processed significantly slower than Gromacs WUs that support SSE.
I think if you specify -advmethods flag you will be getting almost exclusively Gromacs.
Between Gromacs WUs there seem to be 2 kinds, one kind gets done quickly and another kind gets done about 30% slower on the same machine (point wise). There is no way to select which kind you want out of these two. The positive thought is that everyone is subject to the same distribution in the long run so it is fair to everyone.
Hope this helps.
Are you sure about getting ~500 per processor? I've crrunching on 3x2GHz machines for a while and only getting the same type of results BCMoney is getting (~700ppw). And I've been using -forceasm.
From my logs:
From my logs:
[18:18:35] - Calling 'FahCore_78.exe -dir work/ -suffix 08 -service -forceasm -lifeline 1276
...
[18:18:35] Folding@home Gromacs Core
[18:18:35] Version 1.54 (November 26, 2003)
...
[18:18:52] - Assembly optimizations manually forced on.
...
[18:18:52] Project: 850 (Run 8, Clone 25, Gen 19)
[18:18:52]
[18:18:52] Assembly optimizations on if available.
...
[18:19:13] Extra SSE boost OK.
My Barton 2800+ @2.4 Ghz gets about 750 PPW on a good Gromacs core. On other Gromacs cores it might get 500-600 PPW. On Tinkers that number goes way down, this is why you might want to use -advmethods.
If you get 250 PPW from 2 Ghz machine then something is really wrong with your config. Hey, my 2.0 GHz Duron consistently gets above 500 PPW and it's got very little L2 cache.
Try to launch the client with -forceasm -forceSSE -advmethods, make sure the core you are processing is Gromacs and you should get significantly more than 250 PPW.
If you get 250 PPW from 2 Ghz machine then something is really wrong with your config. Hey, my 2.0 GHz Duron consistently gets above 500 PPW and it's got very little L2 cache.
Try to launch the client with -forceasm -forceSSE -advmethods, make sure the core you are processing is Gromacs and you should get significantly more than 250 PPW.