Page 1 of 1

Project: 2170 (Run 29, Clone 235, Gen 4)

Posted: Mon Jun 02, 2008 12:07 pm
by bgd73
Project: 2170 (Run 29, Clone 235, Gen 4)

Code: Select all

[02:24:27] Completed 60000 out of 500000 steps  (12)
[02:27:56] Writing checkpoint files
[02:33:03] Writing checkpoint files
[02:38:09] Writing checkpoint files
[02:43:16] Writing checkpoint files
[02:48:22] Writing checkpoint files
[02:53:28] Writing checkpoint files
[02:58:33] Writing checkpoint files
[03:03:38] Writing checkpoint files
[03:08:46] Writing checkpoint files
[03:14:04] Writing checkpoint files
[03:18:25] Writing local files
[03:18:25] Completed 65000 out of 500000 steps  (13)
I am getting to know pc once again with these work unit trials...

I did learn I had to do some things (like setting priority) manually with task manager, making machine ids, and running 3 clients to stop the gpu one from taking over :roll:

now I am assuming my cpu is not setup for the title of this thread (p4 2.8e- that is the 1mb cache ht 800 etc etc)
but upon real calc of time to complete, after doing what above shows complete is....
31 days.
31 days, 24 hours a day at full throttle to complete this project. No doubt an error given the points. I went through this on a project "2150" already for less than 100 points for days and days of runtime. (weeks in fact)
Should I be chucking my pc in the dumpster? Or can someone inform me that these times to complete are "normal". I am hoping to find errors be honest.

Re: Project: 2170 (Run 29, Clone 235, Gen 4)

Posted: Mon Jun 02, 2008 12:38 pm
by bapriebe
bgd73 wrote:

Code: Select all

[02:24:27] Completed 60000 out of 500000 steps  (12)
[03:18:25] Completed 65000 out of 500000 steps  (13)
It looks to me like it's completing 1% every 54 minutes. That works out to 90 hours to completion. My 3Ghz Conroe-based box is getting a lot of these 2170 AMBER WU's too and churns through one in 38 hours. I agree the 234 points is pretty pathetic for that much run-time :).

Re: Project: 2170 (Run 29, Clone 235, Gen 4)

Posted: Mon Jun 02, 2008 5:35 pm
by 7im
bgd73 wrote:Project: 2170 (Run 29, Clone 235, Gen 4)

I did learn I had to do some things (like setting priority) manually with task manager, making machine ids, and running 3 clients to stop the gpu one from taking over :roll:

Should I be chucking my pc in the dumpster? Or can someone inform me that these times to complete are "normal". I am hoping to find errors be honest.
If I understand you correctly, you are running 3 FAH clients on a single CPU. That would be the first error. The project recommends running one FAH client per CPU core. Running 3 clients will cause the completion times to be 3 times longer than normal.

I don't know how well the new GPU2 client works with HT, but with the GPU1 client, you could run 1 CPU client on the CPU, and feed the GPU client with the HT, and both would perform reasonably well. GPU2 may have changed that.

So, go back to running 1 or 2 clients, and play with the client Priority settings in the FAH setup, and your frame times should be more what is expected. If the GPU2 is taking over, set the CPU client to the "low" priority setting, and leave the GPU at "idle."