Page 2 of 2
Re: Low points?
Posted: Sat Oct 17, 2009 2:26 am
by uncle fuzzy
jrweiss wrote:uncle fuzzy wrote:3- I set it to 30. With a resonably fast cpu, the frame time is short enough it never writes a checkpoint.
AFAIK, it will always write a checkpoint at the designated time. If you don't have the
-verbosity 9 flag set, you may nt see it in the log.
I use verbosity. With it set to 30, it writes the step checkpoint. It does not write the time triggered
extra checkpoint, which cuts the disk write frequency to a minimum.
Code: Select all
[01:51:08] Project: 6304 (Run 708, Clone 6, Gen 11)
[01:51:08]
[01:51:08] Assembly optimizations on if available.
[01:51:08] Entering M.D.
[01:51:14] Protein: p6304_sh3_with_TYR_frags
[01:51:14]
[01:51:14] Writing local files
[01:51:14] Extra SSE boost OK.
[01:51:15] Writing local files
[01:51:15] Completed 0 out of 500000 steps (0%)
[02:06:44] Writing local files
[02:06:44] Completed 5000 out of 500000 steps (1%)
[02:21:51] Writing local files
[02:21:51] Completed 10000 out of 500000 steps (2%)
[02:37:00] Writing local files
[02:37:00] Completed 15000 out of 500000 steps (3%)
[02:52:15] Writing local files
[02:52:15] Completed 20000 out of 500000 steps (4%)
[03:07:28] Writing local files
[03:07:28] Completed 25000 out of 500000 steps (5%)
[03:22:40] Writing local files
[03:22:40] Completed 30000 out of 500000 steps (6%)
[03:37:55] Writing local files
[03:37:55] Completed 35000 out of 500000 steps (7%)
[03:53:09] Writing local files
[03:53:09] Completed 40000 out of 500000 steps (8%)
Nothing in there says checkpoint.
Code: Select all
[23:50:53] Project: 2665 (Run 1, Clone 401, Gen 122)
[23:50:53]
[23:50:59] Assembly optimizations on if available.
[23:50:59] Entering M.D.
[23:51:10] Calling FAH init
[23:51:13] ater
[23:51:13] Writing local files
[23:51:13] rom checkpoint)
[23:51:13] Read checkpoint
[23:51:14] eps (36 percent)
[23:51:14] ter
[23:51:14] Writing local files
[23:51:14] Completed 90000 out of 250000 steps (36 percent)
[23:51:27] Extra SSE boost OK.
[00:21:29] Timered checkpoint triggered.
[00:30:23] Writing local files
[00:30:23] Completed 92500 out of 250000 steps (37 percent)
[01:00:23] Timered checkpoint triggered.
[01:08:15] Writing local files
[01:08:16] Completed 95000 out of 250000 steps (38 percent)
This one did write timed checkpoints. It all depends on the client and the Project.
Re: Low points?
Posted: Sat Oct 17, 2009 4:55 am
by jrweiss
OK. It will write some kind of checkpoint at least every X minutes, where X is the time set in the config. By setting that to 30, you often save timed checkpoint writes. The step checkpoint will be written even if the timed checkpoint was just written a few seconds ago, so you may still get 2 in a row. My current units are right at 30 min/step -- some a bit more, some a bit less:
[00:52:29] Completed 212500 out of 250000 steps (85%)
[01:22:18] Writing local files
[01:22:18] Completed 215000 out of 250000 steps (86%)
[01:52:19] Timered checkpoint triggered.
[01:53:12] Writing local files
[01:53:12] Completed 217500 out of 250000 steps (87%)
[02:23:13] Timered checkpoint triggered.
[02:24:21] Writing local files
[02:24:21] Completed 220000 out of 250000 steps (88%)
[02:54:22] Timered checkpoint triggered.
[02:54:37] Writing local files
[02:54:37] Completed 222500 out of 250000 steps (89%)
[03:24:23] Writing local files
[03:24:23] Completed 225000 out of 250000 steps (90%)
[03:54:10] Writing local files
[03:54:10] Completed 227500 out of 250000 steps (91%)
Re: Low points?
Posted: Sun Oct 25, 2009 9:46 pm
by zotteken
Now, I had this SMP thing running for about 14 days I guess...
And I'm starting to longing back to the days I could just boot up the PC, no worrying about a thing, having the CPU & GPU fold, shut down, reboot & still not need to worry about lost work...
Re: Low points?
Posted: Sun Oct 25, 2009 9:53 pm
by uncle fuzzy
With increased points comes increased work.
They ain't free, ya know.
Re: Low points?
Posted: Sun Oct 25, 2009 10:35 pm
by zotteken
Yeah, I know, but it's starting to get on my nerves...
I bet I haven't even finished 7 WU in the past 14 days or so... I count 3 since 18 oktober... That's about 1 in 2 days, while I should be able to do 2 in 3 days... over 6 days that's a difference of 1... Or a difference of +/-60 over a year...
Nonetheless I have to admit some of the "issues" of the past 14 days aren't SMP related...
Things like updates that require a reboot (and just do it) messing up the WU... Or power-saving features that kick in when not wanted (I didn't even set them) & mess up the WU as well. HD's that cause fatal errors...
Yeah, maybe it's not as bad as it feels after all and also not SMP fault... I gonna give it some more time...
Re: Low points?
Posted: Mon Oct 26, 2009 12:01 am
by uncle fuzzy
The first things I do when I build a new box, or reinstall an OS, is disable all "energy saving" features (that includes in BIOS), turning off screen savers, and turning off automatic updates. You just have to remember to manually check for updates and reboot when YOU want to.
Re: Low points?
Posted: Tue Oct 27, 2009 4:50 pm
by 7im
Windows SMP is known to have an increased risk of losing a work unit when stopping and restarting the client.
Make of copy of the fah smp folder before stopping the client. Then stop it. If it doesn't restart correctly later, try copying the duplicate in over the top. This isn't a perfect fix, doesn't work every time, but it reduces the problem substantially. But again, extra work.
You could always run a Linux SMP client in a VM, more points and more stable.