MacOS Intel S5R2/S5R3 App 4.42 available for Beta Test
Bernd Machenschalk
Moderator
Administrator
Joined: 15 Oct 04
Posts: 4312
Credit: 250193236
RAC: 35038
5 Sep 2007 15:48:37 UTC
Topic 193106
(moderation:
A new MacOS Intel App is available from our Beta Test page.
Probably non-Windows users won't notice any difference between this App and the current official one. We just a added a feature that we'll need for S5R3. As this App can still process S5R2 Tasks, we publish it in advance for people to test before S5R3 starts.
BM
BM
MacOS Intel S5R2/S5R3 App 4.42 available for Beta Test
)
Just switched one of my Core2Duos onto 4.42, no problems so far, first results should be in early on Saturday
Running 2 tasks of 4,42 at
)
Running 2 tasks of 4,42 at the same time on my MAC seems to slows down the app.
Edit
It can also be that the new app itself is slower:( report comming saturday.
First two results now
)
First two results now crunched successfully (partly with 4.38) and returned. One has validated OK, the other still pending. The app does seem to be slightly slower, now taking ~101,400 secs to crunch a 638.33 credit result, up from ~98,700
I concur with the slightly
)
I concur with the slightly slower reports. My mac mini was running 657 credits in about 151,000 cpu seconds. It went up to around 156,000 on 657 credit WUs. And that was with around 20% of the WU completed by the previous APP.
RE: Running 2 tasks of 4,42
)
Hmm sunday by now .
1 or 2 tasks at the same time does not matter.
The new app is about 2% slower on my IMAC.
My 1.66Ghz Core Duo mac mini
)
My 1.66Ghz Core Duo mac mini has 1 WU that just validated OK. 3 others finished ok but have not completed validation yet.
RE: My 1.66Ghz Core Duo mac
)
Hi!
I saw the debugging output of your mac mini in the results pages, what strikes me is that it seems to checkpoint rather frequently.
Is it possible that you changes the setting "Write to disk at most every xx seconds" to a very low value (60 seconds is the default and indeed a good choice).
You should get better crunching performance when you raise that value to something near 60 seconds.
CU
Bikeman
I will check that setting. I
)
I will check that setting. I think I may have changed it. Thanks for the heads up.
I had the write to disk set very low like suspected. It was .01, I set it to 60 like suggested.
After I did this I looked at my CPU temp and my hard drive temp (on my G5 imac, not the mini). Funny, the CPU has gone down about 3 degrees C and my hard drive has gone up about 3 degrees. I would have figured things would have gone the other way.
RE: I will check that
)
I guess there's no simple connection between temps and load, as the imac might have a variable fan control (I know that the mac mini has this). It could be like a saw-tooth (?) curve where temps rise with load until a given point, then the fan is switched "one gear up" and temps are going down. 3 degrees isn't much and could also be influenced by ambient temps.
Anyway the change will only become effective after the next "update" operation when the client has contact with the server and the client synchronizes it's preferences settings. It could even be that this takes effect only after the science client has been stopped once. You'll notice the change by looking into the appropriate slots//stderr.txt file . It should now contain more than one consecutive number per line.
CU
Bikeman
RE: I will check that
)
BTW, I've just noticed that you own one of the TOP 3 most productive computers (in terms of RAC) on Einstein@Home, congratulations. It will be interesting to see how performance is influenced by the chnage of preference settings.
Keep up the good crunching!
H-BE