New WUs quite large?

[Thor]

I'm just a little worried:

I have a workunit that uses around 175MB RAM ans after 40min of running it, I have only 1.04% done:confused: The last days it took only around 2:50Hours to get one unit done...

This one is called CAS_HOMOLOG_ABRELAX_hom001_t287__506_19018_0

anyone else noticing something similar?


Florian

[rshepard]

I have one running at 1.044% after 2hr 16 minutes. I also have a completed one that ran to 2hr 27 minutes. This may be one of those models that sticks around 1% for a long time then finishes up alll in a rush.:dunno:

[Thor]

Yeep, this seems to be the case... after 1:25 it jumped up to 45% so it should be ok...

I just didn't want to waste any cpu cycles:bonk:


So let's rock on!


Florian

[Exci]

Before CASP, the biggest proteins they were working on were 200 AA's long (most were much smaller). Of the 6 targets released so far for CASP7, 2 are 200 long, and one is 287. I believe the cpu time required grows exponentially as the length of the protein increases, so it makes sense these targets are taking so long to do even one iteration. Plus, they look a lot cooler in the screensaver :)

[PY 222]

With the watchdog in place, there is no worry about the client getting stuck with a work unit anymore because it will kill it and grant you points no matter what.

[pfb]

Had 2 crash out as I didn't have enough space - 1 wanted ~7,000GB and another ~1,300GB of disk space!

Need to dig out the errors and post them over at Rosetta...

:jester:

[Thor]

The next problem:

The new WUs seem have only a few save points...

It now happend to me that I wasted about 45min of cpu time because there were no savepoints :bang: :swear:

That sucks!

Should I post that other at the Rosetta forums?

Florian

[LAURENU2]

Yes Thor
You should make your thoughts be known to the people in charge
That is one of the ways good change comes about:thumbs:

[Thor]

Done

Look here (http://boinc.bakerlab.org/rosetta/forum_thread.php?id=1619)


Florian

[LAURENU2]

I also confirmed your post there :thumbs:

[Thor]

Okay it might have been a glitch on the first 5.16 units...

But they'll keep an eye on it as I will do, too

Very responsive mods!:thumbs:

I'm more than impressed by the speed with that my request was handled.

An absolutly A+ way-to-go...others can definitly learn from this example!


Florian

[Thor]

Wow, the new wus are huge!

I'm working on one which uses 273Meg of RAM :scared:

Lets see where that goes...


Edit: It is one of those: FRA_t289_CASP7_hom001_4_IGNORE_THE_REST_dect289_4_624_1.pdb_653_23_0

[Exci]

I had a couple that were taking 350 each (and another 350 of VM). My guess is they're using the 1gb+ flag for certian WU's. I think we would have heard from some 512MB folks if they DL'ed one :)

-Ethan

[Marky-UK]

I had to abort one WU earlier as Rosetta was using over 350MB RAM and the PC only has 256MB physical, so it was page faulting like crazy. I hope they don't keep handing out these huge WUs to PCs that can't handle them. :(

[Thor]

The size is not really a problem for me, as the machine has 2GB of Ram.

But I was looking at the graph and saw that is way larger than the last ones and peeked into the task manager and was a little shocked:D

The last ones were somewhere between 100-120MB ...