Boinc pentathalon race

[Rattledagger]

On an ATI 4850, Collatz tasks have been finishing in 1,000 - 1,100 seconds with a few outliers. In round numbers, let's say that's 17 minutes. There are no CPU tasks left in my history. IIRC, those took many hours.

I'm not the only one who has been running this & hasn't checked in on this thread. Hopefully, the rest will continue to running Collatz as well.

Finding any computer running the cpu-version of Collatz is difficult, but did manage locating two Xeon-systems, one X7550, 2 GHz 32-way, and another X5570, 2.93 GHz 16-way, there the 2 GHz uses roughly 24 hours while faster one just under 15 hours per wu. Meaning the Ati-4850 is roughly 50x faster than a 2.93 GHz and roughly 85x faster than the 2 GHz...

Taking into account #cpu's, the Ati-4850 still beats the 32-way system 2.7x in daily production, and the 16-way system 3.2x in daily production.

The cuda-cards is apparently slower, but still the gpu's has an advantage.

 

[Rudy Toody]

This just in!

The forum link on the TeAm AnandTech page of Einstein@Home has been repaired.

Now, back to "Days of Our Crunch," in progress.

 

[dajeepster]

Nice crunchin' everyone!!

Dajeepster, your AC must be set on high for this race!

I've run my AC so cold that the unit has frozen up on me a couple of times... then i have to wait a day for it to thaw.... those are the times when my O/P drop cause I have to shut down most of my computers so my place doesn't spontaneously combust. I keep the A/C set to 55F... but my place hovers around 85-88F... I'm currently living in a studio (was in a rush, but it worked out for cooling needs since I only have to cool one room)

 

[biodoc]

Here's the latest sTAts!

Keep crunching everyone!!

 

[biodoc]

The forum link on the TeAm AnandTech page of Einstein@Home has been repaired.

Thanks RT!

 

[Rattledagger]

The last project has now been announced, a little later than it should have been:
http://www.setigermany.de/blog/lang/en/category/boinc-pentathlon/

Rosetta@home is a one-quorum project, and has 10-days deadline. Any work returned before the end should therefore be counted. Also, it's possible to start running Rosetta@home beforehand, and disable the network...

But, the problem is that if you're also taking part in the other projects, you can't really disable the network...

 

[Ken g6]

Thanks, RD! If I knew all the Rosetta@home servers, I could download a bunch of work and then disable them in my hosts file. Anyone know if there is a list somewhere, like in the BOINC directory?

Edit: Come to think of it, I only need to block reporting of WUs. And I bet all reporting of WUs goes through the project's web site. So if that's true all I'd have to do is block boinc.bakerlab.org.

 

[Rattledagger]

Thanks, RD! If I knew all the Rosetta@home servers, I could download a bunch of work and then disable them in my hosts file. Anyone know if there is a list somewhere, like in the BOINC directory?

Edit: Come to think of it, I only need to block reporting of WUs. And I bet all reporting of WUs goes through the project's web site. So if that's true all I'd have to do is block boinc.bakerlab.org.

Blocking boinc.bakerlab.org is a good idea, but this will only have an effect if the scheduling-server fails to respond 10 times in a row. This is currently srv4.bakerlab.org

Also, if you're trying to block in hosts file, atleast some BOINC-clients has a bug there they caches a successful DNS-lookup indefinitely. I don't remember if this is also the case when user overrrides with hosts file... So, if you're not running v6.10.4x or later, will recommend to re-start BOINC-client.

As for Rosetta generally, you can increase the run-time upto 24 hours, this should give less to download, and if not mistaken there's some extra initializing on wu-start so the longer each run, the less "wasted" cycles... But, on the flip-side, in case of an error, you'll risk losing close to 24 hours...

 

[biodoc]

Reminder: Milkyway@home starts 00:00 UTC 5/9/2010 (Tonight @ 8PM EDT)

If you have double-precision GPUs, tune them into Milkyway!

Here's Day 3 stats for Einstein@Home.

 

[mk]

Reminder: Milkyway@home starts 00:00 UTC 5/9/2010 (Tonight @ 8PM EDT)

If you have double-precision GPUs, tune them into Milkyway!

Unless things have changed in the past few months Milkyway and Collatz both grant fairly generous credits for CPUs as well. No reason to entirely skip these projects if you don't have a suitable GPU.

Oh, and don't forget the Milkyway Optimized Apps.

 

[Ken g6]

Thanks, MK, didn't know Milkyway had optimized apps. Since I'm running out of Einstein WUs that have already been completed by wingmen, I'll head over that way shortly.

Meanwhile, it looks like one can run FreeHAL in a VM fairly well if one has enough RAM. I finished some 22 WUs in an Ubuntu 9.04 32-bit VM with 768MB RAM allocated to it. I'll flush them late tomorrow and get more.

 

[Rattledagger]

Unless things have changed in the past few months Milkyway and Collatz both grant fairly generous credits for CPUs as well. No reason to entirely skip these projects if you don't have a suitable GPU.

Oh, and don't forget the Milkyway Optimized Apps.

Atleast with the Ati-win64, the "optimized" is the exact same files as you'll otherwise automatically download from Milkyway, so it's a waste of time to use it.

While it's possible to run the cpu-versions of both Milkyway & Collatz, atleast my plan is to cache-up and run Rosetta@home instead.

In any case, it's now 4.5 hours before the start of Milkyway, so anyone that can run Milkyway should be ready.
Milkyway is single-redundancy, except off-chance fails validation, and has a limit of 48 tasks per GPU. This means little or no work can be cached beforehand.
Still, even 1-2 hours cached beforehand can be an advantage.

 

[filibusterman]

hmm i cannot get into my account at the freeHAL. Does anyone else have this problem. I thought I saw in earlier posts you can do many wu at once and I am only getting one assigned at a time so I would like to fix that.

 

[Ken g6]

FreeHAL's servers have been up and down for me alot.

But remember, you won't get credit for FreeHAL work submitted until about 26 hours from now.

Edit: Try this link. They have a tiny link to that from the donation page. :\

 

[Ken g6]

MK, the SSE4.1 Linux binary I downloaded kept giving me computation errors at ~30%. Something also caused BOINC to report malloc memory corruption errors. I had to upgrade to 6.10.17 to fix that.

I'll see how the standard binary does.

 

[Ken g6]

"Reset project" didn't get rid of the app_info.xml. And now the project says no work available.

 

[Rattledagger]

"Reset project" didn't get rid of the app_info.xml. And now the project says no work available.

Once-upon-a-time the "best" method to stop using anonymous platform-mechanism was to detach, re-start BOINC and re-attach... Possible the same rules applies now, it's atleast a method that should work without any problems.

In any case, 2 hours after the start of Milkyway, TeAm AnandTech is up two spots to #11 in the totals.

This even in both Einstein and Milkyway we're at #13...

Atleast for now it seems the server is keeping-up with demand.

 

[Ken g6]

Yeah, I think I got it working now. Just had to remove the app_info.xml. But the default app is at least twice as slow.

 

[filibusterman]

this is what I am getting http://www.freehal.org/forum/viewtopic.php?f=11&t=38

I assume you have to login to set how many wu you can take at one time?

 

[dajeepster]

this is what I am getting http://www.freehal.org/forum/viewtopic.php?f=11&t=38

I assume you have to login to set how many wu you can take at one time?

have you tried to use mozilla, google, or opera?

at least according to the message on front, you'll have problems if you use IE... but they say the other browsers are ok...!?!... ??!

i'm only using IE explorer on my main... and i'm getting the same error.

anyone else having the same problem?

 

[biodoc]

have you tried to use mozilla, google, or opera?

I'm getting the same error on IE, Google Chrome, and safari on my mac.

 

[biodoc]

We're ranked 13th overall: 13th on Einstein and 14th on Milkyway

Just in case there is any confusion our credits count if you are crunching:

Einstein@Home: GPU and CPU (ends tonight at 8PM EDT)
Milkyway@Home: GPU and CPU

FreeHal@Home doesn't start until tonight (8PM EDT) but you may want to start earlier since the WUs take very little CPU%. Hopefully they will fix the link to our accounts before tonight. :hmm:

Here's some individual stats. These aren't entirely accurate since it is difficult to time start and end for these mini-races. It's going to get more complicated starting tomorrow.

 

[biodoc]

anyone else having the same problem?

Looks like it's fixed!

I cranked my preferences up to 25 simultaneous WUs

Quad1 is crunching 1 WU
Quad2 is crunching 2 WU
Quad3 is crunching 4 WU

Methinks FreeHal has taken over the project!!

 

[biodoc]

FreeHAL has work now.

25 WUs on one Quad:

 

[dajeepster]

FreeHAL has work now.

25 WUs on one Quad:

it appears to be down now