It's Alive!
log in

Advanced search

Message boards : News : It's Alive!

1 · 2 · 3 · Next
Author Message
Profile Slicker
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist
Avatar
Send message
Joined: 11 Jun 09
Posts: 2525
Credit: 740,580,099
RAC: 1
Message 10828 - Posted: 9 Jan 2011, 22:56:26 UTC

*The server maintenance has been completed.
*The BOINC server software has been updated to the latest stable release (v22891).
*The workunit size has been increased so that the fastest GPUs should take 15-20 minutes to complete a workunit.

Profile Slicker
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist
Avatar
Send message
Joined: 11 Jun 09
Posts: 2525
Credit: 740,580,099
RAC: 1
Message 10830 - Posted: 9 Jan 2011, 23:16:02 UTC

My original goal was to migrate the boot drive to a RAID 1 array. However, the conversion didn't go well. Debian wasn't identifying the new drives properly and so I eventually ended up pulling the original boot drive and installing Debian from scratch on the new drives. That meant reinstalling PHP, MySQL, Apahce, etc.
Since it was a clean install, I decided to go ahead and upgrade the BOINC server software which took longer than expected due to having to fix several bugs in the server_stable code in order to get it to work and merge the Collatz code back into the BOINC code and upgrade the database and web pages. There are still a few tweaks to be done (e.g. reinstall MRTG, Webalizer, etc.) but those can wait for a while.

Gary Craig
Send message
Joined: 30 Jan 10
Posts: 6
Credit: 3,626,547
RAC: 0
Message 10832 - Posted: 10 Jan 2011, 2:16:25 UTC

It looks like the WU size has been quadrupled, since my old slow GT120 (iMac, 18 months old) used to take 3:20:00 and now I'm on my first new WU and I estimate it will take 13:20:00. Ouch. Well, I've got a new machine on order (w/ GTX 570); hopefully I'll have that powered up in a couple weeks.
Cheers,
--Gary

Profile Slicker
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist
Avatar
Send message
Joined: 11 Jun 09
Posts: 2525
Credit: 740,580,099
RAC: 1
Message 10833 - Posted: 10 Jan 2011, 2:56:49 UTC - in response to Message 10832.

It looks like the WU size has been quadrupled, since my old slow GT120 (iMac, 18 months old) used to take 3:20:00 and now I'm on my first new WU and I estimate it will take 13:20:00. Ouch. Well, I've got a new machine on order (w/ GTX 570); hopefully I'll have that powered up in a couple weeks.
Cheers,
--Gary


I still want to split the app into two apps - one for short and another for long so that people can choose which they wish to run. I just haven't had the time lately to work on it.

Pappa
Send message
Joined: 11 Feb 10
Posts: 3
Credit: 47,790,551
RAC: 0
Message 10834 - Posted: 10 Jan 2011, 4:33:34 UTC

Slicker

I have completed one 31992648 4,979.72 seconds

Previous Average 31974297 1,190.52 seconds

So while it not yet validated, that gives you an idea. Over 4 times the previous application for a 450 GT.

Regards




den160593
Send message
Joined: 12 Dec 09
Posts: 1
Credit: 3,005,909
RAC: 0
Message 10839 - Posted: 10 Jan 2011, 8:04:19 UTC - in response to Message 10828.

May I ask why the workunit size has been increased? Does that mean that it tests more numbers in the same workunit?

Honzyk
Send message
Joined: 21 Oct 10
Posts: 1
Credit: 10,482,938
RAC: 0
Message 10840 - Posted: 10 Jan 2011, 8:25:12 UTC - in response to Message 10839.

Hello, same with me - no workunits receiving

Profile [AF>EDLS] Polynesia
Avatar
Send message
Joined: 2 Sep 09
Posts: 13
Credit: 13,018,423
RAC: 0
Message 10841 - Posted: 10 Jan 2011, 10:26:12 UTC
Last modified: 10 Jan 2011, 11:01:40 UTC

Hello,

thank you for your work ...

Since the units are longer, is that the credit system will change?

edit: 35 minutes to make a 1.05 unit ?????
____________

Profile oldDirty
Avatar
Send message
Joined: 18 Aug 09
Posts: 5
Credit: 8,958,764
RAC: 0
Message 10843 - Posted: 10 Jan 2011, 10:35:13 UTC
Last modified: 10 Jan 2011, 10:39:29 UTC

The new Wu's work good on HD58xx Cards, Runtime ~1400sec for a 5870 stock clocks.
But one thing, the required CPU Usage shows 0.89%, but when i stop the similar running WCG, i nearly had 0% CPU usage by Boinc and 2 CC Wu's are running ( 2x 58xx cards ).
The Problem now ist thats Boinc "reserve" this nearly two cores for cc and WCG for example cant use the now idling cores, in my case 2 cores ( i got a 6 core with 90% in CPU usage settings ).
So if you run 3 or 4 CC Wu's with a few cards in a box, a quadcore for example, is nearly complett set to idle, because CC reserve nearly one core per Wu's but the cores are idle in fact. I hope you understand what i want to explain.
-------------
The credits now around 3000 on a 5870.
____________
Public Energy -Crunch da Power-

[boinc.at] Nowi
Send message
Joined: 1 Nov 09
Posts: 3
Credit: 18,810,085
RAC: 0
Message 10845 - Posted: 10 Jan 2011, 11:49:08 UTC

I see the same behavior. Is it possible to change the amount of CPU used in Collatz-tasks? Before I have three cores dedicated for CPU-crunching and one for GPU-crunching. Now I need an extra core for Collatz without using it. Please make a change...

Nowi

I7I7I7I7
Send message
Joined: 31 Oct 09
Posts: 42
Credit: 36,250,826
RAC: 1
Message 10848 - Posted: 10 Jan 2011, 13:30:54 UTC - in response to Message 10841.
Last modified: 10 Jan 2011, 13:32:06 UTC

To : [AF>EDLS] Polynesia

2600 seconds and 3000 credits :)
Confirmed on my first WU :)

Profile SLAYER OF DEATH
Avatar
Send message
Joined: 12 Dec 10
Posts: 134
Credit: 17,993,517
RAC: 0
Message 10850 - Posted: 10 Jan 2011, 14:28:43 UTC - in response to Message 10848.
Last modified: 10 Jan 2011, 14:29:59 UTC

Much fan fare and khudoz on your speedy recovery of CC Slicker!!..Thanks, I was able to recive WU once you were up and running last nite. On a side note just recived 6 inches of snow here in the ol ATL (Atlanta, Ga)
____________

Proud Member of Team Carl Sagan.

Profile Mark W. Patton
Send message
Joined: 20 Apr 10
Posts: 20
Credit: 791,436,386
RAC: 3,982,739
Message 10851 - Posted: 10 Jan 2011, 14:38:46 UTC

Is there going to be an increase in points allotted for the larger work units?

Profile Slicker
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist
Avatar
Send message
Joined: 11 Jun 09
Posts: 2525
Credit: 740,580,099
RAC: 1
Message 10852 - Posted: 10 Jan 2011, 14:38:49 UTC - in response to Message 10843.

The new Wu's work good on HD58xx Cards, Runtime ~1400sec for a 5870 stock clocks.
But one thing, the required CPU Usage shows 0.89%, but when i stop the similar running WCG, i nearly had 0% CPU usage by Boinc and 2 CC Wu's are running ( 2x 58xx cards ).
The Problem now ist thats Boinc "reserve" this nearly two cores for cc and WCG for example cant use the now idling cores, in my case 2 cores ( i got a 6 core with 90% in CPU usage settings ).
So if you run 3 or 4 CC Wu's with a few cards in a box, a quadcore for example, is nearly complett set to idle, because CC reserve nearly one core per Wu's but the cores are idle in fact. I hope you understand what i want to explain.
-------------
The credits now around 3000 on a 5870.


What? You don't love the new server code? Could someone from Berkeley have made changes without thinking them through and make the CPU percentage dependant upon the WU flops and the GPU flops when that has nothing to do with it? Like that's never happened before.

I'll need to re-write the scheduler code yet again. Sure would be nice if tjhey would stick tinkering with it and leave it alone so the project admins didn't have to keep re-writing their software with each realease.

Profile Slicker
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist
Avatar
Send message
Joined: 11 Jun 09
Posts: 2525
Credit: 740,580,099
RAC: 1
Message 10853 - Posted: 10 Jan 2011, 14:52:14 UTC - in response to Message 10839.

May I ask why the workunit size has been increased? Does that mean that it tests more numbers in the same workunit?


Why? Here are a few:
* Reduce network bandwidth, especially for those who feel the need to return results immediately. For those, it will be a 75% reduction which means faster speed for everyone else.

* Increase cache size for people with multiple fast GPUs -- it should be about a half-day for those with quad GPUs which isn't great, but sure beats one hour

* Reduced I/O on the server (fewer requests for data and fewer WUs downloaded for the same work being done)

* Smaller database size which means backups will run faster and therefore less daily down-time

Size was increased 4 times. Exactly 4 times. Numbers checked per WU has increased from 200 billion to 800 billion. Since credit is dependant upon the total steps calculated, with 4 times the numbers, there will be 4 times the total steps, and 4 times the credit.

Akermit
Send message
Joined: 4 Sep 10
Posts: 2
Credit: 3,184,826
RAC: 0
Message 10863 - Posted: 10 Jan 2011, 16:22:51 UTC - in response to Message 10853.
Last modified: 10 Jan 2011, 16:26:52 UTC

Hi,

I´m fine with 4 times bigger WU´s, but you may take a look at :
http://boinc.thesonntags.com/collatz/result.php?resultid=72713915

Calculated on : ATI HD 6950

I don´t think I found a new recordnumber ... "needed 3324576156 steps for 2367547124171451865412" so could it be, that the new ATI GPU´s are currently NOT supported ?
Also the executed steps is way to high : "2232460673094528190 total executed steps" - which gives nearly 16.297.899 Creditpoints ....

It looks like the calculation is wrong :(

Greets
Akermit

Profile Slicker
Volunteer moderator
Project administrator
Project developer
Project tester
Project scientist
Avatar
Send message
Joined: 11 Jun 09
Posts: 2525
Credit: 740,580,099
RAC: 1
Message 10868 - Posted: 10 Jan 2011, 17:57:22 UTC - in response to Message 10863.

I've run a bunch of tasks on my HD 6970 card, so that isn't the problem.
I checked the number with two other big integer apps and the steps for the number listed should be 459. Usually when it finds such a large number, it is due to the GPU driver resetting while the calculations are being done and/or the GPU being OC'd to the point where it makes errors. It might work OK for the next WU. If not, try rebooting. If that doesn't work, lower the OC on the GPU.

Profile [AF>EDLS] Polynesia
Avatar
Send message
Joined: 2 Sep 09
Posts: 13
Credit: 13,018,423
RAC: 0
Message 10872 - Posted: 10 Jan 2011, 19:01:23 UTC

When you say you increase the size of wus, this means that rather than tell you so slow calculating wus there by 4 or more calculation to do?
____________

Bymark
Avatar
Send message
Joined: 28 Jul 09
Posts: 78
Credit: 592,523,966
RAC: 1,254,000
Message 10874 - Posted: 10 Jan 2011, 21:09:25 UTC - in response to Message 10872.
Last modified: 10 Jan 2011, 21:15:29 UTC

I am not happy with longer work units, but I understand why.
Running boinc for me is only keeping some computer alive, with 1 hour work unit that won’t work. Must switch some computer to faster projects . And what about the validating chase, it going to be big in a month or two. Shorter deadline please!
____________

Dirk Broer
Send message
Joined: 20 Aug 10
Posts: 33
Credit: 178,245,026
RAC: 606,073
Message 10877 - Posted: 11 Jan 2011, 0:26:03 UTC - in response to Message 10828.
Last modified: 11 Jan 2011, 0:40:49 UTC

Since I am running this project on an old Sapphire AGP HD 3850 I am not happy to see the running time increased from 48 minutes to almost 3 hours per WU, nor with the increased use of the CPU by the new Collatz version (.64).
The moment I have my HD 6990 and a 12-core AMD Bulldozer CPU to go with it I will return to this project, but for now BYE!
____________

1 · 2 · 3 · Next
Post to thread

Message boards : News : It's Alive!


Main page · Your account · Message boards


Copyright © 2018 Jon Sonntag; All rights reserved.