Distributing 4-bit workunits

Message boards : News : Distributing 4-bit workunits
Message board moderation

To post messages, you must log in.

Previous · 1 · 2 · 3 · 4 · 5 · 6 · Next

AuthorMessage
Profile Raistmer
Volunteer tester
Avatar

Send message
Joined: 18 Aug 05
Posts: 2423
Credit: 15,878,738
RAC: 0
Russia
Message 58697 - Posted: 20 Jun 2016, 19:08:27 UTC - in response to Message 58695.  

Actually, if everyone would learn to read client_state.xml, there's a perfectly good

Heh, teach to learn to read file called ReadMe (just as "Eat me" in Alice in Wonderland) first... :/
News about SETI opt app releases: https://twitter.com/Raistmer
ID: 58697 · Report as offensive
Zalster
Volunteer tester

Send message
Joined: 30 Dec 13
Posts: 258
Credit: 12,340,341
RAC: 4
United States
Message 58698 - Posted: 20 Jun 2016, 19:37:21 UTC - in response to Message 58697.  
Last modified: 20 Jun 2016, 19:37:48 UTC

Heh, teach to learn to read file called ReadMe (just as "Eat me" in Alice in Wonderland) first... :/


No one told me this was going to be a english reading course, I thought this was computer science!
ID: 58698 · Report as offensive
Profile Raistmer
Volunteer tester
Avatar

Send message
Joined: 18 Aug 05
Posts: 2423
Credit: 15,878,738
RAC: 0
Russia
Message 58700 - Posted: 20 Jun 2016, 19:51:36 UTC - in response to Message 58698.  

Heh, teach to learn to read file called ReadMe (just as "Eat me" in Alice in Wonderland) first... :/


No one told me this was going to be a english reading course, I thought this was computer science!

:D
All real sciences go together. Linguistic too.
News about SETI opt app releases: https://twitter.com/Raistmer
ID: 58700 · Report as offensive
Speedy
Volunteer tester

Send message
Joined: 16 Apr 06
Posts: 13
Credit: 335,690
RAC: 86
New Zealand
Message 58701 - Posted: 20 Jun 2016, 22:02:47 UTC - in response to Message 58686.  

Out of interest is doubling the work unit size being down to lower the amount of returned to the servers?

Nope, apparently it helps reduce the level of noise in the WUs when crunching, making it easier to find signals, and reduce the likely hood of finding false signals).

Thanks Grant
ID: 58701 · Report as offensive
Profile Eric J Korpela
Volunteer moderator
Project administrator
Project developer
Project scientist
Avatar

Send message
Joined: 15 Mar 05
Posts: 1547
Credit: 26,991,474
RAC: 1,371
United States
Message 58702 - Posted: 20 Jun 2016, 23:45:57 UTC - in response to Message 58696.  


I'll block this check in next build.


Thanks.
ID: 58702 · Report as offensive
Grumpy Swede
Volunteer tester
Avatar

Send message
Joined: 10 Mar 12
Posts: 1661
Credit: 12,823,542
RAC: 7,968
Sweden
Message 58720 - Posted: 22 Jun 2016, 12:53:00 UTC

So, what will happen with these larger WU's now. Will they be released on main, and if so, when?
WARNING!! "THIS IS A SIGNATURE", of the "IT MAY CHANGE AT ANY MOMENT" type. It may, or may not be considered insulting, all depending upon HOW SENSITIVE THE VIEWER IS, to certain inputs to/from the nervous system.
ID: 58720 · Report as offensive
Profile Eric J Korpela
Volunteer moderator
Project administrator
Project developer
Project scientist
Avatar

Send message
Joined: 15 Mar 05
Posts: 1547
Credit: 26,991,474
RAC: 1,371
United States
Message 58721 - Posted: 22 Jun 2016, 16:14:29 UTC - in response to Message 58720.  

It's probably a few weeks before I have enough results to compare with the 2-bit workunits for the same data. I'd say no earlier than a month from now.
ID: 58721 · Report as offensive
Grumpy Swede
Volunteer tester
Avatar

Send message
Joined: 10 Mar 12
Posts: 1661
Credit: 12,823,542
RAC: 7,968
Sweden
Message 58722 - Posted: 22 Jun 2016, 16:38:07 UTC - in response to Message 58721.  
Last modified: 22 Jun 2016, 16:41:00 UTC

It's probably a few weeks before I have enough results to compare with the 2-bit workunits for the same data. I'd say no earlier than a month from now.

OK, thanks for the reply Eric.
I take it that you still want as many crunchers as possible here on Beta then, until further notice.

I'll stay here then, 100% for the time being, with my i7-4790K CPU @ 4.00GHz/GTX 980 Strix. It's crunching along just fine, as of now number 4 in top hosts here in Beta, and still climbing:

https://setiweb.ssl.berkeley.edu/beta/top_hosts.php
WARNING!! "THIS IS A SIGNATURE", of the "IT MAY CHANGE AT ANY MOMENT" type. It may, or may not be considered insulting, all depending upon HOW SENSITIVE THE VIEWER IS, to certain inputs to/from the nervous system.
ID: 58722 · Report as offensive
Grumpy Swede
Volunteer tester
Avatar

Send message
Joined: 10 Mar 12
Posts: 1661
Credit: 12,823,542
RAC: 7,968
Sweden
Message 58756 - Posted: 25 Jun 2016, 22:38:13 UTC
Last modified: 25 Jun 2016, 22:41:16 UTC

Wow, another vlar storm of Arecibo tasks. That's a bit painful, even on my 980 :-)
WARNING!! "THIS IS A SIGNATURE", of the "IT MAY CHANGE AT ANY MOMENT" type. It may, or may not be considered insulting, all depending upon HOW SENSITIVE THE VIEWER IS, to certain inputs to/from the nervous system.
ID: 58756 · Report as offensive
Grant (SSSF)
Volunteer tester

Send message
Joined: 15 Jun 16
Posts: 45
Credit: 1,836,741
RAC: 0
Australia
Message 58757 - Posted: 25 Jun 2016, 23:46:25 UTC - in response to Message 58756.  

Wow, another vlar storm of Arecibo tasks. That's a bit painful, even on my 980 :-)

Yep.
I'd moved to crunching 2 WUs at a time, but have gone back to 1 with the Arecibo VLARs. Everything just grinds to a (almost) halt with them.
Grant
Darwin NT.
ID: 58757 · Report as offensive
Grumpy Swede
Volunteer tester
Avatar

Send message
Joined: 10 Mar 12
Posts: 1661
Credit: 12,823,542
RAC: 7,968
Sweden
Message 58758 - Posted: 26 Jun 2016, 1:09:42 UTC - in response to Message 58757.  
Last modified: 26 Jun 2016, 1:09:58 UTC

Wow, another vlar storm of Arecibo tasks. That's a bit painful, even on my 980 :-)

Yep.
I'd moved to crunching 2 WUs at a time, but have gone back to 1 with the Arecibo VLARs. Everything just grinds to a (almost) halt with them.

I'm making no changes. Still running 3 at a time. Just because I'm stubborn, and that it still works :-)
WARNING!! "THIS IS A SIGNATURE", of the "IT MAY CHANGE AT ANY MOMENT" type. It may, or may not be considered insulting, all depending upon HOW SENSITIVE THE VIEWER IS, to certain inputs to/from the nervous system.
ID: 58758 · Report as offensive
TBar
Volunteer tester

Send message
Joined: 2 Jul 13
Posts: 505
Credit: 5,019,318
RAC: 9,689
United States
Message 58759 - Posted: 26 Jun 2016, 5:07:39 UTC

I'm seeing a number of these messages;
Find triplets Cuda kernel encountered too many triplets, or bins above threshold, reprocessing this PoT on CPU...

http://setiweb.ssl.berkeley.edu/beta/result.php?resultid=24167553
http://setiweb.ssl.berkeley.edu/beta/result.php?resultid=24167587
I don't remember seeing these messages before. Anyone else seeing them?
It appears there are more on the Arecibo VLARs.
ID: 58759 · Report as offensive
Grant (SSSF)
Volunteer tester

Send message
Joined: 15 Jun 16
Posts: 45
Credit: 1,836,741
RAC: 0
Australia
Message 58760 - Posted: 26 Jun 2016, 6:17:46 UTC - in response to Message 58759.  

I'm seeing a number of these messages;
Find triplets Cuda kernel encountered too many triplets, or bins above threshold, reprocessing this PoT on CPU...

http://setiweb.ssl.berkeley.edu/beta/result.php?resultid=24167553
http://setiweb.ssl.berkeley.edu/beta/result.php?resultid=24167587
I don't remember seeing these messages before. Anyone else seeing them?
It appears there are more on the Arecibo VLARs.


Found a couple here,
24164653
24164656
Grant
Darwin NT.
ID: 58760 · Report as offensive
TBar
Volunteer tester

Send message
Joined: 2 Jul 13
Posts: 505
Credit: 5,019,318
RAC: 9,689
United States
Message 58761 - Posted: 26 Jun 2016, 7:15:34 UTC - in response to Message 58760.  

Found a couple here,
24164653
24164656

Well, at least I'm not the only one...
I'm having a hard time finding any nVidia cards running CUDA on an Arecibo VLAR.
So far I've only found one message on a BLC task, so it appears to be basically the Arecibo VLARs. Good thing they're not supposed to run Arecibo VLARs on Main.
ID: 58761 · Report as offensive
Grant (SSSF)
Volunteer tester

Send message
Joined: 15 Jun 16
Posts: 45
Credit: 1,836,741
RAC: 0
Australia
Message 58762 - Posted: 26 Jun 2016, 7:53:49 UTC - in response to Message 58761.  

I'm having a hard time finding any nVidia cards running CUDA on an Arecibo VLAR.
So far I've only found one message on a BLC task, so it appears to be basically the Arecibo VLARs. Good thing they're not supposed to run Arecibo VLARs on Main.

I've been going from 2 to 1 then 1 to 2WUs at a time to give the APR a shakeup as the manager ended up with SaH as it's choice.

By APR (before mucking around with the number of WUs at a time)
1 SaH
2 CUDA42
3 CUDA50
4 SoG (way, way in last place).

when in reality it's
1 SoG (by a pretty good margin)
2 CUDA50
3 CUDA42
4 Sah
Grant
Darwin NT.
ID: 58762 · Report as offensive
TBar
Volunteer tester

Send message
Joined: 2 Jul 13
Posts: 505
Credit: 5,019,318
RAC: 9,689
United States
Message 58763 - Posted: 26 Jun 2016, 8:35:00 UTC - in response to Message 58762.  
Last modified: 26 Jun 2016, 9:16:01 UTC

Here's how the SoG & SaH Apps work on an nVidia Mac Laptop;
BLC3 opencl_nvidia_SoG_mac
Run time: 5 hours 42 min 14 sec
CPU time: 44 min 44 sec
http://setiweb.ssl.berkeley.edu/beta/result.php?resultid=24161117
BLC3 opencl_nvidia_mac
Run time: 5 hours 59 min 6 sec
CPU time: 42 min 31 sec
http://setiweb.ssl.berkeley.edu/beta/result.php?resultid=24150968
That's in Darwin 14.5, it's worse in Darwin 15.5 as they give Incorrect Results in 15.4 and 15.5.

Here is a similar Mac Laptop running the CUDA 6.5 App;
BLC2 x41zi (baseline v8), Cuda 6.50
Run time: 58 min
CPU time: 4 min 57 sec
http://setiathome.berkeley.edu/result.php?resultid=4986572842
A BLC3 task isn't much longer than the BLC2 task running CUDA, much faster than 5.5 hours anyway.
As a Bonus, the CUDA App actually produces the Correct Results in Darwin 15.4 & 15.5.
The New CUDA Apps I'm about to post should be a little better than the CUDA 6.5 App.


Hmmmm, this App didn't give any messages with an Arecibo VLAR,
http://setiweb.ssl.berkeley.edu/beta/result.php?resultid=24167781
???
ID: 58763 · Report as offensive
Grumpy Swede
Volunteer tester
Avatar

Send message
Joined: 10 Mar 12
Posts: 1661
Credit: 12,823,542
RAC: 7,968
Sweden
Message 58767 - Posted: 26 Jun 2016, 22:58:25 UTC

Yup, that was ugly, but now it's over :-)
WARNING!! "THIS IS A SIGNATURE", of the "IT MAY CHANGE AT ANY MOMENT" type. It may, or may not be considered insulting, all depending upon HOW SENSITIVE THE VIEWER IS, to certain inputs to/from the nervous system.
ID: 58767 · Report as offensive
TBar
Volunteer tester

Send message
Joined: 2 Jul 13
Posts: 505
Credit: 5,019,318
RAC: 9,689
United States
Message 58768 - Posted: 27 Jun 2016, 3:23:32 UTC - in response to Message 58760.  

I'm seeing a number of these messages;
Find triplets Cuda kernel encountered too many triplets, or bins above threshold, reprocessing this PoT on CPU...

http://setiweb.ssl.berkeley.edu/beta/result.php?resultid=24167553
http://setiweb.ssl.berkeley.edu/beta/result.php?resultid=24167587
I don't remember seeing these messages before. Anyone else seeing them?
It appears there are more on the Arecibo VLARs.


Found a couple here,
24164653
24164656

Well, that was interesting. I booted to Darwin 15.4 and the same App only produced One line in One of Four Tasks. The other three Arecibo VLARs didn't have any "Find triplets Cuda kernel encountered too many triplets, or bins above threshold, reprocessing this PoT on CPU..." lines. Strange. Same app, same CUDA Driver, Different OS.
???
ID: 58768 · Report as offensive
Grant (SSSF)
Volunteer tester

Send message
Joined: 15 Jun 16
Posts: 45
Credit: 1,836,741
RAC: 0
Australia
Message 58769 - Posted: 27 Jun 2016, 3:45:03 UTC - in response to Message 58768.  

Well, that was interesting. I booted to Darwin 15.4 and the same App only produced One line in One of Four Tasks. The other three Arecibo VLARs didn't have any "Find triplets Cuda kernel encountered too many triplets, or bins above threshold, reprocessing this PoT on CPU..." lines. Strange. Same app, same CUDA Driver, Different OS.
???

Just one more variable to add to the mix.


I decided I was going to move my main system back to main, and my old C2D over to Beta and leave it crunching here.
However it's got an older driver (OpenCL 1.1) and needs a more recent driver to run the OpenCL applications. Upgraded the driver, and what do you know, "Display driver has stopped responding and has recovered..." messages- even when running CUDA applications.
I've tried 4 different drivers, I've tried the TdrDelay registry hack, rebooting, those same 4 drivers yet again.
No joy.
There's something about those drivers that sucks, at least on Vista 32bit. With a 64bit installation they might work, but as it is they just fall over when you ask them to do some crunching, yet on the older drivers the GPU can handle 3 WUs at a time; no restarts, crashes or hangs.
Grant
Darwin NT.
ID: 58769 · Report as offensive
Grumpy Swede
Volunteer tester
Avatar

Send message
Joined: 10 Mar 12
Posts: 1661
Credit: 12,823,542
RAC: 7,968
Sweden
Message 58778 - Posted: 28 Jun 2016, 1:57:49 UTC
Last modified: 28 Jun 2016, 2:01:26 UTC

Heh, so it takes 3 AMD Hawaii (4096MB) OpenCL: 2.0, to beat my single GTX 980 (4095MB) + INTEL Intel(R) HD Graphics 4600 (1400MB)

Very interesting, and costly :-)

http://setiweb.ssl.berkeley.edu/beta/top_hosts.php

Edit: And Microsoft Windows 8.1 still rules. Holds places 1-3.
WARNING!! "THIS IS A SIGNATURE", of the "IT MAY CHANGE AT ANY MOMENT" type. It may, or may not be considered insulting, all depending upon HOW SENSITIVE THE VIEWER IS, to certain inputs to/from the nervous system.
ID: 58778 · Report as offensive
Previous · 1 · 2 · 3 · 4 · 5 · 6 · Next

Message boards : News : Distributing 4-bit workunits


 
©2019 University of California
 
SETI@home and Astropulse are funded by grants from the National Science Foundation, NASA, and donations from SETI@home volunteers. AstroPulse is funded in part by the NSF through grant AST-0307956.