Maximum daily WU quota per CPU

Message boards : Number crunching : Maximum daily WU quota per CPU
Message board moderation

To post messages, you must log in.

1 · 2 · Next

AuthorMessage
Profile Royce Moore

Send message
Joined: 23 Nov 01
Posts: 4
Credit: 213,751,400
RAC: 453
United States
Message 863427 - Posted: 8 Feb 2009, 6:29:54 UTC

Any chance of getting the "Maximum daily WU quota per CPU" of a 100 per CPU upped?? Turning CUDA credits in around 3-4 minutes per and is pushing my quad core box right up to the limit or against it depending on how many Astro units I get.
ID: 863427 · Report as offensive
Luke
Volunteer developer
Avatar

Send message
Joined: 31 Dec 06
Posts: 2546
Credit: 817,560
RAC: 0
New Zealand
Message 863428 - Posted: 8 Feb 2009, 6:39:37 UTC - in response to Message 863427.  

Any chance of getting the "Maximum daily WU quota per CPU" of a 100 per CPU upped?? Turning CUDA credits in around 3-4 minutes per and is pushing my quad core box right up to the limit or against it depending on how many Astro units I get.


I do believe that the work unit quota for CUDA devices rose not long ago. Anyway, this feature is built in to prevent hoarding of workunits and the maxxing out the SETI@home bandwidth. I believe it should stay the same.

- Luke.
- Luke.
ID: 863428 · Report as offensive
Cosmic_Ocean
Avatar

Send message
Joined: 23 Dec 00
Posts: 3027
Credit: 13,516,867
RAC: 13
United States
Message 863430 - Posted: 8 Feb 2009, 6:47:34 UTC - in response to Message 863427.  

Any chance of getting the "Maximum daily WU quota per CPU" of a 100 per CPU upped?? Turning CUDA credits in around 3-4 minutes per and is pushing my quad core box right up to the limit or against it depending on how many Astro units I get.


See here..

From: David Anderson <davea@ssl.berkeley.edu>
Subject: Re: [boinc_dev] Work-fetch and gpu quotas
Date sent: Mon, 19 Jan 2009 17:03:50 -0800
Copies to: BOINC dev <boinc_dev@ssl.berkeley.edu>

I checked in a change to deal with this problem.
I added a scheduler config parameter "cuda_multiplier".
If the basic quota is K,
the actual quota for a host with N CPUs and M CUDA devices will be
K * (N + M*cuda_multiplier)
(i.e. cuda_multiplier reflects how much faster GPUs
are than CPUs for this project).

I'll deploy this on SETI@home (with a value of 5 or so) soon.

-- David

Linux laptop:
record uptime: 1511d 20h 19m (ended due to the power brick giving-up)
ID: 863430 · Report as offensive
Fred W
Volunteer tester

Send message
Joined: 13 Jun 99
Posts: 2524
Credit: 11,954,210
RAC: 0
United Kingdom
Message 863439 - Posted: 8 Feb 2009, 8:35:21 UTC

Yes, I changed to Boinc 6.6.3 four days ago and immediately downloaded 1400 WU's to my Q9450 + GTX295 (all CUDA since the 4 CPU cores were crunching Beta AP5.01's). Note that my additional work work cache is set to 1 day...
I've now found a way to live with the download problem and am still crunching through them.

F.
ID: 863439 · Report as offensive
-BeNt-
Avatar

Send message
Joined: 17 Oct 99
Posts: 1234
Credit: 10,116,112
RAC: 0
United States
Message 863456 - Posted: 8 Feb 2009, 9:54:53 UTC

Very good news, it would probably benefit the project somewhat to find someone who has the fastest cpu + fastest cuda setup and find out in a given day crunching 24/7 non-stop how many work units they could actually burn through. That way instead of a cuda device going dry it could continue to crunch. Then again, from what I have seen so far with my various computers the quata system seems to pick and choose between AP's and MB's pretty good. So for me 100 per has been very accurate. So let it ride!
Traveling through space at ~67,000mph!
ID: 863456 · Report as offensive
W-K 666 Project Donor
Volunteer tester

Send message
Joined: 18 May 99
Posts: 19013
Credit: 40,757,560
RAC: 67
United Kingdom
Message 863460 - Posted: 8 Feb 2009, 11:04:26 UTC - in response to Message 863456.  

Very good news, it would probably benefit the project somewhat to find someone who has the fastest cpu + fastest cuda setup and find out in a given day crunching 24/7 non-stop how many work units they could actually burn through. That way instead of a cuda device going dry it could continue to crunch. Then again, from what I have seen so far with my various computers the quata system seems to pick and choose between AP's and MB's pretty good. So for me 100 per has been very accurate. So let it ride!

Look no further than the top computer stats page.
ID: 863460 · Report as offensive
Profile Westsail and *Pyxey*
Volunteer tester
Avatar

Send message
Joined: 26 Jul 99
Posts: 338
Credit: 20,544,999
RAC: 0
United States
Message 863461 - Posted: 8 Feb 2009, 11:12:09 UTC


I'll let someone know if I run out ;) *shrugs*
Tesla powered

"The most exciting phrase to hear in science, the one that heralds new discoveries, is not Eureka! (I found it!) but rather, 'hmm... that's funny...'" -- Isaac Asimov
ID: 863461 · Report as offensive
Profile Woyteck - Boinc Busters Poland
Avatar

Send message
Joined: 3 Jun 99
Posts: 49
Credit: 3,203,845
RAC: 0
Poland
Message 863462 - Posted: 8 Feb 2009, 11:25:02 UTC

This guy is better thou:
Monster!



--
Get up, stand up! Don\'t give up the fight!
Credits will make everybody feel high! ;-)
ID: 863462 · Report as offensive
Profile Sutaru Tsureku
Volunteer tester

Send message
Joined: 6 Apr 07
Posts: 7105
Credit: 147,663,825
RAC: 5
Germany
Message 863532 - Posted: 8 Feb 2009, 17:46:44 UTC
Last modified: 8 Feb 2009, 18:06:20 UTC

NVIDIA Tesla C1060 (1023MB) [4 GPUs inside?*]

[4] NVIDIA GeForce GTX 295 (895MB) [(2* (2* GTX 260)) = 4 GPUs]


Maybe same performance..? ;-D


*EDIT:
After search with Google I got info that it have 4 GPUs..
But on the nVIDIA- homepage it's 'only' a 1 * 240 SPC..

Tesla C1060
Form Factor 10.5" x 4.376", Dual Slot
# of Tesla GPUs 1
# of Streaming Processor Cores 240
Frequency of processor cores 1.3GHz
Single Precision floating point performance (peak) 933
Double Precision floating point performance (peak) 78

GTX 280
GPU Engine Specs:
Processor Cores 240
Graphics Clock (MHz) 602 MHz
Processor Clock (MHz) 1296 MHz
Texture Fill Rate (billion/sec) 48.2

Memory Specs:
Memory Clock (MHz) 1107 MHz
Standard Memory Config 1 GB
Memory Interface Width 512-bit
Memory Bandwidth (GB/sec) 141.7

So a Tesla C1060 have 'only' a performance like a GTX 280 ??


EDIT #2:

@ dnolan

Yes.. :-) ..I think now the same..


---------------------------------------
@ all

What is the difference between GeForce, Tesla and Quadro? Only the prize? ;-)

http://www.nvidia.com/object/cuda_learn_products.html
ID: 863532 · Report as offensive
Profile dnolan
Avatar

Send message
Joined: 30 Aug 01
Posts: 1228
Credit: 47,779,411
RAC: 32
United States
Message 863534 - Posted: 8 Feb 2009, 17:58:57 UTC - in response to Message 863532.  


NVIDIA Tesla C1060 (1023MB) [4 GPUs inside?]

[4] NVIDIA GeForce GTX 295 (895MB) [(2* (2* GTX 260)) = 4 GPUs]


Maybe same performance..? ;-D


From the specs I can find with Google, it looks like the Tesla has a single processor and is similar to a GTX 280 except it has 4G of ram and the video card has only 1G. I'm curious to see what kind of performance it has, given its rather large price tag.

-Dave
ID: 863534 · Report as offensive
John McLeod VII
Volunteer developer
Volunteer tester
Avatar

Send message
Joined: 15 Jul 99
Posts: 24806
Credit: 790,712
RAC: 0
United States
Message 863540 - Posted: 8 Feb 2009, 18:06:14 UTC - in response to Message 863428.  

Any chance of getting the "Maximum daily WU quota per CPU" of a 100 per CPU upped?? Turning CUDA credits in around 3-4 minutes per and is pushing my quad core box right up to the limit or against it depending on how many Astro units I get.


I do believe that the work unit quota for CUDA devices rose not long ago. Anyway, this feature is built in to prevent hoarding of workunits and the maxxing out the SETI@home bandwidth. I believe it should stay the same.

- Luke.

Actaully it is in place to prevent a computer that has a problem from burning through and erroring tens of thousands of WUs per day (this happened early on, and there are still a few computers that would try to do this if the limit were not in place).


BOINC WIKI
ID: 863540 · Report as offensive
Profile Westsail and *Pyxey*
Volunteer tester
Avatar

Send message
Joined: 26 Jul 99
Posts: 338
Credit: 20,544,999
RAC: 0
United States
Message 863549 - Posted: 8 Feb 2009, 18:33:23 UTC

Correct me if wrong. I believe a cards total throughput can be roughly estimated by formula core clock x number of shaders.

Tesla C1060
Form Factor 10.5" x 4.376", Dual Slot
# of Tesla GPUs 1
# of Streaming Processor Cores 240
Frequency of processor cores 1.3GHz

13000x240 = 3,120,000

GTX 280
GPU Engine Specs:
Processor Cores 240
Graphics Clock (MHz) 602 MHz
Processor Clock (MHz) 1296 MHz
Texture Fill Rate (billion/sec) 48.2

602x240 = 144,480

Unfortunately my card came with a tiny price tag and thus is neutered a bit. It is an early engineering sample built on a GX200 chip. It only has 2 core arrays so only 192 cores. Also the clock is only 750mhz...So 144,000 or maybe 12,000 Rac? I think that makes the full production card somewhere in the neighborhood of 35,000 Rac. Not bad for $1600 I would have to say. 900Gflops is pretty nuts. Then to run 4 of them. *grins*
"The most exciting phrase to hear in science, the one that heralds new discoveries, is not Eureka! (I found it!) but rather, 'hmm... that's funny...'" -- Isaac Asimov
ID: 863549 · Report as offensive
Profile dnolan
Avatar

Send message
Joined: 30 Aug 01
Posts: 1228
Credit: 47,779,411
RAC: 32
United States
Message 863554 - Posted: 8 Feb 2009, 18:51:32 UTC - in response to Message 863549.  

Correct me if wrong. I believe a cards total throughput can be roughly estimated by formula core clock x number of shaders.

Tesla C1060
Form Factor 10.5" x 4.376", Dual Slot
# of Tesla GPUs 1
# of Streaming Processor Cores 240
Frequency of processor cores 1.3GHz

13000x240 = 3,120,000

GTX 280
GPU Engine Specs:
Processor Cores 240
Graphics Clock (MHz) 602 MHz
Processor Clock (MHz) 1296 MHz
Texture Fill Rate (billion/sec) 48.2

602x240 = 144,480


Am I missing something? Why 13,000 and not 1,300 in your math above?
Shouldn't it be 312,000 vs. 144,480? And wouldn't that make the Tesla roughly 2X the GTX 280?

-Dave
ID: 863554 · Report as offensive
Profile Westsail and *Pyxey*
Volunteer tester
Avatar

Send message
Joined: 26 Jul 99
Posts: 338
Credit: 20,544,999
RAC: 0
United States
Message 863558 - Posted: 8 Feb 2009, 19:26:10 UTC - in response to Message 863554.  
Last modified: 8 Feb 2009, 19:27:04 UTC

LOL, Thank You!

Yea, that's right. Been up all night and never been too hot with decimals. All this is pure theory of course. In the end all that matters is the RAC a machine sustains.
"The most exciting phrase to hear in science, the one that heralds new discoveries, is not Eureka! (I found it!) but rather, 'hmm... that's funny...'" -- Isaac Asimov
ID: 863558 · Report as offensive
Profile dnolan
Avatar

Send message
Joined: 30 Aug 01
Posts: 1228
Credit: 47,779,411
RAC: 32
United States
Message 863563 - Posted: 8 Feb 2009, 19:32:50 UTC


Ok, so $$-wise, 2 GTX 280s (or even 295s) seems to make more sense. But it will still be very interesting to see what kind of RAC the Tesla can pump out. Power consumption would probably be much lower (if the Nvidia site is correct) for the Tesla (compared to a couple of GTX cards), so the RAC vs. power consumption may well lean towards the Tesla device, though it could take quite a while to reach the point where initial cost vs. power savings equals out...
Have you by any chance measured what kind of power your Tesla system is using?

-Dave
ID: 863563 · Report as offensive
Profile Westsail and *Pyxey*
Volunteer tester
Avatar

Send message
Joined: 26 Jul 99
Posts: 338
Credit: 20,544,999
RAC: 0
United States
Message 863584 - Posted: 8 Feb 2009, 20:13:51 UTC - in response to Message 863563.  
Last modified: 8 Feb 2009, 20:16:26 UTC

The UPS shows 830 watts for the whole farm with everyone crunching. The Athlon with the C1060 consumes about 255-260 watts The card no matter what I can't get to pull more than 75 watts! The specs say 170 for the production version and since mine is like half of one of those I guess the numbers sound about right. Coincidently the best rac/$ it looks like right now will be the Tesla C870. Stock clock it is over 500Gflops and available online for around ~$1000. It is also GX20 based 45nm chipset with 192 shaders and 1G ram. Would like to pick one up when I find maybe a used one...*dreaming*

Right now i'm looking into external PCIe 2.0 x16 powered arrays. You can get a tower that sits next to the puter uses a standard atx power supply and can host from 1-32 cards. I think that will end up being the way to go. Imagine a 32 card monster with phase change cooling OC'ed to the wall. Then they may need to raise the 100/cpu limit. ;)
"The most exciting phrase to hear in science, the one that heralds new discoveries, is not Eureka! (I found it!) but rather, 'hmm... that's funny...'" -- Isaac Asimov
ID: 863584 · Report as offensive
Profile Sutaru Tsureku
Volunteer tester

Send message
Joined: 6 Apr 07
Posts: 7105
Credit: 147,663,825
RAC: 5
Germany
Message 863588 - Posted: 8 Feb 2009, 20:23:47 UTC - in response to Message 863584.  
Last modified: 8 Feb 2009, 20:24:54 UTC

...
Right now i'm looking into external PCIe 2.0 x16 powered arrays. ...

... Then they may need to raise the 100/cpu limit. ;)


Please give a link (maybe pic) of this nice thing! :-)

Where I could buy this?


They changed the 100 WUs/CPU/day limit.
So with a QUAD-CPU and 2* GPU you can download 1400 WUs/day.. :-D
ID: 863588 · Report as offensive
Profile Westsail and *Pyxey*
Volunteer tester
Avatar

Send message
Joined: 26 Jul 99
Posts: 338
Credit: 20,544,999
RAC: 0
United States
Message 863592 - Posted: 8 Feb 2009, 20:34:39 UTC - in response to Message 863588.  
Last modified: 8 Feb 2009, 20:35:15 UTC

Linky

Of couse would have to morgage the house to pay for it..

So how much units can I download with a dual core?
"The most exciting phrase to hear in science, the one that heralds new discoveries, is not Eureka! (I found it!) but rather, 'hmm... that's funny...'" -- Isaac Asimov
ID: 863592 · Report as offensive
Zydor

Send message
Joined: 4 Oct 03
Posts: 172
Credit: 491,111
RAC: 0
United Kingdom
Message 863595 - Posted: 8 Feb 2009, 20:39:50 UTC - in response to Message 863584.  

Right now i'm looking into external PCIe 2.0 x16 powered arrays. You can get a tower that sits next to the puter uses a standard atx power supply and can host from 1-32 cards. I think that will end up being the way to go. Imagine a 32 card monster with phase change cooling OC'ed to the wall. Then they may need to raise the 100/cpu limit. ;)


Dont forget the 4Ton Fridge to go with that rolf :)
ID: 863595 · Report as offensive
Fred W
Volunteer tester

Send message
Joined: 13 Jun 99
Posts: 2524
Credit: 11,954,210
RAC: 0
United Kingdom
Message 863596 - Posted: 8 Feb 2009, 20:43:22 UTC - in response to Message 863592.  
Last modified: 8 Feb 2009, 20:43:51 UTC

Linky

Of couse would have to morgage the house to pay for it..

So how much units can I download with a dual core?

See earlier in this thread.

i.e. 100 per CPU core + 500 per GPU core.

F.
ID: 863596 · Report as offensive
1 · 2 · Next

Message boards : Number crunching : Maximum daily WU quota per CPU


 
©2024 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.