Einstein@home & Aqua@home
![[B^S] thierry@home](https://einsteinathome.org/sites/default/files/20253_avatar_2.jpg?1470151864 "[B^S] thierry@home"){kind=avatar}
9 Nov 2009 21:47:03 UTC
Topic 194616
(moderation:
Just a remark: I'm running Einstein (GPU), Aqua & Milkyway on a Q9300 with a graphic card GTX285.
When an Einstein WU is running on the GPU:
- it uses 1 CPU core,
- because Aqua needs (needs or uses 4 cores when available?) 4 CPU cores, Aqua doesn't run,
- the three remaining cores don't crunch any Milkyway WUs.
[pre] [/pre]
Intel I7 930 - GTX 480 - Windows 7 64
Join BOINC Synergy, the best team in the galaxy!
Einstein@home & Aqua@home
)
You don't say which version of BOINC.
This is a known problem with v6.10.17/.18
I recommend (see AQUA message boards) using v6.10.13 when combining AQUA with single-CPU projects.
There is a code change tonight (changeset [trac]changeset:19522[/trac]) which addresses a typo made three weeks ago: we'll have to wait for a new version to be compiled before we can be sure that this is the problem they think they're fixing.
I use 6.10.17. So it's clear
)
I use 6.10.17. So it's clear :-)
Thanks
[pre] [/pre]
Intel I7 930 - GTX 480 - Windows 7 64
Join BOINC Synergy, the best team in the galaxy!
I am using BOINC 6.6.41 on my
)
I am using BOINC 6.6.41 on my Linux box, without CUDA, and I find no problem in running AQUA and other four projects. It is true that AQUA grabs all 2 cores of my Opteron 1210 but it also switches regularly every hour and uses very little high priority mode, except at start, maybe, so all projects coexist.
Tullio
I'm not familiar with Aqua.
)
I'm not familiar with Aqua. What is that?
RE: I'm not familiar with
)
D-Wave's AQUA (Adiabatic QUantum Algorithms) is a research project whose goal is to predict the performance of superconducting adiabatic quantum computers on a variety of hard problems arising in fields ranging from materials science to machine learning.From the AQUA homepage. For other BOINC projects, see Choosing BOINC projects at the BOINC site.
Gruß,
Gundolf
Computer sind nicht alles im Leben. (Kleiner Scherz)
RE: I'm not familiar with
)
The big difference with Aqua compared to other Boinc projects is that Aqua appears to be the first to be using a multithreaded science application. It will use all available CPU cores in parallel to speed up the processing of a single WU.
Very good for improving the turn-around time for the project. Yet another scheduling nightmare for JM7 to unravel!
Happy crunchin',
Martin
See new freedom: Mageia Linux
Take a look for yourself: Linux Format
The Future is what We all make IT (GPLv3)
RE: RE: I'm not familiar
)
Depending on BOINC Version it can also cause significant issues because the multithreaded portion of the Resource Scheduler is even newer than the GPU portion (which in some cases is still not working correctly).
Jump in cautiously...
Oh, and it is being run by and will support the commercial success of a for profit company... If you only want to support not-for-profit university type efforts, Aqua is not for you ...
On my Q9300 running on all four CPUs the one task I ran recently took over 24 hours to complete ... which I cannot show you because they have already purged it ... :(
RE: Jump in
)
Paul,
Good comment on all counts there.
For my own personal example, I think my machines are busy enough with the various research projects where (preferably ALL) the results are made freely available.
Quantum computing is a very new and very good area of development. It's good that they are in Boinc!
I guess there are similar questions about who profits from the effort for the various pharmaceutical projects...
Happy crunchin',
Martin
See new freedom: Mageia Linux
Take a look for yourself: Linux Format
The Future is what We all make IT (GPLv3)
It took 142 hours on my
)
It took 142 hours on my Opteron 1210 to finish a 72-qubit AQUA run. I think that they are still so far away from a commercial solution that they are still doing basic science. Their approach is very different from all research done in Institutions like NIST, which has recently published an article on Nature describing a two qubit logic gate. I am wondering who is financing them. Must be a very brave venture capital entrepreneur.
Tullio
RE: Paul, Good comment on
)
Thanks :)
I agree that it is good that they are here, and have supported their efforts some. I don't like the way that UCB has done the MT portion and there is still issues there not yet fully addressed as of 6.10.18 (and I am not running that version either, still at .13) ... the problem still seems to be that the lack of BOINC running experience is slowing progress, still ...
You do raise a point at the last, you cannot even be sure of the University efforts in that some of them are just as likely to eventually be sold to the benefit of the Univ. instead of released into the wild. Even those efforts where open publication is promised, well, at the end of the day it is still a matter of "trust me" on our parts ...
Complicated by, on some projects, the lack of feedback by the project staffs ...
Oh, ugh ... looks like Collatz is down ... sigh ...