Message boards :
Number crunching :
New GPU´s Could make SETI FASTER!!!!!
Message board moderation
Author | Message |
---|---|
Wingman2 Send message Joined: 2 May 00 Posts: 4 Credit: 27,810 RAC: 0 |
Taking Adtantig of the graphic card power will be interesting. Take a look at this http://www.tomshardware.com/hardnews/20040902_135943.html http://www.extremetech.com/article2/0,1558,1642832,00.asp?kc=ETRSS02129TX1K0000532 It can help since seti is much of sound analizang http://www.bionicfx.com/ |
Wingman2 Send message Joined: 2 May 00 Posts: 4 Credit: 27,810 RAC: 0 |
No news about the subject? |
keeleysam Send message Joined: 17 Dec 03 Posts: 133 Credit: 60,478,373 RAC: 0 |
turn the URLs into links... as most are too lazy to copy and paste. |
1mp0£173 Send message Joined: 3 Apr 99 Posts: 8423 Credit: 356,897 RAC: 0 |
> Taking Adtantig of the graphic card power will be interesting. > Take a look at this > http://www.tomshardware.com/hardnews/20040902_135943.html > > http://www.extremetech.com/article2/0,1558,1642832,00.asp?kc=ETRSS02129TX1K0000532 > It can help since seti is much of sound analizang > > http://www.bionicfx.com/ The big question is: how many people would have graphic cards that are faster than their CPUs -- and how much work would be involved in trying to use them. |
Benher Send message Joined: 25 Jul 99 Posts: 517 Credit: 465,152 RAC: 0 |
There is some discussion of this at the open source setiboinc development site. Astropulse code already has some GPU code in it. Discussion of GPU usage ...and another |
meckano Send message Joined: 1 Jul 03 Posts: 130 Credit: 48,466 RAC: 0 |
You mean, faster for those that run the screensaver part, and not for those that run the client only. Yes? ----------------------- Click to see my tag My tag SNAFU'ed? Turn the Page! :D |
Benher Send message Joined: 25 Jul 99 Posts: 517 Credit: 465,152 RAC: 0 |
> You mean, faster for those that run the screensaver part, and not for those > that run the client only. Yes? No, GPU = graphics Processor. Latest few revisions of NVidia and ATI chips both have floating point engine built into the GPU. And people have already written some advanced math routines that can run on the chip in the chip/board's memory. So, for these the regular CPU would be doing one WU, and the GPU would be crunching another WU. |
Chilean Send message Joined: 6 Apr 03 Posts: 498 Credit: 3,200,504 RAC: 0 |
I'm not an expert on this, but I've heard that the memory speed is too slow, PCI Express might solve this problem, but I dont know how many people do have it, and if this would be worth it. I personally have an pretty fast graphics card, but it doesnt go any higher than 350MHz stable after overclocking it.... |
meckano Send message Joined: 1 Jul 03 Posts: 130 Credit: 48,466 RAC: 0 |
I read the links above, very interesting. Like it says there, it's more for the music processing bunch of folk for now, but they do mention Seti. - Audio software is like 25 000$ they say, and this software would be under 1 000$ - PCI Express is new and not many would have it, and the cards can be over 1 000$. But next to 25 000$, it is a big gain plus, - Apparently the GPU can do things that the cpu just can't! -- like 40 G flops instead of 5 Too bad ATI isn't in the picture, anywho, I don't have a new ATI. :P lol - ATI was the only thing that ran well with my AMD when I was buying. Edit: There they say it is the transfer-back speed that slows things. ----------------------- Click to see my tag My tag SNAFU'ed? Turn the Page! :D |
-= Vyper =- Send message Joined: 5 Sep 99 Posts: 1652 Credit: 1,065,191,981 RAC: 2,537 |
>I'm not an expert on this, but I've heard that the memory speed is too slow, PCI Express might solve this problem, but I dont know how many people do have it, and if this would be worth it. I personally have an pretty fast graphics card, but it doesnt go any higher than 350MHz stable after overclocking it.... Yes that indeed is right, the bus between GPU and CPU is too slow. But!! If the code is in GPU Ram and the data is in the GPU Ram, then!!! Imagine a sparking brand new 6800Ultra or ATI.. These bad boys have aprox 35-40 GigaByte/Sec in ramspeed compared to about 6 GigaByte/S in a dual channel machine today!! If the code is gpu active and executed in a backbuffer on the graphicscard you can work with windows without even knowing that the GPU is in 100% state buzy running SETI :) _________________________________________________________________________ Addicted to SETI crunching! Founder of GPU Users Group |
Chilean Send message Joined: 6 Apr 03 Posts: 498 Credit: 3,200,504 RAC: 0 |
> >I'm not an expert on this, but I've heard that the memory speed is too > slow, PCI Express might solve this problem, but I dont know how many people do > have it, and if this would be worth it. I personally have an pretty fast > graphics card, but it doesnt go any higher than 350MHz stable after > overclocking it.... > > Yes that indeed is right, the bus between GPU and CPU is too slow. But!! > If the code is in GPU Ram and the data is in the GPU Ram, then!!! > > Imagine a sparking brand new 6800Ultra or ATI.. These bad boys have aprox > 35-40 GigaByte/Sec in ramspeed compared to about 6 GigaByte/S in a dual > channel machine today!! > > If the code is gpu active and executed in a backbuffer on the graphicscard you > can work with windows without even knowing that the GPU is in 100% state buzy > running SETI :) > > Thats what I thought when they told me that... but now, the core speed is even slower than my first PC... |
-= Vyper =- Send message Joined: 5 Sep 99 Posts: 1652 Credit: 1,065,191,981 RAC: 2,537 |
> > >I'm not an expert on this, but I've heard that the memory speed is > too > > slow, PCI Express might solve this problem, but I dont know how many > people do > > have it, and if this would be worth it. I personally have an pretty fast > > graphics card, but it doesnt go any higher than 350MHz stable after > > overclocking it.... > > > > Yes that indeed is right, the bus between GPU and CPU is too slow. But!! > > If the code is in GPU Ram and the data is in the GPU Ram, then!!! > > > > Imagine a sparking brand new 6800Ultra or ATI.. These bad boys have > aprox > > 35-40 GigaByte/Sec in ramspeed compared to about 6 GigaByte/S in a dual > > channel machine today!! > > > > If the code is gpu active and executed in a backbuffer on the > graphicscard you > > can work with windows without even knowing that the GPU is in 100% state > buzy > > running SETI :) > > > > > Thats what I thought when they told me that... but now, the core speed is even > slower than my first PC... > Yes i know that it is but does a modern CPU even have sixteen different shader processors, no but they do have 2 or even 3 pipes of FPU power.. I don't hardly believe that they could do everything but if the cpu and the gpu do different types of operations i really think that it could match up working side by side.. Imagine effecient code that the cpu tells that the gpu should calculate different aspects in the frequency side by side by chunks of sixteen then it could make sense.. _________________________________________________________________________ Addicted to SETI crunching! Founder of GPU Users Group |
Wingman2 Send message Joined: 2 May 00 Posts: 4 Credit: 27,810 RAC: 0 |
Ok. We know it coul help. What does seti developer think about this |
Benher Send message Joined: 25 Jul 99 Posts: 517 Credit: 465,152 RAC: 0 |
> Ok. > We know it coul help. > What does seti developer think about this Theres really only one main "seti" developer...Eric Korpela. David Anderson and others helped in developing parts of seti science code but Eric did most of it, or now manages most of it, and now he can only dedicate 25% of his employed time to seti. David, Rom, Matt and the rest all work mostly on BOINC now. So, there is a public website for improving seti science code. Developers can ask to join and contribute to the code. There is a discussion thread on its forums about using GPU with seti, and some GPU source code already for the "Astropulse" program. Have a look around. |
©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.