HI:

If any help I got a PCIE NVidia Quadro FX-1300 ready for any testing you want (Windows XP64 or Linux)

Just email with details.

RP
rperaza26@comcast.net

Hi all,

Like a few of you already know, I'm currently working on a GPU port of the seti app, targeted at nVidias 8800 series of graphics cards.

I've already ported the 2 most trivial crunching steps - chirping and calculating the power spectrum - and created a little test app.

The app just generates some text output, it is _no_ replacement for the seti app.

If you have an 8800 card, it would be nice if you'd do a test run.

The app, and sources, are available at Anton's site

The whole port is pretty much a work in progress. Comments and suggestions are very welcome.

Regards Hans

---

I have a pair of 7800 GTX running in SLI if needed for testing. The GPU crunching is a very interesting idea. Let me know what needs to be done. Rig specs sould be in the sig.

---

Anyways...to the core of what I wanted to say...

I was and still am looking over the code (pretty slow at it) but it is obvious that FFT is the root problem. It has been about 2 years since I did some work with FFTs, assuming that FFT = Fast Fourier Transform (please confirm) but I was pretty good at it. I will have to pull out a few manuals/texts to refresh my memory and get back with you. But for now I will keep looking at what you have and let you know if I see anything.

-citroja

Thanks.
I'll (hopefully) get a bit further over the weekend.

Regards Hans

P.S: You'll have to register if you want to post.

---

When can someone create a 6xxx supported app? I have a 6150 LE PCIE waiting.

Ummm....we are currently trying to get an app WORKING...once that is done we can START to talk about card support...

-citroja

umm, whoops ^_^'

DIdn't read everything :P

ok after i have read infos on this page brook implementation is gone because its
speed is incomparable to CUDA ...

---

ok after i have read infos on this page brook implementation is gone because its
speed is incomparable to CUDA ...

Thats too bad...it could have made things more interesting :)

-citroja

P.S. i'll register when i get a chance...i have been helping a friend who is in the hospital and that has been taking all of my time.

---

i must correct my last post : difference between brook and CUDA is not so in speed - brook generate direct assembler code for gpu (ogl(cg)/dx9(fxc)),but when you use brook all must be in streams and kernels operations are allowed only on streams and there is a speed penalty - for example matrix multiply is done via three kernels and any kernel is doing a data upload and download ....
in pure cg and i think in CUDA too is for this a mul() function with hasnt doing any multiple data upload ....

brook might be more speedy on older cards(<G80) than CUDA (i canntot test it because i havent CUDA )...

---

i must correct my last post : difference between brook and CUDA is not so in speed - brook generate direct assembler code for gpu (ogl(cg)/dx9(fxc)),but when you use brook all must be in streams and kernels operations are allowed only on streams and there is a speed penalty - for example matrix multiply is done via three kernels and any kernel is doing a data upload and download ....
in pure cg and i think in CUDA too is for this a mul() function with hasnt doing any multiple data upload ....

brook might be more speedy on older cards(<G80) than CUDA (i canntot test it because i havent CUDA )...

I've got an X1900XTX if needed :)

---

Hans,

i have been reading posts on your site...just curious if you had a status update? I haven't seen one in a while. Like I said before I might be able to help with the FFTs and I will have some free time over the next few weeks.

I will join your site tomorrow and start going over things in more detail.

-citroja

---

by brook is a fft sample ... thanx god ....

but now i have a another problem - compilining seti source code on VS2005SP1 give a const error in xmlutil.h . can some send a good file (i am trying fix it for a two weeks and i am definitely lost)???

---

ATI vs nVidia

As far as Folding@Home... They are able to use ATI cards both in AGP slots and PCI 4x (and up) slots. Their code does not work on nVidia cards and returns errors. They've asked nVidia to clean up their drivers in some areas, but they haven't heard back from nVidia (at least they aren't saying this publicly).

Maybe the SETI-BOINC code is simpler so that nVidia cards could crunch it.

ATI cards are currently better (than nVidia) for science projects because their shaders have "4 dimensions" to them while the nVidia shaders have something like "1 dimension" to them. So while the ATI cards have fewer shaders, those shaders are "deeper" and more able to manipulate data.

They've had more success with the 512MB versions of cards than the 256MB versions. The Radeon x1950xtx 512MB does 80-90 gigaflops (compared to 3-4 gigaflops for a C2D). But (for all those concerned about credits), FaH only awards about 3x the credit of their P4 benchmark. If you throw out the credits, it is a huge contribution to science.

[on a side note: I'm building a dual GPU system to crunch for FaH. It'll probably be 150-180 gigaflops because the second card will not be running at 16x PCIe. For GPU folding presently with DirectX 9, the GPU needs to poll the CPU, so my build will have two CPU cores dedicated to the GPU.]

I stand corrected... (we'll see). nVidia's CUDA is now free software.


"GPU computing with CUDA is a new approach to computing where hundreds of on-chip processor cores simultaneously communicate and cooperate to solve complex computing problems up to 100 times faster than traditional approaches.

"This complete development environment gives developers the tools they need to solve new problems in computation-intensive applications.

"The CUDA Software Developers Kit (SDK) is currently available to developers and researchers through the NVIDIA registered developer program."

...and it's free.

This should make it MUCH easier than the kludgy way it is done on ATI cards. For more info, see:

http://www.nvidia.com/object/IO_37226.html

---

I stand corrected... (we'll see). nVidia's CUDA is now free software.


"GPU computing with CUDA is a new approach to computing where hundreds of on-chip processor cores simultaneously communicate and cooperate to solve complex computing problems up to 100 times faster than traditional approaches.

"This complete development environment gives developers the tools they need to solve new problems in computation-intensive applications.

"The CUDA Software Developers Kit (SDK) is currently available to developers and researchers through the NVIDIA registered developer program."

...and it's free.

The main problem with CUDA today is that it requires a signed NDA from a developer. So, developers cannot effectively cooperate with each other, and that is a serious show stopper for an open-source project. I am sure Hans is doing the best he can solo, but he inability to co-operate and inability to disclose his findings must be limiting him.

---

If the motive is to sell more graphics cards, I wonder why nVidia has people sign NDAs. If AMD wanted to, they could easily get someone to look at the code. It's a mystery to me.

---

i will check openvidia - by doc it look good for use (very nice implementation of DFT)....

---

Hans, perhaps you should set up an email account for people to email you and offer their GPU when it comes time to test your apps. Have them send you a blank email with a subject of "7800gt sli" or of the sort. Just an idea, considering how many offers you're getting in this thread. Hitting reply and attaching a file Would also be easier then searching through the thread and scrounging emails if they even have them available.

---

Here are some news around?

When we can crunch with our graphic- cards? :-)

---

Here are some news around?
When we can crunch with our graphic- cards? :-)

There are people working on the issue right now, the problem is that the software would need to be fairly specific to a particular card. So just any card will not work.

---

Here are some news around?
When we can crunch with our graphic- cards? :-)

There are people working on the issue right now, the problem is that the software would need to be fairly specific to a particular card. So just any card will not work.

More exactly...specific types of cards. Right now there is testing going on at various sites to see if things will work.

Cards being tested (AFAIK):
nvidia - 6x,7x,8x (both AGP and PCI-e)
ATI -1x00

If you want to help test check out this thread on Simon's site http://lunatics.at/index.php/topic,152.0.html

or Hans Dorn's site below

-citroja

---

Im gettin a page can not find server for the site http://seti.vivaxgroup.com/forum/index.php?topic=7.0
and
http://seti.vivaxgroup.com/forum/index.php
and
http://seti.vivaxgroup.com/

It has being going on for a while. Cant see any conflicting settings.

BoB

---

Do you Good Search for Seti@Home? http://www.goodsearch.com/?charityid=888957
Or Good Shop? http://www.goodshop.com/?charityid=888957