Einstein support OpenCL ?

Paul D. Buck
Paul D. Buck
Joined: 17 Jan 05
Posts: 754
Credit: 5385205
RAC: 0

RE: RE: The truth is that

Message 90712 in response to message 90710

Quote:
Quote:
The truth is that to do GPU computing you need a card, in general, that costs more than $100 to even get into the game to start at the low end. If you want to do serious work, well, right off you need to start thinking of a card in the $200+ range ... Domination required true commitments of cash ... :)

Sure, but you replace 60 regular computers with one GPU, so even $200+ is not too much.


Which just means that project that were previously impractical now become practical. Take GPU Grid, likely it would not have been started several years ago because of the intensity of the work ... now it is practical...

We get a few more GPU projects and this will actually mean more resources to non-GPU projects.

Simplex0
Simplex0
Joined: 1 Sep 05
Posts: 152
Credit: 964726
RAC: 0

It seams that all of the

It seams that all of the different projects that use GPU optimized code and runs under BOINC have the code optimized for Intel and NVIDIA except Milkyway@home. I wonder why? is it that much harder to write a program that runs on RADEON than on NVIDIA?

Take a look at this page, the speed is enormous.

Jord
Joined: 26 Jan 05
Posts: 2952
Credit: 5779100
RAC: 0

Nvidia ported over the

Message 90714 in response to message 90713

Nvidia ported over the Seti@Home application and helped add the code to BOINC. ATI/AMD wasn't interested at the time, didn't have the manpower or money or whatever. They basically told BOINC and projects to add support and port over the applications themselves, without so much as a peep of help. The documentation to port over the applications to ATI's CAL option isn't really much of a documentation. It's more trial and error.

The Milkyway application was made by a volunteer, a project participant.

Stefan Ledwina
Stefan Ledwina
Joined: 23 Oct 05
Posts: 17
Credit: 2410973169
RAC: 28607

RE: It seams that all of

Message 90715 in response to message 90713

Quote:

It seams that all of the different projects that use GPU optimized code and runs under BOINC have the code optimized for Intel and NVIDIA except Milkyway@home. I wonder why? is it that much harder to write a program that runs on RADEON than on NVIDIA?

Take a look at this page, the speed is enormous.

As for the speed differences and why most projects use CUDA, this forum thread on the GPUGRID message boards might explain it a little bit...

Simplex0
Simplex0
Joined: 1 Sep 05
Posts: 152
Credit: 964726
RAC: 0

RE: RE: It seams that all

Message 90716 in response to message 90715

Quote:
Quote:

It seams that all of the different projects that use GPU optimized code and runs under BOINC have the code optimized for Intel and NVIDIA except Milkyway@home. I wonder why? is it that much harder to write a program that runs on RADEON than on NVIDIA?

Take a look at this page, the speed is enormous.

As for the speed differences and why most projects use CUDA, this forum thread on the GPUGRID message boards might explain it a little bit...

Why not use both like Folding@home?

Simplex0
Simplex0
Joined: 1 Sep 05
Posts: 152
Credit: 964726
RAC: 0

RE: Nvidia ported over the

Message 90717 in response to message 90714

Quote:

Nvidia ported over the Seti@Home application and helped add the code to BOINC. ATI/AMD wasn't interested at the time, didn't have the manpower or money or whatever. They basically told BOINC and projects to add support and port over the applications themselves, without so much as a peep of help. The documentation to port over the applications to ATI's CAL option isn't really much of a documentation. It's more trial and error.

The Milkyway application was made by a volunteer, a project participant.

Hmm... It seams that they need help from volunteers both to write the code and crunsh.

Einstein@Home have also had good help from volonteers to write their optimized
cood if I remember right.:)

Paul D. Buck
Paul D. Buck
Joined: 17 Jan 05
Posts: 754
Credit: 5385205
RAC: 0

RE: Why not use both like

Message 90718 in response to message 90716

Quote:
Why not use both like Folding@home?


Eventually we will get there ... Folding has had years to get where they are now.

OpenCL should make this easier, maybe, ...

But more likely what will happen is that we will find OpenCL is like the non-optimized applications and that on those projects with open source that ports to the specific APIs will produce applications that will be even faster than the OpenCL applications.

astrocrab
astrocrab
Joined: 28 Jan 08
Posts: 208
Credit: 429202534
RAC: 0

RE: That's a pity, as the

Message 90719 in response to message 90705

Quote:
That's a pity, as the ATI HD 38xx and 48xx series seem to be good at double precision

and what is about new 4770 with almost 1Tflops?

going to buy one or several for E@H and MW@H.

Paul D. Buck
Paul D. Buck
Joined: 17 Jan 05
Posts: 754
Credit: 5385205
RAC: 0

RE: RE: That's a pity, as

Message 90720 in response to message 90719

Quote:
Quote:
That's a pity, as the ATI HD 38xx and 48xx series seem to be good at double precision

and what is about new 4770 with almost 1Tflops?

going to buy one or several for E@H and MW@H.


Unless you want to have it to play games might as well wait till EaH actually starts to issue GPU work. Last month there were statements that applications were in the works, and more recently that there were not ...

MW is having difficulty keeping up with demands and there are a lot of complaints about that. If Travis is off coding a GPU application that actually explains that a little bit. In the mean time I would wait to see what happens.

astrocrab
astrocrab
Joined: 28 Jan 08
Posts: 208
Credit: 429202534
RAC: 0

RE: RE: RE: That's a

Message 90721 in response to message 90720

Quote:
Quote:
Quote:
That's a pity, as the ATI HD 38xx and 48xx series seem to be good at double precision

and what is about new 4770 with almost 1Tflops?

going to buy one or several for E@H and MW@H.

wait till EaH actually starts to issue GPU work.

i waiting for this several years already! =)

so what about double precision at 4770?

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.