GPU support for ATI cards

I hope a solution can be found to this as it will not be worth my money in buying new cards cos of this reason, just wasted my money buying a ati HD4890 cos I thought it worked with seti, silly me I should have asked b4 buy.
____________

No, BOINC 6.10 will be supporting ATI GPUs.

Whether to use CAL or OpenCL is something that projects have to decide on, on the application level. It's not something BOINC has to support. (got that info directly from Rom, so he'll probably know what he's talking about. ;-))
____________
Jord

- BOINC FAQ Service
- BOINC User Wiki

Real is just a matter of perception.

David206180, unless you are one who just does SETI, you might look toward other projects. I know some of my team mates are running GPUGrid on ATI cards. Not sure if there are any more out there. You might look around for a project that interests you as a backup that utilizes your ATI card.
____________


PROUD MEMBER OF Team Starfire World BOINC

I know some of my team mates are running GPUGrid on ATI cards.

Sorry bout that, you're right Jord. Our team runs so many different projects it's hard to keep up with them.
____________


PROUD MEMBER OF Team Starfire World BOINC

Hey, given Dx11 ATI card's are expected to be out in September/October (given the october windows 7 release), and dx11 nvidia parts arent expected till next year q1 or q2 no update on which yet, it's likely I and many others will be getting ati card's, anyone got an update on upcoming OpenCL support like an eta, twould be nice to have support in place by end the of October, so seti doesnt miss out, else i'll have to run folding on it till ya get it released?

(also all current CUDA or CAL compatible and future nvidia, ati and intel larabee when it comes out gfx cards will support OpenCL in response to DanielT's q's, both nvidia, ati and intel are supporting it)
____________

The biggest problem for now is that there is totally no example code available to program anything in OpenCL, so if a project wants to use it, they would have to do it from scratch. Quite a learning curve.
____________
Jord

- BOINC FAQ Service
- BOINC User Wiki

Real is just a matter of perception.

You would need to ask the OpenCL consortium as to when they plan on finalizing the standard. Then its a matter of SETI affording the programming tools (assuming they are different) and the time needed to actually write the code for the program.

well given nvidia's support of OpenCL or Intel, all I can suggest is contact Nvidia about an OpenCL client or some example code at least and also contact Intel, who may be intrested in some extra PR about their forethcoming Larabee platform in 2010.
____________

Intel has already contacted the BOINC developers, but on a whole different kettle of fish. And to be honest, that's about all I can say about that. Keep an eye on BOINC news for that. It'll be exiting.

Now then, since Seti is open source, why not contact all these hardware makers yourself and ask for help on porting the Seti source code to OpenCL? CUDA was being tested for well over a year by developers all around (not necessarily BOINC) before it was started at Seti and GPuGrid. it was started at Einstein at the same time and there you still can't use it as they can't make it completely stable. It had time to get somewhat mature, but it is nowhere near out of beta stages yet.

OpenCL isn't anywhere near being mature, else it would be quite easy to get example code. Why hurry and make mistakes? Why not wait until example code is readily available, with which applications can be built without too much trouble?

It's not as if the developers (both BOINC and here at Seti) aren't busy enough already, now you want them to do even more?
____________
Jord

- BOINC FAQ Service
- BOINC User Wiki

Real is just a matter of perception.

I have neither the contacts nor the necessary time, nor the necessary technical expertise or understanding of the SETI analysis app required to make any enquiry of a technical nature or relay any such information in professional manner.

Additionally Ageless please don't give me the oh dear its open source so it ought to be lesser then proprietary software cos our poor overworked staff can't deal with it stuff.

That's exactly the kinda crud I heard when I made the post about supporting GPU processing in the first place, along with the prize tidbit of information from someone claiming comptency saying SETI would run SLOWER on gpu's then cpu's "as its the wrong kind of processing for it" and be like a old 1ghz p4 per.... I did at least have technical understanding of the functions of it to know that was bupkis, but it seems being given a fully written app by nvidia turned out to be what was necessary to dissuade that opinion.

And if by all accounts SETI project's own Dev's didn't actually make the CUDA client, then they're ought not to be as overworked as accounts have made out, open-source software should be superior to close-source not the other way around, the whole point of everyone being able to contribute, if you haven't got enough people maybe ya just not advertising the requirement for volunteer's enough, do so, given a plethora of newly work students with degree level comptency(well ok yes that means beginner with many students, but at least some have to be half-way competent) with nothing better to do I'm sure there are plenty of available, in the UK its expected 20% of the graduating class this year will be out of work, i certainly have heard no mention of recruiting, other then when requesting new features or more accurately information about them, given that creating them should be a given due to the obvious benefits... to spell it out...

Why its simple enough, the vast majority of consumer PC users of xp (and vista)(given xp's no longer mainstream support) will be upgrading to Windows 7 within 6-12 months of release many in the first 6, either with new machines or upgrades, out of these a large proportion will be tempted and a hefty number purchase the new dx11 ati card's along with it, due to the no doubt the many gloating we have dx11 you don't advertising campaigns one can expect to be released from ATI closer to release.

This means just like nvidia when it scooped ati on dx10 in early days, many will be buying ati this time around with ati nearing the next generation whilst nvidia finally produces a dx11 card in 6 months time, their dx10 and dx11 offerings will both be/are better value for money and generally superior in gaming that means ALLOT of dx11 ati card's can be expected to be in use.

And any BOINC project that can capitalize on that will get a significant boost to the available processing power that can be brought to bear on it, which frankly the whole point of BOINC in the first place, to generate the maximum amount of processing power per project possible.

Waiting until a technology is aged is not the way to go at this time, dev's should be trying to recruit as many volunteer folks as possible figure it out for themselves and get the darn thing running asap to capitalize on it in time for Windows 7 release.
____________

Additionally Ageless please don't give me the oh dear its open source so it ought to be lesser then proprietary software cos our poor overworked staff can't deal with it stuff.

And if by all accounts SETI project's own Dev's didn't actually make the CUDA client, then they're ought not to be as overworked as accounts have made out

Waiting until a technology is aged is not the way to go at this time, dev's should be trying to recruit as many volunteer folks as possible figure it out for themselves and get the darn thing running asap to capitalize on it in time for Windows 7 release.