Folding Forum

Rumours abound that Mike has been picked up by AMD/Ati:
http://www.rage3d.com/board/showthread. ... 1335259657

What does this mean for the already stilted GPU client?

I clicked one more link on your link- rumor confirmed.
https://graphics.stanford.edu/~mhouston/

Original news source: http://www.amdzone.com/index.php/home

See last line in first paragraph of Mike's Web Page for an answer.

https://graphics.stanford.edu/~mhouston/

And we're still waiting for a new client that support last ATI boards ...

Maybe the problem is working out how to get certain features of the GPU to work with the hardware. Now he is with AMD, he could choose to liase with Pandegroup and create code, or offer solutions, that could solve known problems.

Maybe the GPU stuff is going the way of the PS3 and he is about to become a project leader.

Either way i wish him luck, and would hope the rest of you do as well. Life doesnt revolve around FAH, even if some of us think it does.

Mike's been a great help on the GPU client and GPU's in general -- he's very smart and knows his stuff and has worked really hard to help us out. But the client itself was primarily written by people in my team, with Mike's help. With Mike at AMD/ATI, he'll be able to continue to help us out, so I am optimistic that we'll see great new results from that continued collaboration. On our side, Adam and Mark are still our GPU gurus and are working hard on getting the newest client/core out very soon.

People will continuously be leaving Stanford (that's the way universities work), but the upshot is that we've been able to get continued help from many *and* get a new crop of bright young students to take up the next challenges.

An observation:

Intel foots his bill for 2 years, one of only 40 or so people annually in the world, and he goes to AMD.

Interesting.

I will be really good for the next phase of whatever happens to the FAH-Brook-ATI client since he'll know the issues from both sides of the street.

The other question is what does this mean for a future FAH client that runs on nVidia? Presumably there are still people at Brook that can support the PandeGroup's needs, with nVidia, but that's an assumption, not an established fact.

Im not entirely sure i would ever run a Nvidia card on GPU. Since it ties up a cpu/core, and running SMP would seem to be of more use.

Besides i honestly dont see the Nvidia client surfacing in the next year. It may be a pessimistic view, but the way the current client works is ok for the most part, but patchy at times. I think they should create a stable ATI client first, before they start thinking about a different manufacturer. Once that is done, if they want they could try and port it to run on Nvidia cards, and maybe onto Linux. But for now, just get a stable client for the current GPU crowd.

as far as I understand it, technically all the gpu f@h client code is standard plain vanilla directX9, although I would think that a least a fair bit of the code dealing with direct caclulations is in assembly, or whatever counts for assembly on these cards... unless some of it is offloaded to the cpu...

the hangup for nv used to be a small (or large) ((un?)documented)) bug in nvidia's gpu's themselves (not the firmware, but the actual gpu die itself), at least from the perspective of geforce 6800's - geforce 7900's, I have no clue if the same holds true for the G80 based systems though, but the lack of news about it might at least hint that it shares the same flaw.

btw, good luck on your new job mike! thanks for all the hard work at making the impossible possible!

I was glad when I heard that mike went to AMD. Should be really good news for GPU folding(if he continues to contribute fah project

wish you the best luck
cheers

Karova_Milkbar wrote:Intel foots his bill for 2 years, one of only 40 or so people annually in the world, and he goes to AMD.

I noted that too. Maybe AMD is hungrier.