There is vast amount of people that wanted to go multi-gpu route, but refrained from doing so because they could not use the second monitor. That changed now with Nvidia, but CrossFireX is still missing in the action, and it’s been almost two years since ATI first demonstrated multi-monitor capability to the press. Well, can’t have it all.

I think we can all conclude there is roughly 1M premium customers every year, even though I would say that number at least doubled in the past two years – my sources told me what manufacturer ship monthly 30.000 high end units in EU and 40.000 high-end cards in NA markets alone. If you go and ask Dell how many 24″, 27″ and 30″ screens they sold, the number is in millions, so there is a market there as well.

But, SLI and CrossFire have to do away with the image of person owning that configuration being a second grade customer, even though he’s a premium one.

the amount of problems that multi-GPU owners had in the past is just incredible, those ppl buy AMG Mercedes, BMW M series, RS Audis… and yetl, the amount of quirks which they go through is well… owning a Daewoo/Chevrolet comes to it.

A lot of SLI customers went away after being burnt, same thing with Crossfire. the problem is, neither of the two were serious enough with the support, and the amount of broken promises is just about enough for anybody, yet alone for ppl that spent thousands of dollars on the GPU part alone.

Steam may not be the best survey to find how many multi-GPU configs are out there, but it is closer to reality than any other.

I think that we have two architectures emerging in the GPU world, and that is something I will address in future interviews with GPU architects.

And I don’t think either of them is wrong.

It’s tough right now for them. AMD too. Times are changing! Even Microsoft is getting screwed up, and Intel has to be looking at IBM at least a little. Buyer’s market, right?!

Multi-GPU is important and it should be important. But, how exactly to deploy multi-gpu that makes sense, that’s whole another ballgame. I find it worrying that out of numerous millions of people that have Steam, only 300K use multi-gpu configuration.
Only reasnoble answer is that multi-gpu wasn’t accessed right. It was absurd to talk with Nvidia and AMD about multi-monitor support on their configurations. No, it was in their heads that anybody that has multi-gpu has 30 inch screen and nu mas. As expected, this approach failed miserably, given that power users have more than one monitor. And in case of some, especially WoW gamer, companies failed miserably. WoW scaled great and is the ultimate high res app, yet the companies didn’t pursue the 15M+ market with multi-gpu products optimized for them.
On the other hand, Skulltrail makes sense for all those situations when you’re CPU-bound. For instance, run Badaboom transcoding with GeForce GTX280 on single, dual, quad and octal core, and scaling will be there. Same thing in gaming – some games+resolution combos require as much cpu as possible, for some particular reason. What the market needs is scaling company such as LucidLogix – if we get 90-100% scaling, multi-gpu is cool.

Perhaps this is the greatest failure of Nvidia – company has nForce 100 and 200 chips and doesn’t offer performance advantages, in fact you lose performance on the account of added latency. We’ll see.

Probably everybody in AMD and a lot of people outside the company know what is the name of exec that killed the platform for minor savings, pissing off one huge motherboard vendor that heavily invested in development of their own FASN8 board.

I think for NVIDIA, it is pretty scary how bad multi-gpu is doing. Obviously, that’s why JenHsun is so excited about ION!
http://www.edbordenblog.com/2009/02/nvidia-shift-in-relevance-away-from.html