To Smegz...I think this is a horrible start. It is painfully unecessary to slap two GPUs on a board when the dual core technology is present. I agree with Tim...Red and Green are just buying time and ripping of consumers with old technology.

Well my 7950Gx2 at the time and still to this day performs flawlessy with very high FPS. Bar Crysis no game has yet dropped it below 30fps. 

In the days when 7900GTX's stuggled in Oblivion and the 7900GT's were being RMAed in their hundreds - the 7950Gx2 shone above the rest.

While inelegant and problematic, the multi-core/die future is the way things must go due to die & package space limitations, so best to play around with those options more and more now with known quantities than trying to experiment with new architectures like the R700/G100 etc. 
With the potential of the many core verse coming sooner than you think, the elegance will need to be found in the software side as well not only the hardware side because the hardware side is highting barriers of physics which are harder to surmount.
If AMD can apply ATi's old super-tiling solution more effectively they may end up with a more flexible solution than just variations on AFR and SFR. This may also help with the issue of copying the frame buffer to all memories involved, which is a small barrier with solutions that won't find themselves easily within the same package on the PCB with shared memory pathways. 

While I personally don't like Xfire/SLi as solutions to a problem that is usually restricted to uber-resolutions / settings and e-Wangs, there are some unfortunate realities we are coming up against with our traditional 1 chip solutions, that need better answers that hopefully these Franken-SLi'n can initiate somehow.

In retrospect the biggest drawback of the old GX2 was it's apparent limited ability to give nVidia any insight into a great than 2 GPU future as demonstrated by their slow progress with the GF8800s. Hopefully these new X2 solutions offer more positives for both AMD and nVidia in perfecting their future products.

PS, Matrox Rulz!! >B~)

? I agree with Jay above; how is one good and the other bad. You tech journos lauded the processor multicore trend and now poo-poo multi GPU's? What is it a case of "been there, done that?" You want something new? Make up your minds and stop whining about this "not being a real innovation."

Having said that, I believe it would have been better for the graphics makers to do EXACTLY like the CPU makers (albeit AMD makes both). The should have made dual-GPU chips and not simply put 2 chips on a board ala Intel (see the first dual core offerings and current quad core offerings.) I will say that this is a good start though.

It is painfully obvious that multi-core is a better solution than multi-GPU. ATI / NVidia both know that. Other than trying to squeeze out more money from the consumers, there is no point to linger around the multi GPU idea. 

I am getting sick of being fed these new cards that are just slight revisions of old technology. NVidia is stalling just because they can, and ATI is a generation behind. It isn't realistic but we need a 3rd company to step in and kick out some new GPU technology while ATI and NVidia's has their head are up their @$$. 

2008 will be a boring GPU year if it is just going to be a multi GPU race. Woopty Freakin Do. New technology please...it was due yesterday.

I know "Pc gaming is dying" has been said before but ....

I've been a PC gamer for about 3 decades now. Up till now, every Xmas, there was always a huge surge of titles as companies rushed to try to get their stuff out. A lot of times we didn't like it because they wound up ruining games by releasing them before they were ready, but still, there was a lot of stuff...

But not this Xmas. I've never seen a Xmas with such a paltry few releases. And looking ahead, it doesn't look any better... in fact, the PC release schedule looks frighteningly slim to me.

As an earlier poster already said, you can buy an XBox or a WII for the price of a top end graphics card these days. And, the consoles just work --- no defragging, updating antivirus or drivers, etc etc. And yes, the graphics are finally good enough (In fact, I felt the PS2 graphics were good enough.)

Finally, the whole DX10 and 10.1 fiasco must have seriously pissed off the early adopters... are you willing to shell out $500 for a 10.1 card, when dx 10.2 may pop up if they discover any suffciently fatal API flaws?

I really hope this doesn't come to pass but it's just not looking good...

I don't think the new approach is misguided

in the HD2XXX generation, AMD, just like nVidia, eliminated the midrange, concentrating only on high-end and low-end

with the HD3XXX generation, AMD dedicated itself to the mid-range and the low-end, leaving the high-end to unelegant and impractical dual chip solutions

I think this second approach is the better one, as low-end and midrange is where the real money is: high-end is only for headline grabbing and consumer mindshare, which are also important but can also be acquired through other means (e.g. through a solid advantage in the sub-$200 market)

just look at valve statistics and see what kind of money people are spending in their GPUs

Nice comment article. But here's what I think is really killing the GPU market.

HDTV and HD movies and HD gaming. Xbox360 and PS3 produce HD graphics that are 'good enough' for most people. The cost of a premium video card for gaming is about the same as a PS3 or an Elite 360. That's ridiculous.

Mainstream cards in the $50 to $150 range are what is needed here. Building a dual, triple or quad GPU solution for gaming is an expensive way to make a game run. What the hell is the point of game devs killing themselves to get DirectX10.1 working when a fraction of the audience has that capability. Why bother making a game that uses so many effects that it obliterates decently performing systems forcing users to upgrade to play a game.

PC gaming has a future if, and only if, game developers start programming to the mainstream. You shouldn't have to add a video card to a standard PC merely to play a few games on your PC. The implication here is that ATI and Nvidia have a future if they can manage to deliver mainstream performance cards that don't cost an arm and a leg, and convince system integrators to offer systems using them instead of integrated graphics. At the same time as this, they have to convince game developers and publishers to stop shooting for the 1% of PC owners who are prepared to drop $500-$1000 in order to play the latest $50 game at 60 frames per second.

We've been hearing this bull since the Radeon 8500 came out in 2001. Intel, ATI, and AMD have all made some serious mistakes, but you try to make NVIDIA's sound fatal. Why not eat a whole lot of humble pie and make nice with the Green Team? They keep winning, and you keep losing out to Kyle, Anand, and Tom's crew. ATI won't be your John forever.

I guess they do the SLI/Xfire thing to claim top honors but very few people buy them - and the drivers dont work right - give me a single GPUany day (until the drivers get better).

Intel and Larrabie? What is this - their 5th, 6th, 10th attempt at a graphics solution? Intel will never succeed because they are driven by committee's. Zero vision. NV goes through serious pain on their drivers - have you seen Intel drivers? Some of the worst on the planet...In the end it will be just good enough to be sucked into a chipset so we can play DukeNukem (the original) at 10fps.
(800x600).
Cheers

why did daamit not use hypertransport between the 2 GPU's? After all they do have some experience with that transport bus. I guess the 2 engineering teams, cpu and gpu, have not been talking to eachother. Is there still some friction in this marriage or do they just not care?

Huh? As the die shrink process gets better and better, the trend will be to multi GPU cards, in a similar fashion to multi core CPUs. And the 7950GX2 was not infamous. It was no better or worse an option than any other SLI or Crossfire setup. The drivers were what limited the hardware.