If the pics have been 'shopped, they're good ones. The other two (close ups) on the TPU site look pretty authentic.
Agreed on the lack of rationale behind the concept (performance, cooling, power draw, price etc.), but I don't think Asus are using the same metric as most right-minded people- hence the original Mars and the Ares. And you just know this is a PR stunt like the previous two...$1k+ , limited edition 1,000 made, >>>Ladeeeeeeeesss and gentleman, roll up, roll up and see the amaaaazing Marrrrrrrs 2. It's bling, It's king, It'll make your ears ring!<<<
Somehow I doubt this will be done for anything other than an "Asus has proved that it can be done" moment, not volume production or economic sense.
From a performance level it would kill a GTX 460 SLI setup, or pretty much any other multi-GPU setup -possibly excepting a CrossfireX HD 5970 , thanks to SLI scaling and the onboard NF200 chip as opposed to the bandwidth limitation imposed by SLI and Crossfire connectors.
Could be the first card that would actually be constrained by the bandwidth of PCI-E 2.0 spec.
It still makes no sense to 99.9999% of PC component buyers obviously, and like the Mars and Ares is guaranteed an exceedingly short reign as the ultimate desktop card. The GTX 460X2 is obviously "do-able", I think will be released and if priced accordingly could sell in a reasonable (for enthusiast cards) volume.
I would think it's still going to suffer from the 256Mb frame buffer in comparison with the 384 and 320 of the GF100 cards at very high res with a heavy tesselation/AO/AA workload.