One thing you have to understand is that when E3 2005 was approaching, there was no such thing as a PS3 prototype in existence, anywhere. The console had been redesigned from the ground up a few months earlier and none of the integration between Sony's and Nvidia's silicon was remotely demonstrable, because there wasn't any except in enginering meetings. They were vamping like crazy.
Some of you might recall that Nvidia's participation in the PS3 was announced very late in the development cycle compared to similar announcement for competing products and Sony's own earlier consoles. This is because the PS3 design originally had no dedicate GPU. Remember the demos they showed where they made a big deal about this being the Cell all by itself? Those are left over fromt he original PS3 concept.
One of the still unrealized big features of the Cell is the ability to gang together multiple chips and have the resulting system achieve mcuh higher efficiency than typical SMP systems that have greatly diminishing returns on throughput as you add processors. (Most such systems run multiple OS instances dividing the system up into several virtual machines rather than try to do one big OS instance across 32 x86 CPUs.) Sony even put forward fantasy scenarios where your PS3 would enlist processing power from idle appliances on your home network that also contained Cell chips. (A mightmarish prospect of unporedictable resources from a console developer's perspective.) The PS3 was going to have anywhere from three to a half-dozen Cell chips whose power developers would allocate to different tasks as they desired. If you wanted more audio processing you could have it so long as you could spare some threads from another operation like pixel shading.
This would be the mahine to beat all others because it essentially allowed developers to design their own machine as they saw fit . A wonderful dream but there were a few problems. IBM as of last summer was nowhere near ggetting the automated cooperation fo Cell clusters working. In fact, according to a quoted IBM Engineering VP in EE Times, they only had Cells working in sets of two. Not exactly a leap for the processor industry.
Also, IBM couldn't make the damn thing reliably. There was no way enough Cells could be produced any time in the next three years to make a consumer product using multiple Cells possible. Unless the price would be in the thousands of dollars. So Sony had to go back to the drawing board and go with a more conventional design using a partner supplied dedicated GPU alongside a single Cell in each system. Starting from scratch on a GPU design would have taken years so Sony got Nvidia to modify one of their upcoming PC products. Sound familiar?
Dual monitor support was a given for any new high-end Nvidia chip and making that HDMI for the purposes of an E3 demo was no big deal. While Sony ran demos solely for the Cell, Nvidia would trot out a bunch of PC demos and say this was what they were bringing to the PS3. (Funny thing, when the Sony and Nvidia partnership was first announced they made sure to say the resulting chip would be unique to Sony's console and not a modified PC design. This turned out to be true only if the names were changed to Microsoft and ATI.)
Sony was desparately in need of demo fodder last year and thought driving two HDTVs simultaneously would wow the crowd, even though this had long been doable on PCs using DVI. Actually going tot he expense of having this in a console is ridiculous. So, like so many times in the past, once it wasn't needed anymore the demo feature was quietly forgotten.