The FPGA it runs on is pretty pricey, though, at ~$320 (from what I can tell), so I haven't bothered trying to run the images yet. I'll likely wait until the hardware exists, and then spend way too much money to have a real looking/acting C65 that also happens to work as a C64. I suspect Amiga and ST bitstreams will eventually show up, as well, since lots of work has already been done there on other FPGAs.
I am not all that interested in emulation (whether hardware or software), and have somewhat limited interest in retro computing (I own a Commodore 64 and a GameBoy for music composition, but long ago sold all of my Amigas and other stuff), but the C65 holds a special place in my heart (and, I imagine that of many C64 kids). It's like a piece of alternate history. Apple fans got the IIGS, but C64 lovers never got our next generation. The C65 would have been that next generation, had it ever gone past the prototype stage.
And if I remember correctly, most FPGAs of that generation and market segment do have differential outputs (though probably not SERDES hard blocks), but can't quite pull off proper HDMI/DVI without an external encoder chip. There might be some way to hack a "good enough" signal for lower bitrates, but if you're going to need to half-ass the signal either way, you might as well go with the simple analog link and not the complex digital one.
1920 x 1080 x 24bpp x (10/8 due to TMDS 8b/10b encoding) 60hz ~= 3.5 gigabits a second, / 3 (3 data lines in DVI/HDMI) = just over a gigabit per data line, plus surely some overhead that I don't know about. I'm not an EE but it sounds like it would be hard to generate a >1GHz signal on a <1GHz FPGA.
Whereas with an analog signal driven by 3 8-bit DACs, you're dealing just dealing with 1920 x 1080 x (VGA overhead) * 60hz = a 120~ megahertz signal that can easily be generated by any FPGA. (the 24bpp is irrelevant because it is being transferred in parallel)
For a resolution of "1920 x 1080 @ 60Hz" the FPGA needs to run only at the pixel clock's frequency, which is "only" 124.416 MHz (1920x1080x60 / (10^6)). Then on every pixel clock cycle, the FPGA needs to present each bit of what makes-up-a-pixel on it's output pins. In case of 24bpp, that would require 24 dedicated output pins. In addition, a dedicated DVI transmitter chip would be required to serialize and encode that 24 wide bus to those DVI-compliant three differential signal lines in addition to doing the 8b/10b encoding. Roughly.
Or the FPGA could have a high-speed serializer built-it which could replace the dedicated chip.
~sigh~ I wish I knew enought about electronics to play around designing such a serializer, it tickles my fancy :)
Remember the original hardware being emulated targeted PAL TVs: 576 lines @ 50Hz.
Lower resolution HDMI is certainly feasible, but you have to care more about the signal integrity when designing the board, and it's harder to debug: http://www.fpga4fun.com/HDMI.html
Something I didn't consider when I posted was that a lot of people will want to use this with a VGA CRT, especially since there is lots of Amiga software (including the Workbench itself) that changes the pixel clock midframe, which would pretty much require upscaling hardware to produce a signal that a modern LCD would accept. VGA+a CRT requirement becomes a much simpler solution in light of this.
http://www.colorforth.com/video.htm
Also see: [1] http://www.colorforth.com/haypress.htm which is in reverse chronological order.