Hi there!
Yep, Elf is right. The issue with the D1 was getting suitably-fast remote display from the card, to the O2. However, it was NOT a bandwidth limit - I did a lot of benchmarking at the time and it virtually never saturated the network connection. The issue was a combination of a latency problem slowing down X11 (even with no routing between the SGI and the SoC, still slower than X over a Unix socket), and that X11 on the SGI was not really fast enough to deal with full-screen bitmaps from a "normal" remote desktop application without the use of OpenGL or other methods of leaning on the graphics hardware, which at the time I didn't really have the knowledge to do properly.
In terms of what the advantage is over just having two computers, it was supposed to allow for a more-integrated user experience, where applications on the SGI and on the SoC could share a desktop and a filesystem. That part worked fine - it was just that display was slow.
I have since written and tested (on my Tezro) a proof-of-concept for OpenGL-accelerated remote display, where the display is sent as a bitmap and rendered as a texture on the SGI, taking advantage of OpenGL for fast bitmap output on IRIX (bypassing normal X11 bitmap handling) as mplayer does when using -vo sgi. It works, and it's fast, but it's not something I've really pursued returning to the D1 to work off of.
If there is still interest, I would still consider working on it - I'm pretty confident it could be made to work now, and I'm a much better programmer and board designer now than I was when I started work on the D1.
However, I will note that one limit to the D1 as planned was that it REQUIRED an O2, not just an SGI with a PCI connector. IRIX only loads the drivers for the TNETE100APCM ethernet chip on IP32. I confirmed this with a 9210105 and, I'm pretty sure, a D1 prototype (which did exist - I still have one in my O2) in an Octane shoebox. I still don't know nearly enough about how IRIX handles drivers to work around this. I also considered, at the time and now, an FPGA-based PCI or PCI-X "video bridge" between the SGI and the SoC, bypassing the network in between altogether, and that's probably the more elegant solution, but again, I definitely don't have the IRIX driver development knowledge to make that happen without a whole lot more research.
Hope this sheds some light on what happened to the D1. It's not dead per se - I never decided or announced that I would discontinue work on it, and I've gone back and done a handful of tests since I last discussed it around 2018, but it hasn't really been worked on as much as it was being for a while. I think I have a better path to making it work the way it'd have to to be any good, but that's still a proof-of-concept and I haven't pursued it too far. If it turns out a lot of people still want this, even with the leaps and bounds that have been made in IRIX software porting lately, I'd absolutely be open to picking it up again.
- Dodoid