> If you can't even lift parts off the shelf and press out a board for a PPC CPU
> except by years of hardware hacking or paying stupid amounts of money to
> design custom parts, then PPC is dead.

As I (and Alan Redhouse) wrote, reliable PPC northbridges were commercially available at the time (October 2000). No "years of hardware hacking" or designing "custom parts" required. It was their decision to go the risky way by trying to design custom parts to supposedly save some money. This unfortunate decision made the released product more expensive in the end (because the original work was done in vain and a completely different, more expensive board was used). The safe way would have been to design the Escena board around a commercially available PPC northbridge like the MPC106/Tsi106/MPC107/Tsi107 (as used by Apple), the CPC700/CPC710 or the GT-64130 (ancestor of Pegasos II's northbridge). (GT-64260/MV64260 and Harrier only became available later in 2000/2001.)
Economic considerations aside, the problem with these commercially available PPC northbridge chips were the lack of AGP support (as mentioned before) and the lack of support for the faster proprietary protocols for southbridge connection (which would have required fallback to slower PCI connection). So yes, the situation for developing PPC desktop machines was definitely worse than for the market leader x86, but it was still good enough for a PPC "Amiga" machine succeeding m68k Amigas (even those with PowerUP accelerators). The northbridge of the Pegasos II, for instance, also lacked AGP support (hence its fake AGP derived from PCI(-X)) and had the southbridge connected by slow PCI, but obviously this didn't mean a reliable product for the desktop market couldn't be built around it.

> Schuler hadn't put any chips on his board.

Exactly. He could have if he had designed the board around a commercially available northbridge chip instead of trying (and failing) to design his own custom northbridge chip.

> he would have had to completely redesign the logic of his board or
> design his own chips to accommodate it.

Designing his own chips for the board was what he was doing constantly and where he finally failed at and what made the Escena AmigaOne fail. Basing the board around commercially available chips instead would have increased the likelihood of successfully completing the development by a great margin.

> Schuler encountered the same issue everyone else has in using it.

No other PPC board design company (except Apple) attempted to develop its own northbridge chip instead of simply basing their board designs around commercially available northbridge chips. Apple succeeded in doing so, Escena tried and failed, everyone else successfully used commercially available chips (including Apple for all G3 Macs, the PowerMac G4 PCI and all G5 Macs).

> Because PPC is dead.

It wasn't in 2000, no matter how much you wish it was. The notion that PPC was dead back then because some guy unnecessarily tried and failed to design his own custom PPC northbridge chip is beyond stupid.

> The Efika and Sam were at the absolute limit of CPU power required
> to run a desktop system.

Yes, I would have never bought one of these slouches for desktop use, not even when they were new.

> Freescale's later SoCs were even less than that.

I especially remember the enthusiasm around the MPC86xx SoCs (and Genesi's plans to use them for new Pegasos and Efika boards) when they were new, also here on MorphZone. These were certainly not less than the MPC5200B or the PPC440EP.