>> Varisys has past experience with PA6T and MPC86xx, which bplan may not have.

> I was trying to point out that the chosen CPU is a mistake.

That's why I also mentioned Varisys's experience with MPC86xx. Or would you consider that choice a mistake as well? If yes: What Power Architecture CPU wouldn't be a mistake in your eyes?

> Powerful (for what?)

See my posting I linked to for some examples.

> already at its end of life

Probably there won't be any other PWRficient CPU than PA6T-1682M, yes. But that wouldn't stop A-Eon from using a newer Power Architecture CPU (QorIQ P5 for instance, see what Varisys responded to Jim's inquiry) in an X1000 successor (should there ever be one, that is). I think PA6T will continue to be manufactured for some years to come, at least until mid-2011 (to mid-2013 maximum) on behalf of Apple, and on behalf of someone else after that.

> Where is this CPU used in volume? I thought it was kept because of some
> contracts with US army, that have systems that use it

Yes, with US DoD (that includes army as well as navy and air force), and exactly that's where this CPU is used in (rather low) volume (see Zylesea's reply). It seems you answered your own question ;-)

>> In September 2008, Apple started volume production of the chip again.

> Sure? Would love sources about this.

See the link in:
https://morph.zone/modules/newbb_plus/viewtopic.php?forum=3&topic_id=7001&start=127

At least in July 2008 that was said to be the plan. I don't have any reasons to doubt that they did what they said they'd do.

> really, that would explain why the specs can't be the ones
> chosen by a sensible engineer.

As I wrote, the exact specs were decided not until one year after Hermans had his rough idea. According to A-Eon, it was the OS4 development team (it was not said who in person) who decided the exact specs. (Mind you, A-Eon was founded in April 2009, i.e. some months *after* Varisys had started the board development.) With one exception: As you might remember, A-Eon said that the suggestion to include the XCore came from Varisys.
Unfortunately, we'll probably never know to what extent the May 2007 announcement of Amiga Inc.'s and ACK's 'Power Design' had an influence on the design decisions for Nemo/X1000 ;-)

>> AFAIK, bplan don't have any past experience with the PA6T.

> That's my way to insist in taking that fantaaaaastic CPU out of the picture.

Same question as above: What suitable Power Architecture CPU is in *your* picture then which bplan has more (or at least the same) experience with than Varisys?
There's really only one coming to my mind: the PPC970(MP), see the TetraPower/Bimini board. But I doubt that the PPC970MP, which is not a SoC and thus requires the CPC945 northbridge (50 to 100 USD), would be a better solution than the PA6T, neither technically nor economically. To quote Neko:

"The BOM cost for a 1.5GHz quad-core board was over $900 (reaching 750 EUR at the time!) including processors for the complex dual-sided PCB. For >2GHz it hit $1500. Genesi would not sell a single unit at that price. Given that it had to use expensive RAM, expensive disks, weird cases (BTX pretty much got abandoned halfway through).. we would have to sell systems at the same cost as an Apple G5 (nearing $3000).. which defeated the whole object of the project."

Using only one instead of two PPC970MP CPUs wouldn't save that much, 250 to 300 USD at 1.8 GHz and 200 to 250 USD at 1.5 GHz.