Could something like this

T2080 board
http://bit.ly/1NODbuO
be stuffed into a PC mobo or maybe even a passive
backplane and be made useful??
maybe even for morphOS , or ppcLinux ??

basically Im thinking of the concept of the A2088 xt bridgeboard
only on a modern scale
I'm going to presume though,the expense involved would be not feasible

Quote:

Well, that isn't really an alternate ISA question (since the T2080 is a PPC processor), but I think I understand what you are suggesting and the short answer would be simply, yes.

PCIe processor cards have already been created for that processor and similar cards have been produced in the recent past for many other PPC CPUs (like the Cell BE and it's successor).
These are generally designed for passive backplanes, but that is only one step away from using it in a multiprocessor system.

The current interest in old Sonnet Allegro PPC expansion boards for the Mac 7200 is a good (if dated) example of this kind of thinking.

If we are talking about MorphOS specifically, then a system with both an X64 and a PPC processor becomes even more attractive.
It could enable the concurrent use of both ISAs without the need for emulation.

And for diehard Amiga fans,there would be the possibility of running Amiga Forever with true PPC support.

Potentially, with enough support for virtualization, any combination of PPC OS' and X64 OS' should be possible.

OK, where does this lead us?
Well, we have actually discussed this idea before, but since it was not one of the multiplicity of options recently rehashed here and at other Amiga sites, I'm leaving you ALL the credit for bringing this one back into the mix.

So, why don't you contact Bill at power2people and mention that there is yet another option to consider?

Or, you could PM Andre Siegel and ask him to relay the idea.

And, of course, if it's adopted as a community developed idea, just your mentioning the concept is productive.

> Allegro

Crescendo :-)

https://morph.zone/modules/newbb_plus/viewtopic.php?forum=11&topic_id=4368&start=26

Sorry Andreas, posted from my cell phone without checking on the correct name for the boards people have asked me to locate.
As far as I can tell there aren't any guaranteed reliable Sonnet boards for sale.

Anyway, I am glad that is the only mistake you have caught (at least so far).

BTW-What do you think, T2080, T2081, or T1042 for a PCIe add in card?
I have ranked them best, lower cost without unnecessary SATA controller, and significantly lower cost alternative.

A PCIe expansion card could be slightly less costly than a full motherboard, but it would require more work by the OS developers.

> What do you think, T2080, T2081, or T1042 for a PCIe add in card?

This may be good for development purposes, but as a mere end user I prefer standalone solutions.

IF those board/cards can be produced at reasonable cost
IF those board/cards have sufficient performance
IF those board/cards can work standalone/with passive backplane
IF the MorphOS-team supports them
IF those board/cards can work in a x86 mobo
IF the MorphOS-team actually goes ahead with an x86 port
IF they build the initial port in a way that coulc use the card for an hosted "ABox"

Then I would buy a lowend/silent x86 MiniITX board and the card.

[ Edited by Kronos 06.12.2015 - 15:41 ]

Hey Kronos, that is a better endorsement then I've gotten from some of the developers (not to mention the posters here) except for the few die hards that agreed that IF it performed better than our current hardware...

But, I'm still unsure how this approach would work best.
A T1042 at 1.4 GHz would lower costs, and a T2080 at 1.8 GHz would raise the cost fairly significantly, BUT if SMP was ever implemented the T2080 would provide well over twice the performance.

Further, I'm not as familiar with AmigaOS and MorphOS coding as I am with other micro kernel OS'.
So I'm curious, how well does this environment support well threaded code?

Under OS9, I would have strived to create the smallest modules necessary that had some flexibility to be used by multiple processes.
But as MorphOS is not focused on re-entrant code, I'm not sure what threading techniques would work best for it.

BTW - Kronos, have I ever mentioned to you how creepy using the name of a Greek Titan is?

@jim

IF that turns out to be the gameplan for the ISA-shift I see no point in supporting SMP or anything similar on the PPC side. Rather just go with the fastest single-core performance on PPC and let the x86 side handle the heavy lifting.



ps:

/me cuddles Jim "don't you fear the big moster under your bed, it won't be hungry for another hour or so"

Quote:


Hey, if I am not mistaken, Kronos castrated his own father and ate his own children.
That makes Freddy Krueger seem like a nice guy. ;)


OK, now the problem.
There are T10XX CPUs with less than four cores, but the T2081 and T2080 are faster four core/eight thread CPUs.
So, is it worth reconsidering other CPUs that could offer a higher clock speed with fewer cores (like the PowerPC 970 used in our G5s)?
I had previously dismissed this idea, but except for power use and heat generation, that still is a pretty good processor.

[ Edited by Jim 06.12.2015 - 11:18 ]

> is it worth reconsidering other CPUs that could offer a higher clock speed
> with fewer cores (like the PowerPC 970 used in our G5s)?

How's availability of those?

Quote:

Not good.
Try looking them up on IBM's website.
And adapting single G5 modules is not really possible.
I doubt there is enough documentation.

And the Freescale single core I'd point to after that is the P5010, but that is way over priced.

So, the T1022 at 1.4 GHz then?
That doesn't make sense with the minimal cost increase for the T1042.

> Try looking them up on IBM's website.

I can't remember they were ever available to order directly from the IBM website.

Quote:

No, but the technical data was there.
With IBM's sale of their manufacturing capacity to Global Foundries I think we are seeing a real shift in their thinking.

Since these facilities are optimized for really large scale production, while it might be technically possible to get them to contract for new PowerPC 970 processors (possibly even the mythical single core 3.0 GHz PowerPC 970GX) the profit they would want to generate (and the volume needed to justify it) would probably be outrageous.

I don't think they will ever resume sales of small quantities of processors.
Hope they don't make too many mistakes in the future (as they already disappointed Apple).
If they are going to rely on really large scale sales they are going to need to stay on the ball.

Because licensing, service and maintenance contracts are only going to bring in so much.

BTW - Does anyone think the T10XX line might have some overclocking headroom?

Looking over this thread, it would be better if we could just incorporate it into one of the previous threads that cover this type of idea.
Like the 'Genesi & PowerPC' thread.
Not that I'd insist that is the only company that could help.

But I do respect Bill and his previous work on the community's behalf.