I know I've posted about the STI Cell/BE processor before, but I halted those posts due to the lack of available support for those processors.
However, the Cell's SPE processing units are available separately on the Toshiba Spurs Engine. Each Spurs Engine processor has four SPE's. Expansion boards featuring this processor are currently available and use a PCI-e X1 interface.
Since a standard PPC processsor (like those used in Morphos and AmigaOS systems) is very similar to the core processor of the Cell, and we can add SPE functionality, what we could build on our own could potentially be more powerful than a Cell driven system.
How could that be when the Cell runs at 3.2 GHz and has 8 SPE's you might ask. Well, there are dual core E600 processors running at 1.5 GHz or higher. Currently, the 8640 and the 8641 have 4 PCI-e lanes that are unused on the Freescale Evaluation board. So, it is potentially possible to run four of these Spurs engine based boards simultaneously. We could have 2 cores running at up to 1.5 GHz (virtually the same as a 3.0 processor) and up to 16 SPE's. If code was threaded properly this would likely be faster than a Cell based system.
What do you think?

Quote:

I think that you might underestimate how difficult it is to write software that utilizes 16 cores in an efficient manner This is even more of an issue if you are talking about highly specialized cores like Cell SPEs.

My main workstation is a quad-core system and there are few applications which manage to come close to a 100% CPU usage on all cores.

A very good point, and I don't see a problem starting out with only four SPEs in one processor on a single PCI-e board. It would still provide powerful floating point co-processors.
And since the SPEs are little more than simplified coprocessors with local memory, programming them should be easier than programming a PPC core. IBM's own Linux Cell SDK might even be useful for some of the initial work.

[ Edited by Jim on 2009/3/16 21:12 ]

Jim, can I ask you very stupid question? :)

Excuse me, but can you produce the mentioned 8610/40D/41D (and so on and so forth) board? May be, you know anybody who can do it? AFAIR, BBRV told that 86xx program is closed by FreeScale... (AFAIR! May be, I'm wrong)

Second. What OS should run on this board? MorphOS? Did you hear news: "QBox is a side project which is no longer important"?

[Bad joke ON]
I think it's better to wait Symbian for ARM-based netbooks. MorphOS has no future due to various reasons.
[Bad joke OFF]

@Jim
Quote:
I strongly disagree with this logic.

> BBRV told that 86xx program is closed by FreeScale

What do you mean with "program"? BBRV told that MPC86xx (i.e. e600) will not be further developed. Already existing products are still manufactured and sold, of course.

> there are dual core E600 processors running at 1.5 GHz or higher.

No, not higher than 1.5 GHz.

> We could have 2 cores running at up to 1.5 GHz

...with MorphOS' ABox only able to use one, as you have already been told.

Quote:

The statement I wanted to write! Sorry for my [beeep] english :)
Anyway, there is no difference for morphers, because of we can not expect new PPC-based hardware.

Quote:

How much money does bPlan ask for, to develop a new computer?

> there is no difference for morphers, because of we can not expect
> new PPC-based hardware.

If so, then it's not because the "86xx program is closed by FreeScale". Nobody is hindered from buying MPC86xx chips from Freescale and building boards with them to which MorphOS could then be ported. Lack of future e600 variants doesn't render current e600 variants non-existent.

Quote:

Absolutely correct Andreas. Further, Freescale is not the only game in town. AMCC, IBM, and several other companies still make PPC processors (some of which are more powerful than the e600 series).
However, I am disturbed by the general malaise that appears to be common in Morphos users. Wait for ARM based systems? Why? They're not even vaguely similar to we've currently been using. New PPC processors aren't being developed? Bull, all three of the top gaming consoles currently on the market use custom variants of PPC architecture. AMCC will hopefully eventually introduce its Titan processor and PPC processors still seem quite popular in other areas as well (embedded systems, communications/network control, etc). What we haven't seen is a lot of support for personal computer applications.
To take this a little further, why can't Morphos support more than one core? Bill McEwen at Amiga Inc. obviously thinks Amiga OS will (in some later variant) be able to support multiple cores or he wouldn't have encouraged ACK Systems to develop a dual core motherboard. Why is Morphos (and Amiga OS for that matter) only 32 bit (when 64 bit processors have been available for years)? Finally, why can't Morphos grow and change?
Frankly, right now I might as well look to AROS for Amiga compatible OS that is continuing to move forward (albeit slowly), because the majority of the posters here don't seem that interested in advancing the current status of their OS of choice.

BTW - I forgot to mention that one of the ideas I mentioned originally won't work. We can't (easily) use IBM's Cell SDK on PPC based computers to develop code for the Spurs Engine's SPEs. The SDK has been developed for several OS', but the PPC variant (running under Linux) requires a 64bit processor.

@Jim, Andreas

Yes, I know about IBM, PA Semi, AMCC (too lazy to check it, but there is several PPC vendors)...
And ARMed MorphOS is not my choise (as well as x86).
Here we have several issues:
1. Processor
2. Main Board (for desktop, not for netbook)
3. MorphOS for this mobo

1. Well, for example, bPlan (or somebody else) choosed suitable processor for new computer
2. How much boards shold be produced? bPlan (or somebody else) want to have a profit. AOS+MOS+Linux fans... Is it enough to make stable, guaranteed market?
3. Ok, we have modern mobo with super-puper-duper PPC. But MOS-Team can say: "It is too much for ABox!". Finish. End of story...

So, there is no Pegasos successor now and in near future.

@jcmarcos

Paragraph 2 in my list is answer to your question. "Invisible hand of the market" :)

Quote:

Who cares what Bill of Amiga thinks? He obviuosly has no clue about technical things and encouraged Adam of Ack (Adam is a joke, too) to develop something to pretend a vivid company for the court case. Maybe Amiga should went the hair dryer business: They are experts in producing hot air.

It's quite the opposite, the SPEs are extremely tricky to program (efficiently) for of several reasons.
For one the only way the can address memory outside their own local ram (analog to cache) is via DMA, DMA transfers has quite some latency so without using doublebuffering (which adds complexity to the software) it's pretty much impossible to use the capacity of the SPU.

SPEs also only work with vector operations internally, having scalar code execute on them is not efficient.

But the biggest problem may be synchronization.

Now I'm not familar with SPURS so I have no idea exactly how it behaves compared to a PS3, but I assume it's somewhat similar.
On the current Pegasos architecture I think it would do extremely poor due to the slow systembus.

> why can't Morphos support more than one core?

Backwards compatibility to OS3 (and thus current MorphOS) API is the argument brought up mostly.

> Bill McEwen at Amiga Inc. obviously thinks Amiga OS will (in some
> later variant) be able to support multiple cores or he wouldn't
> have encouraged ACK Systems to develop a dual core motherboard.

I don't think so. The high-end system was announced for end of 2007 and Hyperion's OS4 as OS. No way that OS4 could have been multi-core enabled by then, even (rather: all the more) if Amiga Inc. would have managed to grab it.
On the other hand, there is this interesting posting by HJF.

> why can't Morphos grow and change?

See above: backwards compatibility.

> right now I might as well look to AROS for Amiga compatible OS that
> is continuing to move forward (albeit slowly), because the majority
> of the posters here don't seem that interested in advancing the
> current status of their OS of choice.

AROS is also aimed at being OS3 API compatible and therefore probably can't overtake MorphOS (regarding multi-core support at least, 64 bit version is there though). That's why Anubis OS was born.

> On the current Pegasos architecture I think it would do extremely
> poor due to the slow systembus.

Jim's idea isn't about current Pegasos, which lacks PCIe anyway.

> So, there is no Pegasos successor now and in near future.

I just wanted to state the fact that *everybody* can *now* buy e600 chips from Freescale (or distributors) and design and manufacture boards with them (to which MorphOS may or may not be ported). Your original posting didn't seem to support that fact. Of course, you are right in that there are also other considerations than just the technical possibility, such as commercial viability.
But, what's the definition of "Pegasos successor" at all? Would the *available* Fixstars PowerStation (*) with MorphOS count as "Pegasos successor"? At least, I see no sense in restricting future MorphOS hardware to bplan/Genesi made gear (and MorphOS Team is obviously with me on this). One might even consider the MacMini G4 running MorphOS a "Pegasos successor".

Edit:
(*) I just got aware that the PowerStation very recently went "out of stock". I very much hope that doesn't equal "out of production".

[ Edited by Andreas_Wolf on 2009/3/20 4:25 ]

Quote:

Yes, you are right. But we have what we have. May be, because of...

Quote:

Quote:

With MorphOS, yes, it will. It slightly expensive, especially with shipping to Russia, but I like it :) Really, every time we talking about "Pegasos successor" we mean bPlan's products because of we accustomed to it. But I think it can be every relatively modern desktop.

Quote:

I don't think the Mini is step forvard. But MacMini better than nothing. Successor? No, it's like you change C64 to C64C with more powerfull processor (1 MHz -> 1,5 MHz).

Neat! A very thoughtful group of responses.

One thought, when I run Aros as a process (hosted) on a computer under another OS that multitasks I have (at least in a crude way) the ability to multitask (at least outside of Aros). Given that I can run multiple copies of the hosted OS, all I should need is some code to communicate between processes in different emulations - crude multitasking.

Couldn't we improve on that without sacrificing compatibility?

Second, as to new designs, while there is less support for PPC based personal computer systems than there has been in the past the hardware is available and there could be more advanced personal computer designs produced with this hardware.

Third, I don't think the Morphos crew is as likely to reject new hardware out of hand as many of you apparently think. A good, low-cost, widely available design that would work under Linux (first), then Morphos (later), and maybe AmigaOS4 (last) would, in my opinion, stand a good chance of receiving support.