You didn't mix them up.

TetraPower *is* the OSW. We dropped the OSW name for various reasons. It was a silly name anyway.

The plan never got off the ground, however as a trade-in deal for complete systems it was sort of viable; what you suggested here on this topic was to 'trade in a Pegasos'.

_________________
Matt Sealey

Here is the Business Plan if you are interested. We did sincerely look to developing this market:

A Disruptive Business Plan – The Open Server Workstation

R&B :)

_________________
http://bbrv.blogspot.com

Erm, no, I did not suggest something. I just quoted (kind of) what was written in the thread I quoted, please see comment No. 3 and No. 4 in that OSW-thread:

-----

takemehomegrandma:

And what about non-ODW (but still Pegasos 2) owners?

-

bbrv:

Yes, PegasosPPC Users are included!

-----
Thank you for the business plan.

So you _did_ look to developing that market in April 2006. Do you still do?

I did not directly answer the vote's question because my answer doesn't match any of the selections.
There are two aspects:
Aspect 1. Performance and usability.
There are two alternate points of view at this:
1a) This could be more powerful machine than Pegasos. With more powerful CPU. Even at the time of Pegasos' birth Intel machines greatly outperformed it (if we consider the same OS - Linux). I had a chance to compare myself - Pentium-III machine ran much faster. At the same time Pegasos contains several not needed components. These are second Ethernet card and IEEE1394 connector. A typical desktop doesn't need more than one network interface; if you need more, you can always fit in a PCI card. IEEE1394 is simply obsolete - if you go to consumer electronics store, you won't find any IEEE1394 equipment - USB 2.0 won the battle (even without any serious struggle). Only some video cameras still retain it in addition to USB2, however it's just a legacy (the same as video camera itself - today's digital photo cameras also can record videos and are much more compact and comfortable to handle).
1b) We live in 2009 and it's already a laptop age. Modern laptops can do the same things as desktops can. Modern desktops come with integrated components - and in this sense they are no different to laptops. The only thing that desktop allows is constant upgrade of a video card (it you're a rich hardcore gamer - this doesn't apply to people gathering here). So a new machine could be a laptop. It could be based even on existing EFIKA design - just add some video chip on the board and you're done at minimum. Personally i would replace my Pegasos with such a laptop, even sacrificing some performance.
Of course the new machine should have several USB 2.0 connectors and Serial ATA.
When i'm telling this i assume this machine would run MorphOS.
Aspect 2. Cost of ownership and feasibility.
This picture is much worse. First, our machines are expensive. x86-based machines with equal characteristics cost less because they are manufactured in greater quantities. And nothing can be done with it. Possibilities of operating systems they run can be not even compared to those offered by Windows OS. This means that discussed machine is a second machine at owner's home. This automatically means that it's nothing else than a "cool toy", which exists just for fun. And here comes the main question - how many people of the communuty could spend ~600EUR "just for fun"? From others' point of view (parents, family, etc - who has what) this would be just throwing these money away into the window. This creates a contradicting requirement - the machine must be *VERY* cheap, compared to today's EFIKA (does it still cost 100EUR?).
As to personally me, i would not be able to buy such a machine even for 10EUR - my family is in permanent deep financial trouble, so noone would let me to spend even 1EUR "just for fun".

a petition doesn't build a market.

It could if there are (10033, 10087) numbers instead of (33, 87) in poll results.

a poll doesn't bring people.

that's about the size of the interested (or uninterested) people you can get here.
you have to learn to live with that : )

We shut that poll off within a few hours of it being posted. It is not that we don't want to develop for PowerPC, we do! The problem is that we have a very hard time controlling our corporate destiny when we don't own any of the processor IP; have very, very little influence on the roadmap (if any); and have had a relatively small market to serve, which generally means IBM and Freescale are not interested. That said, Freescale has been much, much more accommodating over the years. We have dozens of stories we could share...

One of the reasons we started to work with ARM is because 1) IBM was not interested in the 'desktop computing' market in *any* way relative to the user (or the developer focused on that market), and 2) Senior Freescale Executives (C-level) encouraged us to move in that direction. So, here we are.

If we have a chance one day again to push ahead with PowerPC, we will. You can be sure we will be right out front with all our altivec engines firing at full throttle.

R&B :-)

_________________
http://bbrv.blogspot.com

i still have hope for a newer powerpc like efika. but also welcome an arm chip too.

if there is little interest in powerpc, why do the game consoles use it ? and is there any hint of what the next generation game consoles might use ?

matt

Game consoles are completely closed and proprietary systems and have very long development cycles. The devices in the market today have all been there a couple of years at least and were in development a couple of years before that. What will happen next, we cannot say.

R&B :)

_________________
http://bbrv.blogspot.com

I'm sure the next PlayStation will run x86..

Given Sony's reliance on compatibility, I doubt it. Most if not all PS3 games would have to run under the PS4 and this means at least giving it a Cell as well as the x86 - as with PS2 and PSOne compatibility - or of course providing an x86 processor capable of emulating the Cell (at least through whatever APIs they have available) or translating SPU code to x86 (probably Larrabee) and scheduling it properly.

I somehow doubt they will manage that AND manage the difference between the graphics cores (nVidia's GSX to the Intel core). Maybe they'll do it by shipping at $699 again.

Nintendo may move to ARM. They have an ARM core in the Wii already, so moving the console to a full SoC model (ARM CPU + all peripherals in a single unit) would be a benefit, but no ARM could pick up the task of emulating a >700MHz G3 to play Wii games. Again they may well just integrate the entire console into a single chip and then add the G3 for the time being in the same relationship to the way it is done in the Wii now (Broadway is the G3, Hollywood is an ARM SoC and ATI graphics core) just with a MUCH faster ARM CPU inside (Cortex-A9 maybe?).

The alternative is fine a better PowerPC processor to get the performance they'd need to run games in HD (supposedly the next step for the Wii). I can't even think of one and I look at this stuff every day. Maybe they will get a custom core built by IBM based on the G5 or Xenon or Cell (which would then give them instant compatibility with the G3 they have right now) but if IBM are intent on selling their chip division.. good luck!

The key will be using a compatible graphics core, and they did really well with the Gamecube/Wii and ArtX/ATI, so no problem here. The new Wii will probably ship at $249 again. Good ol' Nintendo :)

Microsoft have the same problem as Sony - somehow emulating a 3.2GHz, triple core, triple threaded processor on a new x86 chip. While they don't have to handle the SPUs, I think emulating SPUs by doing code translation would be easier than mapping raw CPU performance to a new x86 chip. I simply doubt they will be able to do it. But the graphics core will probably be ATI again, so they have a good chance of keeping some reasonable level of compatibility as long as they have a reasonable performing replacement they can map 6 concurrent 3.2Ghz-performance-expecting threads to under emulation. Maybe. They have 2 years of development until we even get a hint of what the console might be, plus the reliance on DirectX and managed code (.NET CLI) as development platform may mean they can get rid of a lot of the emulation requirement.

I am pretty certain that nobody is going to stick with PowerPC for any other reason than legacy compatibility with the installed userspace. Nintendo may be the only ones to integrate a real PowerPC in there but it'll be for GC/Wii games and not for new ones.

_________________
Matt Sealey

I'm pretty sure the Larrabee can emulate the Cell quite easily. Don't forget that Larrabee will have 32-cores and each core will have a 256-bit wide SIMD unit itself. Technically, it's feasible. Whether the cost justifies the effort is another issue.
What about ZMS from Ziilabs (former 3dlabs)?
That much is certain, PowerPC is dead -or at least dead outside embedded/car industries.

http://episteme.arstechnica.com/eve/for ... 5008207931
Never or the Playstation brand will die ...

Since SPU code has to be uploaded and scheduled on the SPUs themselves, translating it before it hits x86 core would be easy enough.

What I wonder about is if you can have 32 cores at sub-GHz or near-GHz which can emulate the PPU effectively at the same time. A lot of game code would assume it's getting a certain level of performance and only thread so much. Splitting a single PPU core (which has 3 logical threads) into real x86 cores that won't run up to the speed of even a single logical thread on the PPU might be a real challenge.
Like this?

http://www.ziilabs.com/products/processors/zms05.asp

I think it'd be way too slow, plus, it's too specialized in my opinion, it doesn't fit with Nintendo's "ease of use" ethic on games consoles. I doubt they would inflict that kind of architecture on people trying to write games.. it leaves too much to the talent of the programmer coding those 24 SIMD units, which is exactly why Cell fell over.

You'd probably more likely see something relatively standard like a single or dual core processor like the i.MX515 (i.e. fancy 3D graphics, video offloads) or whatever someone might come up with on the Cortex-A9 sometime in the next year or two (although there is a slim chance the MPC8640D or some derivative could be in the next Wii, it would make some sense :)

Anyway ARM seems to be the way to go for Nintendo, if only because they also use it in the DS, too, and it means you can leverage a common codebase across the platforms (there's no reason they can't run DS games on the Wii right now.. :)

_________________
Matt Sealey