I don't get it.

[AJCopland]

1. People assuming there is no cheap and x86-64 price/performance killing PPC alternative...there is it's called Xenon and runs at 3.2Ghz with the most efficient Power architecture available today for peanuts (Sony/Toshiba funded cost of CELL which IBM derived this CPU from)

So I'll just pop out and buy a Xenon CPU that I can actually run my own code on then... oh wait, no-one sells them except in the Xbox360 which is closed source. In fact even if you wanted too you can't buy them because the Xenon design is licensed only to MS. Sure you could have IBM make you a new CPU based on the PPC core used in the Xenon/Cell but then you've got a bespoke CPU designed for you and that isn't cheap.

Also you'll need a modified compiler for it since generic PPC code runs like crap on it given that it's an in-order architecture unlike every other PPC CPU released since the G3. It's also, just like the CELL, designed to perform best when you're utilising both of the hardware threads per-core. Which is something that AOS can't do either.

You could buy a CELL CPU from IBM but they're not cheap, not by a long way and they're being phased out.

Andy

[AJCopland]

What I'm trying to get at, is that they've picked the best CPU they can get involving their partner company (to reduce costs) who bring in all the experience of designing these boards.

Wishing they'd use XYZ cpu is daft because they'd then be looking at another design house to form a partnership with to spread the costs and get development experience.

This is most likely a compromise and even if it's not to your liking at 1.8GHz even a single core is the fastest OS4 capable machine that we've ever had access too.

It's not what I'm after either, I jumped ship to PCs, Windows, Linux and MacOS years ago, the Amiga is the Classic Amiga for me even if they it is my super-fast 68060 at 50MHz. That doesn't mean I can't appreciate that every design is bound to involved some compromise and that cheap PPC CPUs just aren't available. Even a 400Mhz PPC will run to $50 from somewhere like DigiKey and everytime someone releases an affordable PPC board everyone bitches about how slow it is!!!

Andy

[AJCopland]

There's a lot of arguments against that, and I don't just mean peoples personal feelings.
AOS _still_ wouldn't be able to access anything more than the first physical core and SMT would be equally wasted. Compatibility would be thrown out for old PPC programs, emulation would be required which is not easy. Plus AOS itself would have to be ported including any and all endianess issues.

It also wouldn't bring any greater hardware compatibility because no-one has released any drivers of Amiga OS and the ones we do have would also need porting.

Sure it'd get us access to cheaper hardware but if you're looking for x86 AOS then you're looking for AROS.

Andy

[AJCopland]

The Xenon isn't the same as the CELL PPU, I'm a 360 and PS3 developer, they're slightly different. The Xenon is based on it but it was customised for MS. Furthermore it's only available in a chip package which includes a direct bus to the Xenos GPU or as the new integrated package with GPU on package.

So the argument still applies: you contact IBM, they say they only do Xenon as a package for MS, and you'll need to fork out for a re-packaged Xenon chip. I.e. you need a custom chip package containing the Xenon CPU. You pay through the nose for it and you've still not got a off-the-shelf, and thus cheaper, component. You've got all of the work you'd need to do for a custom chip including evaluation and board design etc.

That's my point.

No, I'm not going to name a sub £100 PPC cpu that matches it because there isn't one.

That's my other point.

It's why there are good arguments for going to x86 (which also has negatives) but you can't just pick a closed PPC chip design and go "that one would be better" because no-one can use it for the reasons listed above, this is why I think they made a reasonable choice for the current CPU.

If AOS4 can be made to use that other core eventually then you've instantly doubled the available resources without any extra work. Also that 1.8GHz core is an Out-Of-Order core which is trivial to extract good performance from, whereas the Xenon/CellPPU is an in-order core which it is difficult to extract performance from.

Thus the 1.8GHz available may well be better utilised than a single core/thread on a 3.2GHz Xenon.

Editing to include benchmark.
I knew I'd read this somewhere:
http://www.primatelabs.ca/blog/2007/05/playstation-3-performance-may-2007/
And I quote:

It’s clear that the Cell processor isn’t all that impressive as a general-purpose CPU; if it’s not executing code designed to run on the Cell processor, it’s generally slower than a PowerPC G5 @ 1.6GHz (the baseline processor for Geekbench).

Now, we've got a 1.8GHz core in this X1000 which is roughly equivalent to the G5 but using less power...

[AJCopland]

And costing 450euro and with no path for upgrade thereafter. This chip is quite simply the literal definition of a dead end. The company that designed them has been swallowed up and the only reason the PA6T is still being produced is because the US army had a contract for them. Simply put, these things are army surplus, which should suit doomy right down to the ground - they really are "mil spec".

The only other PPC on the horizon that might even be remotely competitive that I've heard of is a chip still on the drawing boards at Freescale which may or may not ever come to anything and even if it does will still be years away.

Ah yes I forgot, around here its not that you want the moon-on-a-stick, its that it's not the moon-on-a-stick you were hoping for.

Andy

[AJCopland]

I'm pretty certain most people "around here" are not so bereft of reason that they see PPC as anything other than an architectural dead end as far as the desktop goes. This to us is a hobby and nothing more. Most of us simply want to enjoy our hobby, we neither wanted nor needed the "moon-on-a-stick" and haven't done since the exodus. Even within those on this site who enjoy OS4 have all pretty much in unison said this thing is way too expensive for what it can offer.

Within that context a £1500 piece of kit simply doesn't make much sense. Using a £450 dead end cpu that can be trounced by a £50 AMD or Intel part just makes it a sick joke.

PPC is incredibly expensive, even when compared to other embedded CPUs. Even Coldfire can be bought on dev boards that are significantly cheaper than an equivalently spec'd PPC. MIPS and ARM leave them both standing in terms of price/performance.

Even to the most hard-line BAF the X1000 is a harsh pill to swallow.

But that's just it, it's not for you, they've gone and made it seemingly despite the lack of a market and that's the cost. Some people will buy it though and it'll probably sell some for Linux and MorphOS too since there are die-hard PPCLinux fans too. Just because it's not for you and you wanted X, Y, or Z doesn't mean that it isn't a reasonable thing for THEM to have made.

Yes you wanted power at less cost, but AOS4.x doesn't run on anything other than PPC so it was always going to be PPC despite the cost and lack of desktop CPUs because that's the little box we're locked in architecturally.

If you want x86 or ARM then you've got to go for AROS for the time being.

So what "I don't get" is what all the complaining is about, it seems like everyone has a negative opinion to share about the X1000 but isn't willing to accept that their position isn't going to match the existing software (AOS4.x) or the companies (Hyperion etc) goals.

So why complain about it saying it's too expensive when we know that it's always going to be too expensive when using PPC and that it is _always_ going to be PPC.

You either have to accept that and take PPC hardware for AOS4.x or you switch to your alternatives in the form of MorphOS on existing eMac hardware or AROS on x86/ARM(?) but what's the use in complaining about the X1000 when at least someone has spent time, money and their own hard work bringing it to fruition? It's like slapping them in the face and say "you're shit, you should have done XYZ" despite knowing that XYZ is currently out of their control.

Andy

[AJCopland]

Even to the most hard-line BAF the X1000 is a harsh pill to swallow.

Also... what's a BAF?

[AJCopland]

No one has said or implied that. Also quit making it personal. It isn't.

My apologies, I just get really worked up when people seem harsh towards peoples achievements and bringing a PPC computer to market running AOS4.x is an achievement these days. It may not be a wise one, but it's their choice to do it or not and peoples decision if they want to buy it or not.

Look the way I see it is that £1500 for what you get is overly steep for me, I can't afford it, but your criticism's are harsh. You basically stated this CPu is a dead end, but they don't have to use this CPU again, any compatible PPC (from AMCC, Freescale, IBM, etc) chip will do in future if they're still around. Or it's not a good power/performance/price relationship, neither is a Mac but it's the only way you're running OSX.

Simply slamming them for doing it at all though seems churlish.
I'm gonna bow out 'cos you're right though I am making it personal when you don't mean it to be, so my apologies once more.

Andy