druck (+3.2)
5/5/06 4:49PM |
Intels embedded devision are as obviously clueless as their x86 counterparts not they are on their back foot. The majority of XScale shipments are in for the PXA variant (PDAs) and IOPs (which replaced their i986 I/O controlllers in servers and workstations and network fabric), they've made absolutely no ground in pure mobile/smartphone sector which is dominated by TIs ARM based OMAP chipsets. They are well behind TI and others on moving away from a seperate application and stack processors to a single processor package.
But anyway if the XScale goes, it will be no great loss. Their divergance from DEC's StrongARM Mk2 development path has lead at a glacialy pace to increasingly inefficent and bloated processors (no surprise there being Intel), whcih are now easily bettered on both performance per MHz and per Watt by native ARM designs. Good luck to their competitors on continuing to run rings round the tired old giant. |
AW
5/5/06 7:36PM |
Nice to see a positive take on the issue, Druck.
Hopefully there'll be a future for RISC OS on efficient ARM chips yet. |
markee174 (+0.2)
5/5/06 8:51PM |
So is there a faster non-XScale chip for future Iyonix models????
|
druck (+2.5)
6/5/06 2:29AM |
Any new Iyonix motherboard would have to be substantially redesigned as newer IOP series XScales aren't pin compatible, also other compontents on the board such as the southbridge need to be replaced with new versions. So from a hardware point of view, using an alternative ARM9/10/11 based SOC to the XScale would not be infeasible. It would however require software changes to the HAL and device drivers to cope with the the different chip, but this is doable. |
gvrace
6/5/06 8:31AM |
In reply to druck:
In fairness I am not that clued up on hardware but you clarify one thing.
I thought the HAL in RiscOS 5 was supposed to be hardware independent so why does the HAL need to be modified and not just some drivers that it uses.
Greg |
jbyrne (+2.6)
6/5/06 9:19AM |
I'm afraid you've got it the wrong way round - the HAL is the hardware dependent bit. It means that the parts above it can be more hardware independent. When you move to a new chip, you write a new HAL to deal with the specific hardware on that chip, but the core OS stays the same. Also, the HAL only covers a fairly low-level set of functions, there are other higher level modules (e.g. network drivers, video drivers) that are also dependent on specific hardware.
James |
AMS
6/5/06 2:08PM |
As is the tradition in these things parts availability dries up slowly so there is no major concern here, there is more than enough time for Castle to do some redesign the CPU part of the MB to support either later Intel chips or perhaps even CPU's with ARM11 support.
Trouble is this could represent a problem for A9Home as much (or more than) for Castle - as Castle will to some extent be expected to being an alternative design to use newer ARM parts. This may mean people will hold back from purchasing anything (A9 included) until they see what developments arise from the Castle camp.
In actual fact in the medium term this may actually prompt further development - not a bad thing all in all.
|
tweety (+0.5)
6/5/06 3:49PM |
Why would it affect the A9Home? It uses an ARM9 from a different manufacturer... |
AMS
6/5/06 4:14PM |
In reply to tweety:
If Castle have to change their CPU (and if this article is right ultimately they will have to) and if people know this then it may affect sales of both Iyonix and A9 computers. This sort of thing happened before when Acorn let slip that Phoebe was on the way - sales of the RiscPC collapsed. If there were a reasonable expectation that Castle were going to upgrade the IOP321 to 333 OR were switching to an ARM from another source then that could effect sales of RISC OS machines in the interim (A9 included).
|
Jwoody
6/5/06 4:48PM |
I thought the article said that IOP333 would be the last i.e. Both IOP321 and IOP333 would go. So I don't see Castle upgrading to another Intel produced processor they will have to look else where.
This sort of problem has been on the cards since the day ARM was split from Acorn. With ARM free to develop low power processors for phones and no real wish to compete with PowerPC, Sparc, HP PA, AMD and Intel in the general CPU wars. The second nail was when Acorn sold all its ARM shares as part of the winding up and lost any influence it had left over ARM.
Personally I think its time that RISC OS looked to a different processor other than ARM. Like I have said before I would like to see VRPC enhanced so that the system part of SWI's were running as native x86 code. |
AMS
6/5/06 5:37PM |
Jwoody wrote>"I thought the article said that IOP333 would be the last i.e. Both IOP321 and IOP333 would go. So I don't see Castle upgrading to another Intel produced processor they will have to look else where."
Yes that's true. But the inferrence was that IOP321 would be going first. Intel in the past usually issued a message to the effect that a particular component not be used in "future designs", meaning that it will continue to be available for a while but will at some point go out of production. Unfortunately the article seems to point to unsubstantiated rumours (yes they could be true - but the timescale could be a year or years plus any residual stock held by distributors). Yes I also agree it would be prudent for Castle to design for a non-Intel alternative.
Jwoody wrote>"Personally I think its time that RISC OS looked to a different processor other than ARM. Like I have said before I would like to see VRPC enhanced so that the system part of SWI's were running as native x86 code."
The JIT end of VRPC helps to an extent, but to wholesale translate to x86 would be a major pain - expensive and (arguably) not the best use of resources.
I would not disagree with you on the characterisation of what Acorn did - not so much a case of nails in coffins as bullets in feet I think  |
Revin Kevin
6/5/06 9:04PM |
I thinked that I read in one of last years Wakefield's report that Castle were showing their ARM powered computers used in the embedded market, so perhaps the work to change chips to ARM ones is allready done. |
ROHC
7/5/06 12:34PM |
lets hope that somebody decides to make a decent cortex chip before the IOPs run out.
...would be nice to have a CPU that was designed for a desktop machine not one that "could also run RISC OS" /dream  |
druck (+0.2)
7/5/06 9:19PM |
Lets not forget that Intel is completely irrelevant to the ARM market, they make an almost insignicant number of processors compared to all the other licencess. The only market section they have real traction in, is in PDAs, and now almost every manufacturer has a model featuring an alternative processor, particularly for converged devices with GSM/3G capability where the Intel is week.
Anyone worried about future Castle machines, should be aware they have far more experiance with non-Intel (ARM9) designs from their set top box work. The Iyonix is their first and only XScale based system, the reason being it was the fastest clocked amd most suitable system on chip at the time. It hasn't been the fastest ARM processor for some tine, but Castle have said they wont be replacing there Iyonix until there is a suitable replacement that offers significantly more performance (say 3x), and neither Intel or other licencess have brought a suitable chip - yet.
|
| 1 comment(s) are below your moderation threshold. Login to view them. |
| |
Chip giant Intel is expected to axe families of its XScale processor in a cost cutting move, a 
