JGZimmerle (+1.0)
28/9/06 3:31AM |
Another ARM chip wich is not quite suitable as a desktop CPU. Only 8 PCIe lanes, that is only half of what PCIe graphics cards require. |
gdshaw (+3.0)
28/9/06 6:24AM |
In reply to JGZimmerle:
Having 8 lanes available does not exclude all PCIe graphics cards. Some require 16 lanes, but others are available which use only 1. Not ideal, I'm sure, but not a showstopper either. |
lostamarble (+3.0)
28/9/06 6:27AM |
[Link: www.intel.com] - of course, every PCIe graphics card must be a 16 lane bandwidth monster. there is absolutely no way that it could work in a physical 16* slot with 8* electrical lanes, so all those budget mobos, or SLI mobos that do this are of course a waste of your money and will not be able to drive your grapchics card - if they do, or you have seen it, it's a bad dream. if some slack company decided to put 8 16* physical slots on a mobo and only wire a single lane to it, your graphics card would never work. physical characteristics of slots/cards do not fracture the flexability of PCIe (though perhaps my spelling does). it's a bandwidth 'requirement' of graphics cards that ideally they should be used in a 16 lane slot. |
bucksboy (+4.0)
28/9/06 8:45AM |
The increase in memory speed over the Iyonix chip (approx. 2.5x) would further multiply the benefit of the higher clock speed, presumably. |
JGZimmerle (+1.0)
28/9/06 12:46PM |
I know you could get PCIe graphics cards to work with 8 lanes, but did you ever actually use such a crippled device? They are slow. Really slow. And in a computer, you would need at least four lanes for the other expansion slots, wich leaves us with only four lanes for the graphics card.
For me the biggest issue with the current Iyonix has always been the non-availability of an AGP port, because PCI graphics cards always lagged years behind. Now the world has moved back to a unified expansion interface for all cards, and we get a chip that again would severely limit the graphics capabilities of a RISC OS computer based on it. |
adrianl (+2.0)
28/9/06 1:15PM |
Well, that's what happens when you misuse chips that have been designed for something else. |
ninja
28/9/06 2:13PM |
Your sub-heading there seems to have broken the rss feed, by the way. Your ampersand needs escaping. |
hubersn (+5.7)
28/9/06 2:39PM |
I don't think that the PCI graphics inside the current IYONIX is a problem. I don't think that a slower-than-the-fastest-PC PCIex graphics inside a hypothetical IYONIX II would be a problem.
But I am sure that a doubling in clock speed, much faster RAM access and (for the first time in ARM history, at least as I remember it) a comparatively large L2 cache would give us a large performance boost comparable to the ARM2 -> ARM3, ARM710 -> SA and the SA -> Iyonix upgrade. |
hubersn (+5.5)
28/9/06 2:41PM |
I don't think that the PCI graphics inside the current IYONIX is a problem. I don't think that a slower-than-the-fastest-PC PCIex graphics inside a hypothetical IYONIX II would be a problem.
But I am sure that a doubling in clock speed, much faster RAM access and (for the first time in ARM history, at least as I remember it) a comparatively large L2 cache would give us a large performance boost comparable to the ARM2 -> ARM3, ARM710 -> StrongARM and the StrongARM -> Iyonix upgrade. |
7thsoftware (+4.0)
28/9/06 3:30PM |
It's worth noting that the RAM will be significantly more than 2.5x faster than in the existing, due to the broken memory interface in the existing XScale not getting even close to the memory speed that you'd expect from the specs. This, of course, assumes that Intel have got the memory interface working correctly in this XScale...  |
arawnsley (+6.2)
28/9/06 4:08PM |
Regarding PCI express, you'll find that many older SLI motherboards (allowing two graphics cards) include two PEG slots (ie. two full size PCI Express grpahics card slots) but didn't give maximum bandwidth. This was because last generation northbridge/southbridge combinations only offered (say) 24 lanes, as they thought "16way for a grpahics card, and 8 lanes for everything else".
As such, many systems are already running graphics cards in slots which are 8x not 16x, and it didn't limit choice of card nor performance (to any great degree). Indeed, it was only latterly that nVIDIA issued a version of their Nforce 4 chipset with 2 full 16x slots.
Only last week I was reading about the lane arrangements on the most recent Intel 975 boards and ATI R600 systems, which in some cases still didn't offer 16x to all PEG slots (although to be fair, they are now talking about having 3 PEG slots on a single motherboard, to allow for a third graphics card for physics processing).
Translation - number of PCI express lanes isn't really that big a deal right now, and won't (shouldn't) affect choice of available cards. |
hzn (+11.7)
28/9/06 4:24PM |
So Intel has a new, significantly faster IOP which would give CTL the chance to build a new, faster IYONIX pc or the like. But instead saying wow, that is an opportunity - let us hope CTL or some other manufacturer builds a new, fast RISC OS system with that we're back to usual: Complain about what ever deficiency is there to be found!
But even with that alleged deficienty the new IOP is better on all accounts as far as I can tell (well except for the fact that no IYONIX using it is there yet, but I don't complain about CTL not having finished it off yet) so why start off with complaining. This way of looking at things is beyond me!
|
lym (+2.0)
28/9/06 5:01PM |
In reply to Hzn:
you're right. Who knows what will come of this - Castle may or many not decide to build a new Iyonix - but if they did, then this new processor would surely improve things enormously. Let's not get carried away, but it does seem strange to knock the prospects of an IOP341/348-powered RISC OS solution before anyone's had a go at making one! |
SimonC (+4.0)
28/9/06 5:34PM |
Nah, far better to have a good moan in the tradition of RISC OS.
If there is some impact on potential graphics performance, how serious would that be for an Iyonix II? I can't see it doing much to stretch a graphics card anyway, so would a far from theoretically maximum performance there be all that noticeable? |
AMS (+5.5)
28/9/06 7:45PM |
Thing is intel are aiming these things clearly in places where *I/O performance matters* (keeping x8 SATA drives fed and watered is going to require some umph). That being the case these chips are not going to be slouches
All in all good news, hopefully Castle (given their experience with the related IOP321) will be in a position to produce a machine based on one of these CPU's. The performance should be somewhat better than the simple arithmetic of dividing the new CPU clock rate by the old (the faster memory interface and the large L2 cache should see to that). |
martin
28/9/06 8:58PM |
The Iyonix : Four years old - fancy that. Amazing how time passes.
Hmm, yes, I suppose it is about time we all bought a new RISC OS machine - one every four years seems to be what those windoz folks go for. |
JWCR
28/9/06 11:27PM |
As someone on a limited budget, the question that occurs to me is:
If Castle were to produce a new machine using one of these chips, could I just buy one of the new motherboards and replace the one in my current IyonixPC, or would I need to buy the whole new machine? |
tamias (+1.0)
29/9/06 9:37AM |
In reply to JWCR:
Practically it should be perfectly possible (what is the rest of the Iyonix apart from drives, CPU, and the case?) but remember how long it was before Castle sold DIY Iyonix kits; and only for a limited time too. I would be surprised if they offer a straight motherboard upgrade in the first couple of years of any potential new system. |
hEgelia (+3.0)
29/9/06 10:24AM |
I just hope Castle still thinks it's worth their while to build an Iyonix II - technically it's certainly possible and I assume enough progress has been made now to make it a substantial improvement over the original Iyonix. The question is, do they still believe it's feasible to invest the time and effort in our current market? How many people will buy it within, say, the first year of production?
I'm sure that if they see an economical possibility to do so, they will. Perhaps it also depends on how RISC OS 5 works out - if parts of it actually are to be open sourced and development thereof would be considerable, it could take a load of Castle so they could concentrate more on Iyonix II and its hardware support...
By the way, it would be nice if they'd lowered pricing of the first Iyonix to tempt those still using legacy hardware. |
dgs (+4.5)
29/9/06 12:09PM |
In reply to hEgelia:
"it would be nice if they'd lowered pricing of the first Iyonix to tempt those still using legacy hardware"
They already have, several times. The Iyonix is now priced at not much more than half its original list price.
I think there are quite a few people who don't bother upgrading because they're just not really all that serious about using RISC OS.
dgs |
chrisj
29/9/06 1:03PM |
In reply to dgs:
"I think there are quite a few people who don't bother upgrading because they're just not really all that serious about using RISC OS. "
Or because they can't justify spending GBP600+ for a new machine (especially if they're one of the many people who have to have a MS or Apple machine in addition to their RISC OS hardware). Or because they can't justify spending GBP600+ on a new RISC OS machine in preference to spending the money on software upgrades. |
dgs (+2.5)
29/9/06 1:22PM |
In reply to chrisj:
"can't justify spending GBP600+ for a new machine"
Well, exactly the same thing was said when the Iyonix was GBP1500+. That doesn't mean that Castle (or indeed the A9home vendors) should go on reducing the price right down to ten pence, just because there will always be someone who "can't justify" buying a new RISC OS machine no matter how cheap they're made.
If Castle did make a 1.2GHz RISC OS 5 machine, I imagine it would be well over 1000GBP... so we'd have yet another round of moaning about prices, despite the massive advantages that such a machine could bring.
dgs |
flibble (+0.1)
29/9/06 3:11PM |
In reply to dgs:
I've not upgraded, because I'm no longer that serious about using RISC OS. I've found other OS's and other apps to enjoy, so that paying for an Iyonix I wouldn't use would just be an act of charity towards Castle. About the only thing left that could grab my attention back to RISC OS would be if they Open Sourced it and I were able to experiment freely with the code.
Re the price, after the early adopters have paid full price, you either need to drop the price to encourage those with lesser budget, or market it to new people willing to pay the full price. |
bucksboy (+2.6)
29/9/06 6:48PM |
An Iyonix II with 3-5x the performance of the current machine at GBP 1000 or thereabouts would certainly be interesting to me. |
fwibbler (+0.1)
29/9/06 7:54PM |
Even I might be interested in that.
Apart from the processor, what other specs would people want from such a machine? |
druck (+2.5)
30/9/06 2:24PM |
So its straight in to knocking the new chip with the first comment. Would anyone like to explain how a RISC OS machine with a modest processor performance and no 3D software (i.e. games) avilable, is going put any appreciable load on the most budget PCIe 1x card never mind utilisng 4GB/s of 16x.
The point of PCIe is to allow access to modern cheap graphics cards, with some degree of future profing, now that AGP has become legacy, and the PCI cards that the Iyonix supports are becomeing rarer (although still available). |
jbit (+4.0)
2/10/06 2:25AM |
""in a computer, you would need at least four lanes for the other expansion slots""
The 348 and 341/342 can use PCIe and PCI-X concurrently, so you could use eight PCIe lanes for a graphics card, and a plain PCI-X bus for sound/USB/whatever else.
And I don't get the argument that eight lanes would "severely limit the graphics capabilities"... A) it'd be replacing a PCI bus on the iyonix, B) eight PCIe lanes is faster than AGP-8x, C) RiscOS isn't known for its large selection of 3D games which require every ounce of graphics power possible. I think a much bigger limiting factor would be the drivers....
All in all, I think the 342 looks like it could be a very nice desktop processor, even if it's not intended to be so. I wonder how much the IQ81342EX (ATX board with a 342 on it) will cost..... hmmmm.... |
| 1 comment(s) are below your moderation threshold. Login to view them. |
| |
Chip giant Intel has officially launched the latest additions to the XScale IOP family - including the 81348 which boasts eight high-speed hard disc access ports. At the Intel Developer Forum in San Franciso, the 1.2Ghz ARM-compatible CPUs were today touted as 
