Hi Atheist

To add insult to injury, most said, "Why so many PCI slots? Are you mad?", well, look at what happened to the XE. ALL the PCI slots are used to replace flawed motherboard components/features.

I'm watching the A1XC-development very carefully. I know this is under development, but the info is quite limited at the moment. I guess we may expect to hear something more soon, but i also know that Eyetech want to make sure that The MicroA1s become a great success, before thinking of A1XCs...

Eyetech mentioned that A1XC could become the ideal development platform for
the developers that want to make use of the latest hardwares, writing AmigaOS-drivers for them, and even develop better and better software with more to support than present.

I'm looking forward to the possibility to buy A1XC, but before that, i will go for the possibility to buy an 1.4Ghz G4 CPU-module for mye A1XE-system when available sometime around spring this year

@Atheist

Quote:

Atheist wrote: ALL the PCI slots are used to replace flawed motherboard components/features.

Small exagerration there... how were we (me and Bodie) able to fit in a TV card?

_________________

@Agafaster

Quote:

its supposed (!) to be the Articia-P Northbridge, which is alleged to handle DDR SDRAM, so I imagine the FSB will probably take a hike-up too. 150 MHz ?

It's supposed to be 200MHz. The part was originally specced for 166MHz but MAI seem to have bumped it up.

Of course, this may be completely irrelevant as the Articia P is still very much missing in action. The only shred of evidence I've seen that suggests it has any substance other than press releases and brochures is this picture grabbed from an MAI presentation file that I found lurking on a chinese language site.

This design clearly isn't based on the Articia S, as that chip only supports 32-bit PCI and four 64-bit slots are visible on the board. Also, the memory slots have just one visible keying notch, which suggests DDR memory as SDR has two keying notches.

It's a pitty that picture isn't a bit bigger

And sadly, the possible existence of a prototype reference board doesn't tell us much about the state of the Articia P project. Given the chip's increasingly outdated specs and the small market for a desktop-type G3/G4 support bridge, they would be fully justified in canning it and concentrating on the Articia I.

_________________
Who do you serve, and who do you trust? - Galen

If an XC became available, had 5 PCI slots and was bugfree, I would
ditch my XE straight away and plug in my G4 module.

That is if you could buy it without a CPU module of course. I'd pay
between £80 - £100 for such a board if the features were good.

Last edited by Rob on 23-Jan-2005 at 01:00 AM.

@Rob

I'm not sure £80 would even cover the VAT!

_________________
Who do you serve, and who do you trust? - Galen

@DrBombcrater

I did say a board with no cpu module. I have a G4 module already in
my XE.

@Rob

Alan Redhouse posted a rough estimate of costs for the uA1 a while back, and I think the CPU card came in at about £175. Using the price of the XE as a guide, that means an XC board would still cost at least £514 sans processor card. The VAT portion of that price is, if my sums are correct, £75.

And I think it will cost more than that because if Eyetech follow their usual procedure of adopting MAI's reference design then the XC will be copy of the Teron PX2, the Articia P reference board -- which is the board pictured above.

The final version may be cost-reduced a bit, but it's still not going to be cheap to build something like that.

_________________
Who do you serve, and who do you trust? - Galen

@DrBombcrater
I am trying to understand your approach to things here re CX possiblity but more so the Articia P chip.

From what I can work out it is not the combination of features supported by the chip, but the bus context in whiich it works. I am just unaware of any major moeves elsewhere towards a bus centered design. Any missiing features of wither the S or P chips does niot seem to effect the massive benefit so having the chips addressed by the OS -- surely this gives a practical performance hit beyond any mere increase in CPU speed.

Now I thought the Pegasos had a decent but essentially old fashioned bus design (I could be wrong on this) the board has more features but to my way of thinking cannot exploit HW resources to the possible extent of Articia chips -- I could be wrong but this difference always seemed crucial to me.

Hence in a previous post your suggestion of relabling pegasos boards as a cheap option would seem to be exactly that -- a cheap option but not necessarily a good one.

The present Teron designs make large commercial board designs a possiblity. Banks of G3s exploiting Articia buses, is a pretty powerful combination, especially in any media related area. I think you would need a lot more Pegasos boards running to effect the same practical output in high bandwidth media servers, but again I might be wrong on this.

To me the hope of getting radically cheaper boards for our use better relies on a few big commerical runs, then aiming for a handful of sales on boards which are marginally cheaper but do not have the same commerical potential.

I missed any talk of the CX model by Alan, or forgotten what I had seen. Is there a reference?

If and it seems likely the CX follows the TERON PX2 and the picture you supplied is its prototype, yes it ios no cheapy, but it does look impressive.

_________________
Greg Schofield, Perth Australia

@DrBombcrater

Alan Redhouse posted a rough estimate of costs for the uA1 a while back, and I think the CPU card came in at about £175

He also estimated a cost of £130 for the uA1 boards with a video chip. I sincerely hope it wouldn't cost £384 extra for a larger board with a few more PCI slots and and AGP slot.

Last edited by Rob on 23-Jan-2005 at 03:40 PM.
Last edited by Rob on 23-Jan-2005 at 03:38 PM.

@GregS

Quote:

Any missiing features of wither the S or P chips does niot seem to effect the massive benefit so having the chips addressed by the OS -- surely this gives a practical performance hit beyond any mere increase in CPU speed.

I don't pretend to understand how MAI's designs work in great enough depth to give a definative answer as to the effectiveness of their 'bus-centric' idea, but the proof of the pudding is in the eating, and there's been no proof so far that it is of any benefit.

The performance of the ArticiaS seems to be fairly lacklustre, although I'd freely admit that might be down to implementation-specific problems in other parts of the chip (such as the slow memory controller) that negate any advantage from the floating buffer architecture.

Quote:

Hence in a previous post your suggestion of relabling pegasos boards as a cheap option would seem to be exactly that -- a cheap option but not necessarily a good one.

Anything that would get OS4-capable hardware into people's hands is a good thing, I think. The Peg2 would certainly fill a gap in the market if, as seems likely, the A1-XC is an expensive high-end board. An ATX sized XC with lots of PCI slots would remove some of the arguments for using the Peg2, but not all. There's bound to be a number of potential customers who would feel cramped by a uA1, but can't justify splashing out for an eight-slot monster.

Quote:

I missed any talk of the CX model by Alan, or forgotten what I had seen. Is there a reference?

He talked about it in one of his speeches. The one at Bath in 2003, I think. Back then it was supposed to be a MicroATX model, but that plan seems to have fallen by the wayside. I can't find any signs of an Articia-based uATX board anywhere, but information on the full-ATX PX2 board is not hard to come by.

Quote:

If and it seems likely the CX follows the TERON PX2 and the picture you supplied is its prototype, yes it ios no cheapy, but it does look impressive.

Indeed, it's an interesting product. I've seen information that suggests the board in the picture is an old revision, with the newer cost-reduced version losing a PCI slot, but it'd still be a significant step-up from the XE.

_________________
Who do you serve, and who do you trust? - Galen

@DrBombcrater
Thanks for the reply and I hope in the end the CX has legs and is produced, those 64bit PCIX slots could be real killers, one less PCI (ordinary slot) would not be missed.

Quote:

I don't pretend to understand how MAI's designs work in great enough depth to give a definative answer as to the effectiveness of their 'bus-centric' idea, but the proof of the pudding is in the eating, and there's been no proof so far that it is of any benefit.

The touble we have is that the meal has finished cooking. The bus architecture is one of the main bottlenecks in computing and deserves as much attention if not more than the CPU.

My problem is I have no real idea how well the OS is exploiting its features, it may well be the case that this is a good as it gets, or that OS4 is already exploiting it fully, or not at all (probably something in betwen). Maybe someone from Hyperion could give a clue on this one.

I assume at the moment we are not doing a lot with it, or those parts most useful don't show up in any obvious way. Besides which it is the P chip that seems to have an excess of goodies and onchip media codecs (best place for them as pass throughs within the bus itself (personal bias).

Quote:

Anything that would get OS4-capable hardware into people's hands is a good thing, I think. The Peg2 would certainly fill a gap in the market if, as seems likely, the A1-XC is an expensive high-end board. An ATX sized XC with lots of PCI slots would remove some of the arguments for using the Peg2, but not all. There's bound to be a number of potential customers who would feel cramped by a uA1, but can't justify splashing out for an eight-slot monster.

I like the extremes an XC at one end and a mini-one-slot at the other. Though I think we need to see some progress in commercial deals before it is likely to have both.

I disagree with you logic here. It sounds agreeable, we would like to see OS4 all over the place but at this early stage I think we could be doing a disservice to ourselves.

1) if the exploitation of the bus system is significantly exploited in nthe future, then we would have an offocially santioned "new" amiga that was significantly below performance levels, of course this point balances on an assumption, the Peg2, if my assumprtion is wrong would not be a performance deficit.

2) we could be endangering the link between the OS and a specific hardware combination. I beleieve this could have many effecvts both in marketing and develoipoment in the future. A single set of specs of the officially sanctioned machines centers development and makes it clear ion a developing market what is what. To have two types of new boards that do not share a similar archiotecture (just a common CPU) will at this stage give them equal status. Any how if cost is really the biggest issue, then forget Peggies and wirte it for Mac boxes (Apple would hate it but what the hey).

Plans may have changed, but there is some logic in maintaining OS4 as the community versioin nof the OS, and OS4.1 as the consumer version. In short we should not think about jumping the gun until OS4 is official released, throoughly tested in the real world, a full digest of technical information can be found and we can rewally assess just what we have.

As much as it hurts to watch this thing slowly hatch, why rush things just as the egg is cracking open, moving nests is not advisable, give tyh eofficial release just a little while to recover and then we might be able to asses things. At the moment I think we are too much in the dark to asses the real strengths and eaknesses (besides price) of the HW we have, to go jumping towards an admitably cheaper and good, but "standard" board, like the Peggy., why not go all the way and target Macs.

_________________
Greg Schofield, Perth Australia

It seems to me that the current plan for Eyetech might be to let the MicroA1s become a huge success, providing the special DMA-patch for the current line of A1XEs, even probably making a new batch of A1XE with fixed DMA/USB-issues,
as well as providing high-end CPU-modules like the G4 at 1.4Ghz! That should fill the current "AmigaOne-gap" for at least a couple of months more

@Helgis
It bwould indeed keep things rolling for a bit.

News of fixes for current machines would be very good indeed.

_________________
Greg Schofield, Perth Australia

@GregS

I definitely agree with you There has been much talk about this special hardware-patch, and i would like to know when and how this patch will be provided for the current A1XE-owners, and how they will become a "standard"
in a new batch of A1XEs. That would be very interesting to see.

What will definitely helps a lot, is a great speed-boost, and that is an 1.4Ghz G4 CPU-module. I'm considering one myself, as well as a Sapphire ATI Radeon 9800 Pro Atlantis with 256MB DDR SDRAM - an AGP GPU-card for my AmigaOne

Should be very promising!

@Theodosius

Quote:

Small exagerration there... how were we (me and Bodie) able to fit in a TV card?

With a shoehorn?

_________________
Site admins are people too..pooff!

@GregS

Quote:

News of fixes for current machines would be very good indeed.

Agree, as long it is not a "get a pci-card or external usb-hub.

Also curious about how much it will cost to get a fix for the usb and Via, I guess it will not be covered by a warranty.

_________________
Site admins are people too..pooff!

@Helgis

That 1.4 Ghz cpu-module would certainly be interesting, especially if there will be a dual-cpu one available.

_________________
Site admins are people too..pooff!

@GregS

Quote:

Thanks for the reply and I hope in the end the CX has legs and is produced, those 64bit PCIX slots could be real killers, one less PCI (ordinary slot) would not be missed.

The more slots the better, I say

And I agree the PCI-X slots could be an ace in the hole. Being able to dedicate 1GB/sec of bandwidth to devices like a GB ethernet card or SATA controller without tying up the rest of the system is an excellent facility.

Quote:

My problem is I have no real idea how well the OS is exploiting its features, it may well be the case that this is a good as it gets, or that OS4 is already exploiting it fully, or not at all (probably something in betwen). Maybe someone from Hyperion could give a clue on this one.

My impression, from trawling through the available documentation, is that there isn't anything in the chip to be 'switched on' as it were. The touted bandwidth-friendly features of the Articia exist almost completely transparently from the point of view of the system software.

The performance we have now is probably the best we'll see, except for small optimisations to memory speed or something like that. There appears to be no way of gaining any significant extra speed.

Quote:

we could be endangering the link between the OS and a specific hardware combination. I beleieve this could have many effecvts both in marketing and develoipoment in the future. A single set of specs of the officially sanctioned machines centers development and makes it clear ion a developing market what is what.

I don't think is was ever intended that OS4 should be tied to one very specific architecture. Support for PPC accelerator cards, both the old Phase-5 stuff and the new ACK cards, pretty much blows any idea of that right out of the water.

And I'm not sure that would be desirable, anyway. Nobody cares what kind of North Bridge is in the system provided it works okay and performs in line with similar devices. The only way to differentiate between an Pegasos2 running OS4 and an A1-XC would probably be to take the lid off the system and check what's inside.

Quote:

As much as it hurts to watch this thing slowly hatch, why rush things just as the egg is cracking open, moving nests is not advisable,

Well, there are undoubted advantages to that approach, but lots of potential pitfalls as well. The obvious one being we get what Eyetech (or more correctly, MAI) chooses to give us. Right now this amounts to one not very expandable motherboard, and that must be hurting the market to some extent. Some people are going to keep their money in their pocket because the uA1 simply doesn't suit their needs.

And Eyetech's relationship with MAI means we could miss out of some potentially exciting developments. Freescale has chips now that support PCI-Express and 667MHz DDR-2 memory, but MAI won't be able to offer those features until the Articia-I ships, which could be years from now, and Freescale puts the CPU, memory controller and PCI/PCIe hosts all on the same silicon. That gives a considerable performance boost and is something MAI will almost certainly never be able to do.

The other big gotcha is, of course, the colossal void left if Eyetech were to fail. They are a very small company with little in the way of reserves. Businesses like that go down all the time, often for reasons they have no control over.

_________________
Who do you serve, and who do you trust? - Galen

@DrBombcrater
Good solid replies and thanks.
Quote:

My impression, from trawling through the available documentation, is that there isn't anything in the chip to be 'switched on' as it were. The touted bandwidth-friendly features of the Articia exist almost completely transparently from the point of view of the system software.

Probably my iompression as well when I think about it, but I support a tremendous load of ignorance, my thinking is this, we need some working everyday software examples to see the benefits, or soime OS switches or additions, they may well be there but invisible for these simple things.

Long before I rejoined the Amiga community, I have seen the Bus architecture as one of the main hurdles in actually realising the potential of boith software and CPU. From day one (once I found out about it) the MAI approach sounded spot-on.

Freescales (which I have not looked at) may well do things better, but that is always the problem when you ride a fast horse, a faster one can join the field at any time. The bus architecture to me is a distiunguishing feature of leading edge OS4 development, these are early days still we can try and chase every faster horse gain the saddle and ride on, or stick to what we have and make the best of it -- I am cautious so I follow the latter path, we are on the right breed olf horse, it runs fast even if it is not the fastest and up ahead there will be room to add to our stable if needed.

Yes we are at the mercy of Mai and others HW, the obnly thing that changes that is by becoming a bigger Hardware market ourselves. The worst thing is being a begger at every body's table because we have not stuck with one source long enough to become important to it.

Mind you if we do hit the big time them we start calling the tune and we can easily have HW suppliers competeting to give us the best fopr the OS.

My fear is that chasing better and cheaper weakens our medium term standing.

If Eyetech fails we are in a worse but not fatal position. All we have to do is remain steady for the first phase the initial release of the OS, a bit of time for Eyetech to make some commerical deals to lower prices and then re-asses things. Applying pressure to change at this stage simply does not help, we need to let things mature and then see where we are.

My nightmare is not Eyetech dying, but the OS dissapating amongst HW platforms, none of which need the OS, and we never being for any one of them a big enough market to have any power. The BeOS fate is not a pleasant one.

At the moment and untill some comercial deals are or are not made, Eyetech needs us (the immediate OS HW market), that is essentially good for us. A commerical success or two and the HW becomes cheap and then there will be a lot more of us.

We grow as it grows, each commerical success effects the pricing, each drop in pricing will increase the community. We simply cannnot get this sort of relationship elsewhere, we would always just be buying someone's cheap boards but never become mouch more than a dumping ground.

So above criticisms and legit or otherwise, one thing stands out in my mind, the good logic of the alliances that together are bringing together OS4 as HW and SW.

_________________
Greg Schofield, Perth Australia