log in | register | forums


User accounts
Register new account
Forgot password
Forum stats
List of members
Search the forums

Advanced search
Recent discussions
- Rougol June 2024 meeting on monday (News:1)
- WROCC June 2024 meeting - Mark Moxon dissects Lander (News:1)
- WROCC June 2024 meeting on wednesday - Mark Moxon on Lander (News:1)
- June developer 'fireside' chat is on saturday night (News:)
- May 2024 News Summary (News:5)
- Archive Edition 27:2 reviewed (News:)
- RISC OS 5.30 arrives (News:12)
- Rougol Talk May 2024 - Andy Vawer (News:)
- Rougol May 2024 meeting on monday with Andy Vawer (News:1)
- WROCC May 2024 meeting - Gerph talks games (News:)
Related articles
- Wakefield 2006 show report
- 50,000 shares, Iyonix Select and a Belated Happy Birthday
- Wakefield 2001 show report
- Wakefield 2005 show report (pictures)
- Wakefield 2002 show report
- Wakefield 2004 show report
- Wakefield 2005 show report
- Omega LegPuller at ROUGOL meeting
- RISC OS Expo 2004
- Omega network shipped
Latest postings RSS Feeds
RSS 2.0 | 1.0 | 0.9
Atom 0.3
Misc RDF | CDF
View on Mastodon
Site Search
Article archives
The Icon Bar: News and features: The XScale has landed

The XScale has landed

Posted by Richard Goodwin on 10:22, 2/4/2001 | , , ,
More interesting news from the new and improved MicroDigital website - the first batch of 600 MHz XScale processors have arrived.

Although work on getting these processors working has already well under way using the manufacturer's data sheets, actually getting hold of the chips themselves is a crucial step. The daughterboards that have been designed to hold these chips can now be produced and tested, with an expected timescale of three weeks.

If everything goes to plan this means that there should be something to show the public well in time for the Wakefield show on the 19th and 20th of May.

Thanks to Jon for the tip!

Source: Microdigital news page

  The XScale has landed
  (10:25 2/4/2001)
  Sendu Bala (17:34 2/4/2001)
    Michael Stubbs (19:07 2/4/2001)
      Ian Redman (19:23 2/4/2001)
        Jon Hall (21:40 2/4/2001)
          Sendu Bala (21:50 2/4/2001)
            Chris Williams (00:53 3/4/2001)
              Rob Kendrick (00:58 3/4/2001)
                Paul F. Johnson (01:22 3/4/2001)
                  Sendu Bala (02:19 3/4/2001)
                    Richard Goodwin (10:14 3/4/2001)
                      Paul Johnson (19:04 3/4/2001)
                        Michael Stubbs (19:11 3/4/2001)
                          Nathan (22:44 3/4/2001)
                            Steve knutson (06:34 4/4/2001)
                              Chris Williams (08:08 4/4/2001)
                                Robert Richards (08:58 4/4/2001)
                                  Richard Goodwin (10:06 4/4/2001)
                                    Rob Kendrick (16:58 4/4/2001)
                                      Annraoi (20:33 4/4/2001)
                                        Michael Stubbs (14:52 5/4/2001)
Richard Goodwin Message #88426, posted at 10:25, 2/4/2001
Unregistered user Although there's still some work to do - like getting some of the link-type images at the top of the page actually linked - the new site is a definite improvement on the old, don't you think? Not exactly flash, but a lot cleaner and together.
  ^[ Log in to reply ]
Sendu Bala Message #88427, posted at 17:34, 2/4/2001, in reply to message #88426
Unregistered user Ugh, no they _are_ linked, it's just that only the little gif dots are linked, not the text.

Is all the pricing info new? I don't remember that. Good price. Thank god.
  ^[ Log in to reply ]
Michael Stubbs Message #88428, posted at 19:07, 2/4/2001, in reply to message #88427
Unregistered user Well, it's an improvement over the old one. Personally, I would have kept the old, old one (before the new one that came with Omega announcement). That one looked nice and worked well.

There's a few things the Omega 'does' that has yet to be tested by anyone including MD - such as using XScale and switching modes. That they have claimed it does this without a chip to find out is, in my opinion, a questionable thing to do.
  ^[ Log in to reply ]
Ian Redman Message #88429, posted at 19:23, 2/4/2001, in reply to message #88428
Unregistered user Surely they could have tried it with two normal SrongARM's?? one running in 26bit and the other in 32bit?
If they did though, is another matter....
  ^[ Log in to reply ]
Jon Hall Message #88430, posted at 21:40, 2/4/2001, in reply to message #88429
Unregistered user They can't spell 'meticulous' though!
  ^[ Log in to reply ]
Sendu Bala Message #88431, posted at 21:50, 2/4/2001, in reply to message #88430
Unregistered user Speaking of pricing, why on earth are their memory upgrade prices so high? Crucial is selling unbuffered PC133 SDRAM (256MB) for around £70 and registered for around £80. Quite a disparity then for Microdigitals asking price of £256!
2100 DDR costs only £90 for goodness sakes.

Oooh, but finally there's a short description of how the dual processors will work. Now I _am_ excited. I was thinking I'd have to wait till RISC OS 5 (ie. till the end of the world) before that XScale would actually do anything. Hoorah! Might even be worth buying now...
  ^[ Log in to reply ]
Chris Williams Message #88432, posted at 00:53, 3/4/2001, in reply to message #88431
Unregistered user You are forgetting that we also need some 32bit software to run on a 32bit machine. Wave goodbye to your expansion cards, artworks and possibly OvationPro. It is going to be one heck of a step and I don't think (IMHO) MD are taking it responsibily.

Chris (speaking personally and with an opinion I am entitled to.)
  ^[ Log in to reply ]
Rob Kendrick Message #88433, posted at 00:58, 3/4/2001, in reply to message #88432
Unregistered user Yes, very little software currently available will be able to take advantage of the (possibly very expensive) XScale upgrade. This causes a problem. If there are no customers, software houses may think twice about spending time and money making there software 32bit OK, and with no software, who will want to pay a fortune for a useless 32bit processor...
  ^[ Log in to reply ]
Paul F. Johnson Message #88434, posted at 01:22, 3/4/2001, in reply to message #88433
Unregistered user While it's great that we're getting a 32 bit processor, it does worry me the amount of delays and state of MD's former machines. The Mico is (as far as I am aware) unfinished with drivers and software desperately missing.

I would love to see the 32 bit system working, but with this sort of past history, it does not fill me with confidence that we will not get all that is gold, but something which is gold in colour, but distinctly "leady" underneath.

Don't get me wrong. MD are on the right track, I just think that it's being approached the wrong way. The boards should be fully tested with both chipsets running a 32 bit version of RISC OS. It hasn't. I personally would have done as Castle have been doing and play everything close to the chest instead of stifling the market (as Pheobe did) waiting for a machine which promises lots, but as yet, is months behind schedule (wasn't it promised for delivery before Christmas?)

(speaking personally)
  ^[ Log in to reply ]
Sendu Bala Message #88435, posted at 02:19, 3/4/2001, in reply to message #88434
Unregistered user No Chris, I'm not forgetting. Most software will need a simple recompile at worst. Even for software that is no longer being developed, I won't have to 'say goodbye' to any of it because of the onboard SA. That's the whole point of having the two processors.
But for software - new software esp. like Vantage - there'll be a damn good reason for getting that XScale upgrade. That's what I'm excited about. I can get the speed increase without having to wait for RISC OS 5.

I don't think there'll be a problem going to 32bit. Software developers can easily make the change with their next revision and XScales can only sell like hotcakes. We haven't had a new processor since '96 for heavens sakes!
  ^[ Log in to reply ]
Richard Goodwin Message #88436, posted at 10:14, 3/4/2001, in reply to message #88435
Unregistered user IIRC anything written in BASIC should work in 32bit mode straight out of the box (if BASIC works...), so my stuff like HTML3 will fly :))
  ^[ Log in to reply ]
Paul Johnson Message #88437, posted at 19:04, 3/4/2001, in reply to message #88436
Unregistered user Software broken by 32 bit currently in large scale use = anything by Computer Concepts. Any software written using libraries (such as RISC_OSLib) which have not been recompiled using the 32 bit clean versions, modules which are preserving flags or using old 26 bit trick to return to user mode

Anything in BASIC will still work. Anything in C (which have been recompiled using 32 bit clean libs) will work. Anything in Assembler which is APCS-32 compliant will work. Anything else - won't.

Sorry about that.

I imagine it will almost be as bad as moving from RISC OS 2 to 3.1 again.....
  ^[ Log in to reply ]
Michael Stubbs Message #88438, posted at 19:11, 3/4/2001, in reply to message #88437
Unregistered user I have to agree with Paul about cards close to chest. Much more professional and better for the whole market. Maybe MD are worried that Castle will release an Omega-killer of a machine anytime soon ;-)

*If* it really is possible to do this 26/32bit switching, then we should retain all old apps (unless apps like Sibelius and CC apps don't like the MD graphics chip). But a lot of apps like Sibelius and CC stuff is written in assembler which requires more than a simple recompile! Also, most of these companies refuse to do simple things like that as the most they want to do with the Acorn market is flog their old products to anyone who asks. Which is quite sad :(

Personally, I will keep my Kinetic to run the 26bit stuff and buy a new Castle machine when they release one to run XScale stuff. Until RISC OS 5 appears, Omega offers no benefit (to me at least) over a Kinetic RiscPC.

  ^[ Log in to reply ]
Nathan Message #88439, posted at 22:44, 3/4/2001, in reply to message #88438
Unregistered user If there isn't some way of using 26-bit apps then they are shooting themselves in the foot. I don't think they are going to be that dumb.
  ^[ Log in to reply ]
Steve knutson Message #88440, posted at 06:34, 4/4/2001, in reply to message #88439
Unregistered user Interesting discussion.

If we are to get more horse power to run CPU intensive jobs such as MP3 compression, MPEG video and faster games, then we need 32 bit processors.

Just like any other change in processor technology some older apps will nolonger work. Microdigital have choosen an excellent option in my opinion to provide backward compatiability. Apple used emulation when the Mac when from 68K to PowerPC. Parts of the OS were run in an emulator until the code was move to PowerPC.

The only way forward is 32bit addressing, a side effect will be some software won't run in a pure 32 bit environment, hence the need for 2 processors.

If a large number of people are using Impression and the like, then I see that as an oppurtunity for someone to give those users an upgrade path. Some may move to other OS's, but if new (good) software comes along to replace it then people may move back to RISCOS too. Are Artworks users on of the markets Vantage is aimed at? I would guess YES!
  ^[ Log in to reply ]
Chris Williams Message #88441, posted at 08:08, 4/4/2001, in reply to message #88440
Unregistered user Let's take a step back and look at things from a greater perspective. Who owns RISC OS? Pace. Why do Pace need RISC OS? For (atm) STBs. Does a STB need dual processors? No. So will RISC OS see dual processor support? No.

Thus, anyone wanting dual procssor support will need to write that part of the OS themselves. Will ROL do it? I don't think so, they have other things to fix/update. Will MD do it? If yes, they will need a license to modify and distribute RO. Have they applied for such a license? No comment.

As you can see I don't like the idea of a dual processor system. It is a nightmare since the OS will be in 32bit, the XScale and the hardware will be in 32bit but the SA will be 26bit.

But erm, anyway to end my negitive ramble I'm afraid that if something is written in pure BASIC it'll work in 32bit. The language we've all been trying to move away from has turned out to be the most secure langauge for RISC OS...

Chris (IMHO, speaking personally and all to the best of my knowledge)
  ^[ Log in to reply ]
Robert Richards Message #88442, posted at 08:58, 4/4/2001, in reply to message #88441
Unregistered user If you read the info on the MD website - it says that switching tasks betwen 26 and 32 bit processors is done by their custom hardware, independantly of RISC OS.
Doubtless someone has designed an extremely clever piece of logic to implement this and fool RISC OS into thinking that it's only running tasks on one processor.
So RISC OS doesn't necassarily need to 'support' multiple processors if it's being done with custom logic.

The Omega sounds like a relly nice machine - excellent graphics capabilities and hardware JPEG and MPEG decoding. Nice!
  ^[ Log in to reply ]
Richard Goodwin Message #88443, posted at 10:06, 4/4/2001, in reply to message #88442
Unregistered user I don't think anyone needs a RISC OS license to add custom hardware support - I thought one of the thrusts of the OS was to supply a basic OS on ROM, and then have additional ROMs to add functionality like graphics hardware support? That way the different hardware companies can innovate without RISCOS Ltd being swamped with work trying to hack support into the main OS.

Mind you, that was last years developers conference (or possibly even the year before that), things could be completely different now :)
  ^[ Log in to reply ]
Rob Kendrick Message #88444, posted at 16:58, 4/4/2001, in reply to message #88443
Unregistered user One other interesting point that nobody seems to have mentioned is that the API in a 32bit RISC OS will have to be different. Any SWI that takes flags in a register that also has an address in it will now be completely defunct. Does this magic chip of MicroDigital all translate SWI calls for programs running on the 26 bit processor so they work on the newer RISC OS? If you recall what is said in ROL/Pace's own 32bit developer update, there will be *NO* intermediate state between 26bit and 32bit. It will be one or the other. If this holds true, then a system that allows 26bit software to work on a 32bit RISC OS might need to a lot more munging than just sending it to a different processor...
I'm not sure about the above technically, but it might be worth thinking about.
  ^[ Log in to reply ]
Annraoi Message #88445, posted at 20:33, 4/4/2001, in reply to message #88444
Unregistered user I humbly suggest that MD focus on getting Omega up and running (even showning an SA only prototype at Wakefield would be good).

According to their own (old) website they weren't due to receive their motherboard PCB's until April 12th, so it seems a little premature to be talking about testing the xScale daughterboards as there's nothing for them to be plugged into yet !

Best not to raise peoples hopes until the thing is all assembled and running and when there's something to show.

And yes the website is much MUCH better !!
  ^[ Log in to reply ]
Michael Stubbs Message #88446, posted at 14:52, 5/4/2001, in reply to message #88445
Unregistered user MD obviously don't share your views, Annraoi. Get something designed and working before getting people's hopes up? That's silly - best take deposits, stifle sales and *then* think about all that nonsense.

So far, MD have promised a lot and delivered nothing.
  ^[ Log in to reply ]

The Icon Bar: News and features: The XScale has landed