Mike Clarkson
The creation of a workstations division in Acorn's most recent restructuring caused some surprise to many observers, since it rather seemed that Acorn was moving away from making products (especially anything which could be called a workstation) towards licensing technology. The recent Wakefield show gave the opportunity for Acorn to explain the move and a little more of their future plans: what follows is what I gleaned from the various presentations and from making some individual enquiries.
Acorn see the NC market splitting in two towards the end of this year.
The 'low end' will be aimed at the home market, and will resemble the work formerly done by Acorn's OnlineMedia division, namely set-top boxes, plugging into a television and one or more of a satellite, cable and telephone socket. The box may offer digital TV, TV-on-demand via cable, and/or internet access (web browsing and email). My impression was that Acorn are unsure which of these facilities will be required when - certainly there is a feeling that a "one box does it all" solution only confuses the end-user, who tends to want one thing at a time.
The good news, from Acorn's point of view, is that they can deliver pretty much whatever is required as it is wanted, currently based around the production NCs running on the ARM7500FE. It was interesting to see, for the first time, NCs being demonstrated which allowed access to the RISC�OS desktop - one school apparently has a network of NCs being used for RISC�OS, running pretty much as an A7000+ network machine would, without the interfaces, for rather less cost.
For video work, the 7500 hardware is limited, and Acorn have therefore developed the SA1500/1501 StrongARM processors (and Peter Bondar was keen at Wakefield to emphasise that Acorn played a large part in developing these chips).
Although it wasn't on display at Wakefield, Bondar gave details of the system which has since been on show at the Cable and Satellite show in London which uses these new chips to decode twin MPEG videos in software; Acorn's competition are all still using hardware solutions, and (were) claiming that decoding even one MPEG stream in software wasn't yet possible. Acorn are claiming that to do this in software can be cheaper, and is much more flexible - so a video-on-demand STB can very quickly be turned into a Web box etc. Once again, Acorn seem to have the edge in the technology; they now seem rather more intent on marketing it than in the past.
The 'high end' NCs are the machines Acorn see as being adopted by companies, running over an intranet. At least in this market "we're staking Acorn's future on Java" (Peter Bondar) - and therefore a great deal of Acorn's corporate energy is being spent on having the best Java solution in the market. They are therefore expecting to be second only to Sun in producing a JVM (Java Virtual Machine - !Java to RISC�OS users) for Java version 1.2. The current Acorn !Java supports version 1.0.2, while other platforms are on 1.1, but Acorn are bypassing 1.1 altogether, to be first with 1.2. Secondly, they are coding the JVM in ARM code, rather than C, to make it as fast as possible.
Further, my understanding is that StrongARMs are�particularly well-suited to running Java because (I�think this is right!) the compact nature of ARM code means that the kernel of the JVM can reside permanently in the processor cache. All of the above Java developments should be good news for users of RISC�OS desktop machines - although, of course, they all depend on Java being accepted and adopted as widely as Acorn hope and expect. Bondar did mention great interest in Java among large corporations, such as banks, and others have commented that such companies, perhaps currently based around Unix, are unwilling to put all their eggs in one basket, and are therefore wary of Microsoft-only solutions.
So what of actual NCs? I asked Dave Walker (Acorn's Engineering Support Manager) if the CoNCord StrongARM-based NC was going to be commercially released. It isn't, because Acorn plan to release a new StrongARM Java-oriented NC based around the design of Phoebe, whereas CoNCord was essentially RiscPC technology. They see this new NC being available around the end of this year - it sounds as though it may be Phoebe without the CD, hard drive, podule and PCI interfaces. If there is a faster StrongARM around by then (as seems highly likely), Acorn might have a very powerful, cheap Java machine.
Reasonably full details of the spec. of Phoebe appeared in last month's magazine, so I will only add a little. The speed of the processors inside the first machines is still open to debate. Officially, the current 233MHz chips will be used, but it was quietly being said that "others will be available pretty soon". Not the SA1500, though, which is an integrated chip aimed at different markets - NCs and set-top boxes, as mentioned above. The board speed of Phoebe (64MHz) is chosen for two reasons: any quicker causes real problems in reliable construction (apparently, above this speed, you start losing electrons round corners etc - indeed, the current board has been designed with no ninety degree corners in the circuitry to avoid this even at 64MHz), and secondly 64MHz is also the fastest speed that current StrongARMs can synchronise with, externally.
One of the features of Phoebe that has caused most interest is the "support for multiple processors". The exact nature and extent of this is unclear (to me at least); it does seem clear that RISC�OS is unlikely to support multiple processors for some time, if ever (see below). Dave Walker was keen to see a flavour of Unix (either RiscBSD or ARMLinux) running on Phoebe, since Unix does have support for multiple processors; it certainly seems likely that Unix will be the first, if not the only, operating system to use multiple StrongARMs in Phoebe. Of course, multiple processors were on the agenda a few years ago when the RiscPC first came out, in the form of the Hydra multi-processor board; the unexpected arrival of the StrongARM made that work appear redundant, but some of it may now be called out of cold storage.
An interesting comment from Chris Cox was that "for various reasons" RISC�OS is going to have a longer shelf life than expected - "and quite a lot of work has therefore gone into developing it"; the fruit of this can be seen in the new filecore (virtually no limit to files per directory, or to length of filenames, and better support for large hard drives), and the fact that various bits of the kernel have apparently been rewritten, as has the floating point emulator (which has been re-coded in ARM code to take advantage of StrongARM processors).
It is slightly hard to tell what is going on here, and there were one or two hints that Acorn is going to consider the situation again after the launch of Phoebe (perhaps also depending on the success of Java), but it now looks increasingly likely that Galileo - Acorn's new operating system - may never make it to the desktop at all; it has, after all, been designed for "deeply embedded" applications, and a RISC�OS emulator (necessary to run RISC�OS software on top of Galileo) didn't run fast enough to be particularly viable. So RISC�OS may be here to stay rather longer than anyone expected.
Not surprisingly, ARM are working on it (and so are Intel), and are actively working on finding new semiconductor partners. Acorn/ARM (Bondar kept talking of both companies being involved in development of e.g. SA1500/1) are currently talking with Intel, and happy to be working with them, but are also looking to avoid the problem of relying on only one manufacturer. Bondar's comment was that no-one had expected a giant like Digital to "disappear overnight", but ARM and Acorn got their fingers burnt and don't want it to happen again. Expect some announcements "in the next few weeks" over StrongARM licensing/partners. Presumably he was referring the Samsung deal which has been rumoured but not confirmed, because Matsushita (=Technics/Panasonic) have only, as far as I can see, licensed ARM7, and not StrongARM.
So where does that leave the workstation division? Well, the NC family comes under this banner, although it is also very much part of the licensing division, and Phoebe is clearly the showpiece. Chris Cox was keen to stress that Acorn would like to lure current PC/Mac users to the Acorn platform with Phoebe, which rather looks as if it will depend on Java�software becoming available, although they may�feel that Photodesk, Ovation and the like can compete with the PC world. However, Phoebe is also�expected�to be used as a server for an NC network, presumably running intranet software either�under RISC�OS, or perhaps under Unix (with�multiple processor support?). It will be an interesting year before Wakefield '99.
Again, just a few details for those who didn't make it to Wakefield and haven't read about this elsewhere. The Peanut is the new RISC�OS portable, to be produced (as Paul hinted last month!) by Innovative Media Solutions (IMS). In many ways, this is an A7000+ repackaged as a portable (much as the A4 was a portable A5000). However, that isn't to belittle the work done by IMS, which has included designing�a motherboard (the A7000+ is too big), writing keyboard and video drivers, and tweaking RISC�OS 3.7 (licensed from Acorn) to add features associated with a portable (sleep mode, battery management etc).
The Peanut is expected to ship with 16Mb of RAM (expandable, probably, to a maximum of 32Mb), a CD-ROM and IDE hard drive (both at whatever spec. is most cost-effective at the time). The processor will be the ARM7500FE running at 48MHz, as in the A7000+, while the screen will be a�12.1 inch TFT LCD at 800x600 in 256 colours. A�glidepoint will be used rather than a mouse, and there will be a standard floppy drive and ports for an external monitor, keyboard and mouse, together with serial and parallel interfaces. The price is expected to be around £1500+VAT.
Altogether, this looks like an excellent machine, and there are many who hope it will do well - it certainly generated a great deal of interest at Wakefield, where a working prototype was demonstrated running various RISC�OS applications; it should, I think, be available around September (like Phoebe). There are, however, also many who have been asking questions of the specification, some of which I will try to cover.
The board has been designed by IMS, and designing�a StrongARM board would have been much more difficult and expensive. The 7500 chip is an integrated one, whereas the SA110, as found in RiscPCs, requires additional chips to control memory, input/output etc. True, the SA1500 is a (very fast) integrated chip, but there is no support in�RISC�OS for this chip, and it doesn't quite fit the bill anyway.
32,000 colours does remain a possibility, but would require further design work, and on the LCD screen seems a little unnecessary. The full range of resolutions/colours found on an A7000+ should be possible using an external monitor.
PCMCIA is a standard PC interface for portables, modems, ethernet etc. Although Acorn began developing RISC�OS support for this when they were�developing their now defunct portable (Stork), this work was never released and doesn't seem to be available to IMS. The hardware support is expected to be there, but the software would require quite a bit of development; it might happen - and maybe someone other than IMS might do it?
This seemed to be the most frequently asked question on the IMS stand at Wakefield. They certainly seemed to take on board the number of requests, and are currently looking into solutions, the most likely being one based around the parallel port, it seems.
I have to confess to understanding this less well (and�I�don't think there's been a definitive answer from IMS), but it seems that 16Mb of RAM will be soldered to the motherboard, with sockets for another�16Mb. I think the reasons for this are that higher-capacity RAM chips would consume more power (shortening battery life, and possibly causing problems by running at a different voltage?), and also�would take up more space on the board - space in a portable is very precious!
Of all the concerns raised, this would be my greatest one, except that RISC�OS applications are so compact that 32Mb should be enough in most situations. I rarely use that much, even for high end DTP - but then I still remember wondering why the chap was trying to persuade me to buy a disc drive capable of holding 200Kb per 5¼" disc!
Also on display at Wakefield were some of the consumer devices made for Samsung, which are aimed at the under 7s - some sort of multimedia interactive games with educational content - edutainment, if I must.
The NetFax was also on display, although I didn't quite see its use. My previous understanding was that it was a device to compress faxes and send them at much higher speed than a conventional fax machine, possibly over the internet, thus making for much lower phone bills (but only if both sender and recipient have NetFax capability - which is why the US commissioning company are keen for it to be adopted as a standard by fax makers). However, what was on display at Wakefield seemed entirely different, with an LCD display about the same width and half the height of a Psion 3a, and a very small keyboard (smaller than a 3a). Others may have rather more ideas about what the expected use of this device was? Watch this space.
Bondar also said that Acorn were working with lots of OEMs, "like Sony", quickly followed by "just by way of example" - Sony, who have recently licensed the ARM7 from ARM.
Finally, it seems that the National Lottery machines are "based on RiscPCs" (Bondar) - according to the newsgroups, it's not the vending machines (although this is the sort of embedded appliance that Acorn would love to get into soon) - but Lancelot, Guinevere etc - the ball choosing machines.
Author: Mike Clarkson
Published: Archive 11.10, July 1998.
Contents - Archive Magazine - Articles Index - Phoenix News - Peanut News