Basically, we're going for a strategy that re-converges the two operating systems that we created for the STB and NC - that's going to converge with the operating system for the StrongARM, and we're going to bring those together for a new generation of StrongARM chips. Then we'll go for a final version for both StrongARM and other ARM chips. The Galileo microkernel that exists today will then move forward, which will first of all be hosted on StrongARM and ARM7500FE, though some future generation of chips and then at some point in two years' time, er, two and a half years' time Galileo takes over from RISC OS. A very nice way of moving forward over two and a half years. Not to mention this is all subject to change.
A new system controller chip is being fabricated with Cirrus in 0.3 micron technology, and that will replace IOC. The system will be PCI based, and podules will be able to be supposed via an adaptor bridge. It will use a non-PC case, but the case itself is yet to be designed. Whether it's a Risc PC case or not, the jury's still out on but we hope, er, it definately won't be a PC case. It'll support contemporary DRAM technology, so a minimum of EDO and maybe SVRAM, and it will support a minimum of VRAM and it may go to syncronous SVDRAM as well. There will be processor support for 64bit CPUs - so Pentium type processors will be supported by the architecture. There will also be DVD, which is the new standard for replacing the current CD ROM standard.
(Ed: Rumour has it that the 'sugar daddy' from America ordered 150,000 units!)
I went through the clippings from last year - we run a clippings service internally - and here are some quotes [in the press] from last year: "Acorn will not exist by December" and a generous person said April 1996, "Acorn will not make any more products", "RISC OS is dead", "Acorn's technology has no use in the real world, and Windows PCs are the way forward." So we said "Sod you". Today we are far stronger than we've been for a long, long time. We're spending more on R&D; than ever in our history. We have customers and credibility on a world wide basis, and we think we have a brighter future than at any point in the last five years.
So, in 1996, we're very grateful to you for supporting us through our transition and change as we move forward, and keeping buying our products. In 1997, we'd like to repay that buy giving you some new products that you have asked for - we'll give you some more technology - and we obviously hope for more commercial success.
One of the things that make StrongARM so good is, first, it's near enough identical to previous ARMs - so software works! Really amazing stuff, that. The next thing that really makes a huge difference is that it's clock at - instead of an ARM710 clocking at 40MHz - StrongARM clocks at 202MHz. That's where the vast bulk of the brute speed comes from. But in addition to that, they've also done a lot of tidying up and refining of the internal architecture like the way that instructions are pipelined, and there's a separate 16k instruction and data cache. When you start talking about ARM code, the size of that cache is really very large for the sort of applications and software we run. When we first started running StrongARMs, we were completely amazed that some applications tended to get ten times speed increases - BASIC programs - over a 710. And we were trying to work out why this should be. It turned out that after a bit more investigation that the reason they were going so fast was because the interpreter sits inside one of the caches. When it interprets basic everything is internal - that's the reason they fly so well.
In addition it has a write buffer on the output, and it's the combination of the caches and the buffering that means within the Risc PC we get such a high level of performance, because it's actually buffering quite a lot of the output back into the system. The multiplier on the StrongARM is dramatically better because of Digital technology - they know how to do these things... Remember that this is running at 200MHz, so you're looking at 10 to 15 nanoseconds per mulitplier, and they've generally tidied up, and cleaned, and done a few more constructions.
One of the really neat things about StrongARM as opposed to our previous ones is that the clock is actually generated internally. Internal clocks are are an extremely difficult things to handle, and it only has a 3.68 crystal on it. And you can change the clock speed - not when it's running, but when the power is off. We've put the dip-switches on the board to allow people to change the clock speed of the StrongARM (Ed:from 88MHz to 287MHz!).
We've updated the version of RISC OS. The basic refinements are to make it StrongARM compatible... we actually had to do quite a lot of work. One of the things we've done is to improve the task swapping - the kernel is now aware of application tasks, and we've improved the task swapping there. We actually included a major rewrite of the internet stack, which goes a lot faster. But, as well as getting the StrongARM benefits, the whole machines goes faster as a result of some of these things we've done. The font manager is capable of working over the top of multi-coloured backgrounds.
The StrongARM is just brilliant to use - there's just nothing like it. Once you've got it going everying goes faster. Most applications work; some need upgrades (details are available on the web site)... Once you've used the StrongARM, if you go back to a 710 it's like ugggg! You can't use that.
Please report any problems to:
[email protected]
© Quantum Software 1996