As it appears that RISC OS compatibles come now with Podule, Microbus, ISA, USB, etc it would be very useful if there was a piece of software that would allow you to create a driver for the ISA (etc) device that you had just plugged in.

If the suite could interrogate the card after being told what type of card it was (ie network or modem etc) and using the various standards (for example tcpip for a network card) allow a driver to be created to use the device. (Maybe using basic or something else in the public domain).

Any input on this one, as my knowledge of driver writing is minimal and I am aware that manufacturers of cards and other devices spend large amounts of money getting drivers written.

Hiowever if such a piece of software were to be possible then programmers could concentrate on other things leaving knowlegeable users to create device drivers and post them on an ftp site......

Also your suggestion implies that the suite knows something about the type of card it is trying to help create a driver for. This means it needs to be updated in someway when new types of cards become available. (Although I can imagine this wouldn't be too difficult to do).

I suspect the easiest way (as ever) would be to port drivers from linux where the source has been made available.

Point about hardware valid and noted, but it's still necessary to be able to use things like Excel and Access (yes, YUCK, but still necessary) and a RISCOS machine (PCI slots or no) is not going to provide that without the 2nd Proc. slot.

Ya think?

Such a card could feasably be made to work on a RiscOS based machine.

I could be completely wrong about that, but the card does exist.

When RISCOS programs do crash or throw up errors, it would be useful for the error message to make it clear which program was at fault (this is almost never the case).

Not that we should be concentrating on errors with RISC OS (but my RiscPC has some timing/RAM/PCcard illnesses so I get them more than most :( ... )

For all the work Microsoft has put into Win2K it still had 63000 bugs (its own figures) at launch.

Although there may be difficiencies in RiscOS it is also difficient in the number of bugs it has - and that is no bad thing

Having said that, there software being so crap is a tool that can be used against them. RISC OS is incredibly reliable and should be touted as such to the public at large.

1) Win must run on a lot of machine type... Risc OS not!
2) Win Structure is VERY DIFFERENT from RISC OS structure... Win run both 16 and 32 bit apps RISC OS only 26 bit apps, Win have protected memory, multitrhreads, etc... and RISC OS not!
3) when we will have 32 bit architecture we will talk again about an "emulator"! how can peoples think about an "Emulator" of 26 bit? well now i explain something about the future...

We will have S.A. at 600 Mhz (about) with low electricity use (this is the Intel way now... ) maybe we will have out structure a 64 bit (like pentiums 32 in and 32 out in parallel) so actual boards will never use this processors... (maybe it will have also write-trought chace)
ALL ACTUAL BOARDS AND EXPANSION CARDS will not function with the new processors... Intel market is not interested to our world (RISC OS i mean) it is too little...
Maybe Castle etc.. have not enginers for work arround new processors... and this is also the reason for what aleph1 are try to develop Linux and did make an offers for enginers (i think)
IF RISC SO Ltd will survive.. well they have to do a full 32 bit system (but i don't know the future machines so...) for have 26 bits compatibility they will make a second kernel at 26 bit (like the 16 bit core of Win) this is the best solution in speed terms... this will port a lot of problems in the system so they will have to use protected memory and multithreads for let the machine go without crash... (or maybe without a lot of crash) and for use old expansion cards...
For PCI etc... well the most work of an hardware company is about the driver not on the hardware of an expansion card so there is something one here that think that PC company will make RISC drivers of their cards PCI etc...? company only now started to make drivers for WinNT and WinNT 5.0 use a tecnologi called VDM (Virtual Device Manager) that let it use win9x drivers otherwise it have not drivers etc... and so why company have to start to look at RISC OS? well friends i'm honest i don't think that we could use PCI cards also if we will have PCI tecnologies...
The true is this... we have to wait that our companies find their mind otherwise RISC OS will die... we are not more users so the companies have to find an accord and work togeter and we have to help they in this hard work... we have not a big company like Intel or Microsoft so how can we think about the future... computer market is first a high business place now and after tecnologies... (this thank to Gates! :-( )

Running RISC OS on an ARM card on a PCI bus is better but still problematic. It would require a major RISC OS re-write (to support hardware abstraction) some existing software would still not work. In addition you would still have to BOOT windows first (windows would handle the I/O). So you wait 30-60 seconds to boot Windows, do a further RISC OS boot, you're running a RISC OS app and Windows crashes taking RISC OS with it - you reboot another 30-60 seconds ----- is this MAD or what ?

On top of that to develop drivers for Windows and PCI tools are available and there is (a proprietary) model called VXD (as Paulo points out) in windows. The problem is you would probably have to develope it under Windows VXD's are bound to windows so its a non-solution for Acorn users (unless they dump RiscOS that is). Driver software also requires the active assistance of the OS vendor (RISC OS Ltd) and card developers and I just can't see Creative or Matrox or others spending their money developing software or giving their time to a minority platform.

Our best solution is to improve the platform we have and realise that just because something is used by a lot of people does not mean its the best or only solution.

However, Evolution is coming and this should solve many of the hardware problems and limitations of the current RPC.
Depending on it's price, I'm hoping to buy one.

Well running BASIC V on the 3010 and SA you get just over 8 and 2.2 seconds execution times (3010 and StrongARM respectively) and on the P-III running MS QBASIC (full screen dos session under NT 4) a less emphatic 5.3 seconds. What is wrong with this picture ?

Mad Nurse and others may drool over the clock rate of the P-III but the reality is that PCs underperform (and by some margin) for the clock rates they run.

In short 1MHz on an ARM is NOT EQUIVILENT to 1MHz on a P-III.

Even doing a check on the hardware differences it strikes me how much hardware and high clock rates are required to make a 667MHz processor (P-III) just about 1.5 times the speed of a 6MHz one (the ARM-250 in the A-3010).

Please consider this:

ARM-250 6->12MHz {Intel P-III Coppermine (667MHz + 133MHz FSB)}
ARM-250 NO CACHE AT ALL {P-III 667MHz (Full speed) CACHE 256Kb}
ARM RAM access speed 60nS {P-III RAM access speed 6.7nS (nearly 10 times faster)}
NO accelerated video graphics {AGP card based accelerator}

The ARM250 managed just over 8 seconds the PIII manged 5.4, a lot of hardware for a minimal advantage eh ?

And of course the SA-ARM with ONLY 32Kb of 233MHz cache (1/8th the size and about 1/3rd the speed of the cache on the P-III) and yet winds up TWICE AS FAST as the P-III.

If an SA clocked at 1/3rd the speed having memory and I/O bandwidth about 1/10th that of a PC and a cache 1/8th the size of and 1/3rd the speed can still trounce a PC - a machine like the IMAGO which has a much more compeditive bus speed should embarrass the PC.

As for Evolution, if it happens (and I hope it does), it should yield a performance somewhere between that of the current SA machines and Imago. Just people don't focus on the PC's clock rates - clock rates are only comparable for Processor intensive tasks on IDENTICAL processor architectures (and the ARM is NOT a Pentium so such a comparison is meaningless).

The last time (1987) an independant review of Acorn based hardware (ARM-2 based archimedes) and PC based hardware (the then high end IBM-PS2 Model 80 (386 @ 16MHz)) showed the Acorn to be twice as fast when only clocked at HALF the speed of the PC (suggesting (non-scientifically) that 1MHz on the ARM is the same as 4MHz on a PC !), the review was in an issue of PCW.

As for AGP hardware remember much of that is there to circumvent the problem of slow processor access to the video RAM space on the PC through the PCI bus. This (from the scroll test) does not seem a problem for even the old Acorn hardware and for something like imago (approaching 1.6 Gb per second transfer rates) should be even less of a problem - so the need for AGP is just meaningless I suggest.

[Edited by 144 at 19:28, 30/7/2000]

Similarly, a 486 DX133 is roughly as fast as a Pentium75

Benchmarking aside, In the real world use, there is no way in hell that my SA233 is anything like as fast as my relatively old PC @ 400mhz.

I find myself constantly sitting here waiting for the Harddrive to finish shifting meagre amounts of data around, or for the processor to finish it's number crunching.

Make no mistake, I am well aware of why this is so, which is why we so desparately need up to date hardware to unleash the StrongARM chip and also faster SA chips.
I believe it is the lack of faster SA chips that is holding development back at the moment.

I take your point about the scrolling test, but it's hardly indicative of real world use.

The only reason why it is 'sometimes' quicker for me to use the Acorn is that i'm more familiar with the 'some' of the software I use on it.
(The less said about the Acorns Internet abilities the better!)

As far as Evolution goes, I have spoken to CTA a couple of times lately (when buying my Blitz interface) and have been told that it is progressing nicely and should be with us 'soon'.
The sooner the better for me!

- QBasic has never been known for it's performance (unlike BBC BASIC)
- NT doesn't have DOS - it emulates it, so it's paying for a performance hit there
- NT is an inherently more complex system than RISC OS and doesn't have the advantage of being written in handcrafted ARM code (and boy are we paying for that now). Simply switching out to a DOS session does not prevent the usual NT services running in the background - some of which RISC OS isn't even capable of supporting.
- Assuming both programs accessed the screen memory directly then it is harder (less efficient?) to write directly to a video card on a PC outside of standard VGA, whereas it's easy on an architecture that is tied completely to its own custom chip. If the two versions used the various libraries available (is sprite manager on RISC OS) then the test becomes even less valid.

On the other hand we could do something similar written in C/Assembler but using 3D graphics instead. I don't think the results would be popular with the RISC OS fraternity 8)

Finally, as for AGP being pointless, old Acorn hardware doesn't have the problem because its capabilities are somewhat more limited and as for Imago, well it doesn't matter how fast the bus is if I can't even plug a 3D card in. (Go on, mention viewfinder someone - I'll be convinced when I see a GeForce card running on it that matches a PC).

Still, the comments that Evolution is coming along nicely are nice

The main performance issues with any computer system are the access to main memory, permanent storage (harddisks etc) and video system.

So based on this the board will need to contain the CPU, dedicated memory and of course RiscOS in ROM.

I am sure something could be done to give harddisk dma using a dedicated partition (or entire drive).

When it come to the video Acorns have always had performance problems with top end graphics because the cpu has always done all the work. (Apart from hardware scroll which was built in even on BBC micros). We need to take full advantage of PC graphics cards with built in engines so that the work is taken away from the cpu.

Surly its possible to write a driver even for a specific PCI graphics card (as viewfinder) and for the cards to talk to each other with minimal cpu overhead. Would it be possible for RiscOS to use a PCI card graphics card while the PC used an AGP card?

Access to other pc peripherals is a bonus even if the performance is impaired.

The overall simplicity of the concept must make this worth considering.

P.S. in the previous post, when Paulo was talking about PC board clock speeds all being at 100mhz, we now actually have the 200mhz Athlon 750 motherboards, and the upcoming 400mhz ones.

True QBASIC is not earth shatteringly fast but on a machine clock at 667MHz versus one clocked at 6MHz it should manage to beat it by more than the meagre amount it does.

You are right MS-DOS is run as a session on NT (I had most services turned off for the test). But still NT is not that inefficient in running DOS. I have run the same test (under DOS on an Cyrix 166MHz and its slower than my 6MHz ARM).

I take your point about FP (the SA sucks) and yes I would not compare the ARM running without FP against a PC with a 3D coprocessor - but my point is that once the bus bandwidth problems on Acorns are solved (Imago) and faster ARMs are used (ARM10 or SA2) people may be surprised how much the performance will improve.

As to AGP being pointless I am afraid it IS. The point is that your video ram space is limited to what's on your card. The Imago can address (I think) 512MB the bulk of which could be used for animation and video. The bandwidth (around 1.6GB/Sec) is such that 2 or 3 ARMs could share the bus with video and with the FP (such as the Vector Processor equipped ARM-10) the need for a 3D card would decrease. Remember an ARM is a general purpose processor a 3D accelerator is NOT, give me flexibility anytime.

And No I would not hold out the Viewfinder as an example. Yes it can improve performance of certain operations but is still subject to the same limitations as an AGP card on a PC (plus its hampered by the slow IOMD bus). So the chances of seeing a GeForce card running on it are slim and in any event some FP is required on the main CPU for 3D page set up (in short the GeForce doesn't do all the work - some processing is needed on the main processor - and FP on a standard ARM is weak).

[Edited by 144 at 21:25, 1/8/2000]

In fact the 667MHz P-III was the FIRST PC that I HAVE EVER FOUND to outperform the humble A-3010 on the scroll test.

Yes it is just one test and the PC will beat the ARM (by a long way) on FP.

The slowness of I/O (disk) is down to the peripheral controller chip used by the RPC, as far as I know it's still using methods that are out of date. Given a spanking new UDMA controller that would cease being a problem (its not an ARM or IOMD limitation this !).

The scrolling test I gave was mentioned for two purpose (i). Some of the Acorn hardware is good and gives excellent performance with low clock rates (ii). To get people not to assume that just because a PC is better at certain things that it is better at everything. With a little hardware upgrading the performance of the ARM might well be truely unleashed (and it may just surprise a few people too).

The cleanest approach to video would be for Windows (and the P-III/Athlon) to handle the video and honour requests from the ARM based PCI card to do video I/O. The problem is all of that is running under Windows (with some loss of performance). The bus speed would also be a bottleneck (PCI=32 bits at 33MHz current ARMs are 32 bits at 66MHz) so you lose (at least) half the ARM I/O performance. You would also need an ARM with FP to make the most of 3D cards (these still require some FP work by the CPU).

On top of that RiscOS would need to be modified so that it directed all video operations to the PC side through the PCI bus. All this takes time and money, and I respectfully suggest that getting RiscOS ready for faster 32 bit hardware and getting the Imago and Evolution machines in the market place is a better use of the available money and resources.

Seems like a very convoluted way to market RiscPCs and RISC OS. Especially when for a couple of hundred more you can buy a whole RO computer!

Maybe if you gave them away for free it would increase interest in RISC OS.
However, half the attraction to RO (in my opinion) is the way the software interacts (technically wrong but hey) with the GUI - and so you'd have to give them software with that PCI card.

Don't get me wrong (and I'm sorry to sound so negative - ROR!) I wouldn't mind seeing RO on a PCI - but I've seen nothing to convince me that it's a really good idea (in fact it feels like a step backwards).

But keep on ROR ing! :)
-See geekware thread and comments on the news...


[Edited by 94 at 15:38, 3/8/2000]