A6 review part 2By Martin Hansen. Published: 29th Jan 2004, 20:04:30 | Permalink | Printable
Benchmarks and opinionPart 1 of the A6 review
Review Welcome to part 2 of my confession. As you may recall, I previously admitted on drobe.co.uk to buying an A6, one of the new "hybrid" computers that is set up to run RISC OS 4 as easily as Windows XP. Either operating system can be given the full screen, although RISC OS can also be run within a window of the XP screen. These "Windows RiscPC" machines have prompted heated debate within the RISC OS world, not least because of the long held view that a user has to choose; Are you using RISC OS or Windows?
I though it would be interesting to open part 2 by first looking more closely at the two modes in which RISC OS can run - either as full screen or in a window. Moving from one to the other is no big deal; simply press the two keys "Alt" (held) then "Enter" to toggle between. After some thought, I decided to dig out one of my ancient software masterpieces (he said, modestly). It's a program called Laser Bounce that I sold to BeeBug magazine in 1989, for the princely sum of 50UKP. I selected Laser Bounce because is does not multi-task on the RISC OS desktop, and does not use a mouse to drive its menu system. The code is pure BBC BASIC; simple to write and, looking over it now, highly readable.
VirtualRiscPC to virtualisation?
I have thought a few times about making it multi-task on the RISC OS desktop. But this takes time and effort and there will be nothing elegant about the resulting listing. Now, under Virtual RiscPC there is no need, for, with RISC OS running in a window on the XP start-up screen, Laser Bounce 'multitasks' and the menu, one of the old press the number of the item you want types, works fine. I have often thought how useful it would be to have an option to run BBC BASIC in a window on the RISC OS desktop so that some of these gems from the BBC Micro days could have a new lease of life. Under VirtualRiscPC we have, sort of, got this for free.
Maybe in the future, VirtualRiscPC, or some rival software, could run two or more RISC OS programs in seperate windows and multitask them in the WindowsXP desktop. This "moving of the goal posts" is, to me, typical of what Virtual RiscPCis about. As a further example, let me ask you this; Why now rewrite, as a Windows XP version, any RISC OS application? Is it not now more sensible for the major applications authors to throw their endorsement behind VirtualAcorn as the way to "convert", without doing anything, RISC OS' 'killer apps' to work under WindowsXP? In full screen mode, you can even forget that WindowsXP exists, underneath, supporting the emulation.
Laser Bounce running in WindowsXP
Part 1 of my A6 review prompted a lot of drobe.co.uk readers to cry out for benchmarks. Ever keen to please, I decided, on their behalf, to benchmark the new machine. To do this, following the advice posted, I downloaded RISCOSmark from Richard Spence. Quite reasonably, some readers were not satisfied with comments along the lines of, "It feels much the same as a Kinetic RiscPC". They were after an opinion founded upon the indisputable truth that arises from hard, cold, unemotional numbers. Compare and contrast: What should we be pitching against the A6? Well, obviously, a standard 202MHz StrongARM RiscPC. This is, after all, what the A6 is an emulation of. How well does Advantage6's "Barton-cored 2500+ CPU" newboy match up to an eight year old StrongARM? As I've got a Kinetic RiscPC handy, I've got a second comparison on offer. And just in case Advantage6 is thinking this is going to be a walk over I shall add in the figures for an Iyonix. The stakes are high because it's not just the A6 under test here, it's VirtualRiscPC itself and the whole dazzling caboodle of machines now being sold with it pre-installed. In many respects, on drobe.co.uk, the A6 has become their representative.
The table of results from running the ten different tests of Richard's program is presented below. In this table all figures are percentages and as the A6 is trying to be a base StrongARM RiscPC, I've related each machines performance to that of "being a 202MHz StrongARM". This also explains why the SA202 results are 100%.
Benchmark results - all figures are in percentages and relative to the SA-RPC base machine
Consider the top row, the one labelled 'processor'. Richard describes this as being closely related to MIPs; a measure of how fast the instructions are being processed by the (virtual) CPU. The real CPU on the A6 is running at 1.8GHz to emulate a 0.202 GHz StrongARM machine. The A6 is clocking up under half (41%) of what the StrongARM can manage. I was a little surprised by this. It didn't seem like a very promising start for the A6. Had I wasted my money?
The second row suggests where the A6 may be making up on lost ground. Under emulation, RISC OS on the A6 is accessing memory 13 times faster than on StrongARM. Now that is impressive. Round two to the A6, perhaps? Next, the speed at which a rectangle can be copied from one part of the screen to another is probed. Here, again, the A6 shines: it's almost ten times faster than StrongARM; the Iyonix managing to be six times faster again making it 58 times faster than the machine it succeeded. In the creation of a simple 16 colour icon with mask and palette, it is the Iyonix that looks embarrassed, whilst, remarkably, it's the Kinetic that is top dog at plotting thin lines and then filling in the area between them.
Confusingly, on his website, Richard stops using percentages at this point, but I'm sticking with them for the last four tests which probe the speed of access to the hard disc. Windows machines are well known for their persistent use of the hard drive and for the disc interface being fast. The figures here show that if one gives the A6 a 1 MB file to read or write to the Hard Disc under RISC OS, then it's over 52 times faster than StrongARM: It is only by reading or writing the file one byte at a time, that we can slow the A6 access time to 'only' four times faster than StrongARM. VirtualRiscPC seems wired to take advantage of what Windows XP can do well.
Benchmarks often get criticised for not being relevant to real tasks and therefore productivity in the real world. I don't agree. For example, I recently realised that the A5000 has a real Achilles heal if you ask it to write, in my case, a long string of results from a mathematical algorithm, a byte at a time to hard disc. I'd programmed the code on a RiscPC600 where the byte by byte write to hard disc took a couple of seconds. On the the A5000 it took a ridiculously long time of thirty-five seconds. These last few tests would have picked up that weakness.
Furthermore, the fast memory access figure for the A6 made me think again about my previous (in part 1) observation regarding the listing of a 450 Kilobyte BBC BASIC program. Why was it so slow when the benchmarking suggested otherwise? I retested; 6 seconds to list the program on the Kinetic, 44 seconds on the A6. Then I realised that on the A6, BASIC had fired up in a 16 million colours mode. I switched to 256 colours and it took -pause for effect- just 4 seconds.
As in any good scrap, none of the combatants have emerged without having their dignity ruffled. However, it is easy to become microscopic about the figures. In fact, I would say that it is only once something is around twice as fast or as slow, that it becomes an issue worth talking about. Keeping a sense of perspective, I didn't buy my Kinetic RiscPC because it was 25% faster than a StrongARM. I bought it because it was 2900% faster than my best machine at the time, an A5000, and it could drive modern monitors properly and was vastly more expandable.
The benchmarks confirm that the A6 is worthy of RISC OS 4. What it lacks in MIPS, it seems to pick up elsewhere. So from the speed point of view, which is what benchmarking attempts to examine in isolation, the A6 is right in claiming to be; a software emulation of a StrongARM RiscPC.
Resolution, resolution, resolution
Let me start by reminding you that the latest version of VirtualRiscPC routinely works in 16 million colour modes and, with the monitor supplied, at 1280 x 1024. What you may not have considered is that a screensave will now result in a 5MB sprite file. To match or exceed this the StrongARM would require a Viewfinder graphics card. Even with 2MB of VRAM fitted a StrongARM machine can only manage 256 colours at this resolution, and with 1MB of VRAM, or none at all, only 16 colours. Not good at all if you are into digital photography.
So long podules
At least with the StrongARM you can add on additional hardware such as Viewfinder. Users of Virtual RiscPC will find that the vast array of podules for RISC OS are not accessible to them. No 1MHz bus for that LOGO floor turtle. No teletext card. [Best point out that the WinXP side of the A6 will be able to handle PCI cards and devices that provide things like this, although you'll have to use Windows - Ed]. To be fair though, I asked for R-Comp's "UniPrint" to be included with my A6. This provides access for RISC OS to many printers through WindowsXP for which pure RISC OS drivers do not exist. Not until, that is, Martin Wurther's GemPrint project kicks in. On the A6, networking works from within RISC OS, access to the internet is not a problem either.
It has been an interesting experience to buy a Windows XP machine. I would not have done so where it not for the inclusion of VirtualRiscPC. My twelve year old daughter is wildly enthusiastic about her new computer. She gives the A6 a score of 9 out of 10, and the StrongARM RiscPC she has left behind, 1 out of 10. I asked her why, and it was down to how it looks, and the games available. Of course, she is not into programming or after a deep understanding of the inner workings of her machine. Windows seems to place several layers between the user and what's really going on which I find annoying. I, to my daughter's delight, will probably only occasionally want to use "her" A6 for demonstrating and testing software. For my day to day computing needs, I remain very happy with my Kinetic RiscPC, and feel no urge to use WindowsXP other than in a casual way. Switching to VirtualRiscPC would be a sideways move and, to me, the upgrade machine I aspire to own remains the Iyonix.
I do admire the cleverness of VirtualRiscPC on the A6. One does feel that this is a robust and professional piece of software with great strength of design behind it. All of the substantial extra free software is carefully catalogued and includes some "big name" items. The electronic User Guide too is of a high standard. The emulation of one computer within another can never be exact in all respects, but the A6 is very impressive in how close it has come.
A6 website - specifications, pricing etc.
VirtualAcorn website - VirtualRiscPC, etc.
Previous: DIY Iyonix mobo mod job to cure freezes
Next: APDL hunts for coders
DiscussionViewing threaded comments | View comments unthreaded, listed by date | Skip to the end
Please login before posting a comment. Use the form on the right to do so or create a free account.
Search the archives
Today's featured article
RiscPC breaks 1.5GHz barrier
Calm down dear, this is a case mod
23 comments, latest by flypig on 13/5/04 1:56PM. Published: 12 May 2004
ARMalayser updated to version 0.62
Cross-platform compressed AIF support and more
Discuss this. Published: 18 Oct 2007
News and media:
RISCOS Ltd •
RISC OS Open •
MW Software •
Advantage Six •
CJE Micros •
Liquid Silicon •
Chris Why's Acorn/RISC OS collection •
The Register •
The Inquirer •
Apple Insider •
BBC News •
Sky News •
Google News •