RISCOSmark is a program to give a benchmark for various parts of your computer's hardware and I/O, compared to a 202MHz StrongARM RiscPC running RISC OS 4.02. It will run on RISC OS 3.1 up to 5.03 and on any hardware/emulator capable of running those OS's. For details see the doc file in the zip archive.
Download version 1.01 from: http://www.richardspencer.freeuk.com/riscosmark/romark101.zip
There have been problems with Select/ViewFinder; you may need to disable SpecialFX, I am told.
The following table is based upon other peoples postings of benchmarks to usenet or in emails. Some of the percentages have been altered a little to match the new baselines of 1.01 (where 100%=SA202 RiscPC running RISC OS 4.02), and data rates are shown for harddisc tests rather than percentages. All tests were done in SVGA 800x600, 256 colour screen modes apart from the Iyonix (which was 1280x1024 16m), amd the A4 which was in VGA16. The machines tested were as follows:
Iyonix: [UPDATED] RISC OS 5.03 with DMA enabled, recent (early 2004) firmware
Omega: RISC OS 4.03, firmware ~Oct 2003
SA287: RiscPC SA 287MHz, RISC OS 4.02
SA287 VF: RISC OS Select 4.33, 8MB ViewFinder, Blitz IDE
K233 VF2: RiscPC Kinetic233 RISC OS 4.03 with 32MB ViewFinder2
P4-2.4G: Pentium 4 desktop 2.4GHz running VirtualA5000 RISC OS 3.11
Alpha: MicroDigital's first laptop running VirtualRiscPC-SE emulator software
A7000+: 48MHz ARM7500FE, RISC OS Select 4.33 (shows how bandwidth is eaten up with no VRAM)
A4: Still the only production RISC OS portable, running RISC OS 3.1, but can't really compare graphics
ArmSI version 4.00 (NOT Iyonix compatible) : http://www.richardspencer.freeuk.com/riscosmark/armsi.zip
Dhrystone 2.1 (Castle's 32 bit compiler version) : http://www.richardspencer.freeuk.com/riscosmark/dhry_cc32.zip
To be more 'real-world' for the graphics tests, maybe the DrawFile module could be used to render a complex filled and outline file, a JPEG and a deep sprite, in a true colour mode. Also perhaps test the Font Manager. If there's an interest then I'd develop the program, but at the moment its appeal is that it's compact and only takes 10 seconds to run.
Take a look at http://www.riscoshardware.tk/ for a more in-depth look at real-world benchmarks specifically comparing SA vs Kinetic vs 7500FE. Steffen Huber has done an extensive timed Omega/Iyonix/RiscPC comparison which is at http://www.huber-net.de/iyonixvsomega/bench_e.htm
Tests on environments without a hardware timer will give better results than they should. Therefore VirtualRPC-SE benchmarks can only be accurately determined with a stopwatch. The Iyonix would normally be run in a 16m mode as this means full 32bit words per pixel, so 256 colours does not show it at its best. However as I have neither an Iyonix or VRPC I leave it to others to devise and carry out benchmarking tests.
See the doc file in the download for an email address.