please, if you can get ARMSI to function under this set up, include memory transfer results in your benchmarks. i am fed up with seeing 'the desktop is noticeably quicker', or 'redraws the artworks apple x times faster' in 'benchmark' results. a while back i lashed out and bought a risc pc so i could move development of some software from an a4x0 machine - reached an impass regarding bus speed and video data volume. i was aware that the risc pc has a higher clocked bus, and would solve my problem. (this project has stalled at the mo. holidays, you know there may even be others out there still using large software/data files on RISC OS systems, and probably want to hear about improvements obtainable from both software emulation or new hardware solutions. (essentially interested to see if VRPC slows memory access to RPC like timings for interrupt/timer use). would be great to see iyonix and omega figures too. multiplying the bus speed by bus width does not often give an accurate idea of how quick a computer is going to be able to shuffle data in memory, and some of us old fogies still use mips and cpu <-> memory MB/sec values to compare computer systems. (after all, look at the pc world where a 2.2GHz processor with a slower bus from amd can give intel's wonder chip at 3GHz such a fright - one system is great for games, the other is good for multi-media). since the RISC OS market has become so diverse now, are we going to see if the situation is the same for us ? probably not if the half baked benchmarking continues.
it's all very well saying things happen quicker on screen, we want to also know what is happening behind the scenes too.
(oh, if this has been covered in the printed media, i've missed it - please direct me - and sorry for the rant !)