emmo@moncam.co.uk (Dave Emmerson) (08/03/89)
>WORM drives beat this problem. Maybe a proper loading mechanism >will come out with machines such as the ATW, where one processor >loads, while the other is already running. While the other processor is already running WHAT? Memory speeds are still increasing, but I haven't seen any clairvoyant RAMs advertised. The topic of bolting a 68020 onto an ST looks intriguing. I have a homebrew PCB which does this on a Sinclair QL. The cache can be disabled with both software and hardware on the 68020, my default system only enables the cache for QDOS calls (Read : TOS), this still allows a significant increase in execution speed without any problems with self-modifying programmes (of which the QL has it share, mostly only on loading, as part of the copy protect stuff. The QL is multi- tasking as standard, so once you've loaded all the self-modifying progs, you can enable the cache everywhere if you want.) I can't see why this can't be done on the ST, it's a very simple hardware job, but I'm not an ST expert - I do know a bit abot the 68xxx. The only remaining problem is software delay loops, (if there are any) and this MUST have already been overcome by those companies offering faster 68000 add-ons anyway. Sadly, the ST's MMU would make 32 bit wide RAM virtually impossible, but you could move the ROM (TOS) onto the full width data bus, with a significant increase in speed. No, I'm not about to do it, just food for thought... Dave E.