Originally Posted by alexh
If you do thermal tests, please have a go with a HxC Stand Alone (SD) Floppy emulator unit fitted.
Sorry, that's not going to happen. I will do the thermal tests in order to have something on file for warranty cases where people make funny claims. In any case, I will not do tests with any foreign hardware, as I have no influence on their quality control, component sources or component tracking (I doubt that they have implemented any QA of that level).
If the user decides to combine certain pieces of hardware, my responsibility ends whenever I do not make that hardware. Only exception is the Amiga itself. I'm not liable for anyone else's hardware, so I don't spend time on testing. I'll do my thing, and I'll keep some safety margin. It's up to common sense of the user not to exceed those safety margins.
As always at the end of a working day: Results!
I noticed that the memory chips got pretty warm (over 40 deg. Celsius) and found that I have implemented the "automotive" refresh timing, which makes a lot more refresh-cycles than the "commercial" refresh requirements. I have now reduced to commercial timing, and the memory temperature came down to 37 degrees. In theory, memory performance should have increased, but it seems that the effect is below measurable threshold when using the bustest tool.
Like I wrote this afternoon, memory is now added to the system memory pool automatically on startup using Z3 autoconfig.
I also implemented a non-MMU Maprom function for $f8 kickstart (re-mapping to $40f80000) and it seems that performance is considerably higher than using an MMU. Could anyone verify this? Result looks really pretty, almost 6 MIPS out of this poor little 25MHz EC030!