A mixed bag of results

Post date: 29-Jan-2010 17:09:08

Good news and bad news from yesterday's session with 3 beebs, two and a half accelerator cards and some updates to CPLD and ROM.

good news: Ed's unreliable beeb is rock-solid when running entirely in accelerated memory.

bad news: It's become so unreliable that getting booted into an accelerated state is a matter of luck.

good news: We were able to load and run Elite at 4MHz, and the frame rate is visibly better.

bad news: It's entirely academic because the game runs too fast (not just smooth.)

more bad news: We couldn't load the game at 8MHz: the disk reads seem to return partial sectors.

good news: Our loaner beeb is in fact able to run at 8MHz - Ed had mis-identified the pins and was trying to run at 16MHz.

bad news: We're not doing as well as we thought at arbitrary frequencies: the accelerator is fine but there are rogue writes to the host RAM, clearly visible on the screen but we'd missed them because we run over a serial port most of the time.

good news: We're stable at 12.5MHz, running from a 50MHz crystal oscillator.

The acceleration and memory mapping is working well enough that we updated the ROM to populate and switch in the overlay at boot time: it takes no appreciable time to copy 64K. We haven't yet made high-speed clocking the default at boot time..

Overall, we can run at about 6 times the speed of an original beeb, we have the equivalent of sideways RAM and a patchable copy of the OS in RAM, and we have a 65816 CPU with at least 64k memory free for experiments. That's good news.