pgraf--- via Ql-Users wrote:
>> Well, I said I'd do it and apparently I did it:
> What surprises me is this: "With the original Verilog code my clone
> didn't work at all, so it's definitely not just the different chip."
> With a normal GC, the original QL-SD worked relatively well,
> compared to the SGC.
I've heard, but bot on my system. There was one driver version where
it briefly managed to display halve a directory before it turning into
scrambled screen, but that was about it.
> On some GC systems it was even fully stable hardware-wise. So if we
> assume that the electrical characteristics of the XC9572 play a
> minor role, why did it not work _at_all_?
> A significant difference between Original GC and Tetroid?
Well, the Tetroid-Card has the 30ns displacement between the adress
bus changing and ROMOE being cleared. I think it's entirely plausible
that other GoldCards might have a different timing, making the "it
works or not" a matter of pure probability.
As Tobias wrote "I have mixed experiences with GoldCards - One doesn't
work at all, the other one (at least in one specific QL) rather well
for days together with a QL-SD. But not well enough to go through the
nuisance of destroyed file system every other week or so." You might
get lucky for a while, especially if timing is very tight, but then
one day you might not.
I've also heard that for some people it only works when the driver is
loaded into RAM, which reduces the ROM accesses and once again reduced
the probability for a spurious actions.
Mind you I don't consider this a bug in QL-SD, it's more a bug in GC
and I'm trying to create a stable workaround. It'd be very interesting
how an SGC handles this.
QL-Users Mailing List