From Newsgroup: comp.sys.cbm
Cameron Kaiser <
ckaiser@floodgap.com> wrote in message news:<40dece18$0$43386$
45beb828@newscene.com>...
References trimmed. Crossposted to comp.sys.ti (this is on topic there).
anoneds@netscape.net (Ben Yates) writes:
And yet, the TMS9995 was faster than the 9900. The Geneve was 4-5
times faster than the 99/4A, both through design of the computer (the
9995 was fully exploited whereas the 9900 on the 99/4A was multiplexed
[... excellent tech discussion snipped ...]
So, the 8-bit computer was 4-5 times faster than it's 16-bit
predecessor.
Strictly speaking, it's a 16-bit CPU with a muxed 8-bit data path, but I
know you knew that :) ... I'm just being pedantic, since my Tomy Tutor has
a TMS9995, and I have to be picky where my first computer is concerned. :)
Yes, you're right. But I guess I didn't really want to bring it up,
because there are those out there who will now join this discussion
and argue 'til they are blue in the face that a uP with an 8-bit
connection to the outside world is 8-bit even if it has every feature
plus more than a 16-bit internally and works internally the same way
as its 16-bit predecessor in the same processer family. Kind of hard
to argue either way, which is why many call the 9900 8-bit just
because it was muxed externally on a 99/4(A) to 8-bit for _most_ of
its address space.
I almost bought a Tutor once at Kmart when they were dumping them. I
was _sure_ it was TI related, just by the looks of the manual, but it
didn't list details. I'm sad now that I didn't pick one up...
I really wanted some in the TI community to do comparisons between the
Geneve and the SGCPU (another computer on a board for the TI expansion
system utilizing a fully exploited 9900). No one ever did. But I
believe I remember Michael Becker the designer of that board saying
the Geneve was still faster, esp. because of the pre-fetch.
And there was a lot of misdirection in the TI community about the
Geneve being 12mhz, when that was the clock rate and it is a
four-phase cycle, giving 3mhz, same as the 9900 on the 4A. But the 4-5
times speed difference does appear to be real, mainly due to the
pre-fetch and the on-board registers I guess. And the removed
wait-states (there are 4 between the TI console and P-Box). I still
haven't heard the results of the mod to remove the wait-states that
Thierry Nouspikel posted. That is one mod I would really like to try,
esp. the easy version that switches between 4 wait states and 0. The
version that allows choices between 0, 1, 2, 3 and 4 is a little
complicated (an understatement).
Ben
--- Synchronet 3.18b-Win32 NewsLink 1.113