On 5 Dec 2009 at 10:41, Fred Cisin wrote:
If you let me design the race-course, I can beat a
Ferrari with a
1950's VW bus. Or, to be closer to on-topic: If you let me design
the benchmark, I can prove that a TRS80 is faster than a current Dell.
Machines were designed to perform to certain benchmarks. Anyone
remember Saxpy (the company, not the LINPACK deck)? Anyone have one
of their machines?
We used to "cook" benchmarks with clever compiler optimizations.
Landmark was extremely popular with vendors, because
it made the most
trivial enhancements seem major.
At 80's computer "faires", you used to see aisle after aisle of
vendors selling Far East-origin systems all running Version 0.99 of
the Landmark test.
Although it was CLOSE, usually acceptably so, it was
not COMPLETELY
plug-in compatible. For example, the Gavilan required some trivial
hardware mods to use it.
On most systems, this isn't an issue, but ISTR that the RESET timing
is a bit slower on the V20 than the 8088 and that trips up some
systems. AAR, if you have trouble with the V20 in your 8088 system,
the first place to look is the reset timing. There are also some
nits when an 8087 is involved, but I think the workarounds are mostly
software in nature.
--Chuck