12 Nov
2006
12 Nov
'06
11:31 a.m.
Chuck Guzis wrote:
I believe that benchmarks tend to have a first-cousin
relationship to
unvarnished mendacity, but given the combination of faster I/O and
processing speed, I think an 8 MHz 5170 comes very close to being
almost 5 times as fast as a 4.77 MHz 5160. Perhaps one of the list
members can verify this. I've got XT and AT "clones", but that might
not give an accurate reflection of performance on IBM hardware.
If you're talking about pure CPU processing with it's interface to RAM,
it is not 5 times as fast, and no benchmark will properly reflect that.
The only thing about an 80286 that truly is 5 times as fast is the
MUL and DIV instructions: On 8088, MUL reg16,reg16 takes at least 118
cycles while on 80286 it's exactly 21. If a benchmark makes heavy use
of MUL/IMUL/DIV/IDIV then yes, it will report artificially inflated times.
The only thing that could substantiate "5 times faster" between 4.77MHz
8088 and 6MHz 80286 would be a faster I/O subsystem, like a hard drive
with a 2:1 or 1:1 interleave. But that is outside of the core claim.
I have a 5160 sitting next to me and can easily drag out my PC/XT 286
(6MHz 80286) if someone really wants me to run benchmarks for
comparison. But some old MIPS.COM numbers show that the actual
improvement across general instructions, integer instructions,
memory-to-memory operations, register-to-register operations, and
register-to-memory operations is about 3 times faster... not 5.
--
Jim Leonard (trixter at oldskool.org) http://www.oldskool.org/
Help our electronic games project: http://www.mobygames.com/
Or check out some trippy MindCandy at http://www.mindcandydvd.com/
A child borne of the home computer wars: http://trixter.wordpress.com/