Interpretation of Performance Figures
=====================================
Notes on performance figures
----------------------------
These are based on the Norton SI benchmark (1 = speed
of original PC), and the SPECint benchmark.
A given CPU benchmark in both benchmark regimes was used
to determine a factor relating the two benchmarks.
This isn't very scientific but I think the results ring
true to some extent.

Intel
-----
(All figures independent of MHz, percentages are increase on previous family member (386sx excluded))

The leap from 8086 to 80286 produced an increase in
performance of 400%.
Presumably this increase was caused more being done per clock cycle:
increase due to more die area being available, and more parallelisation.

The 386dx was a minor improvement again (40%), probably due to the 32 bit datapath.
The 386sx was slow probably due to the 32 bit to 16 bit converter at the chip to
motherboard boundary.

The 486 was supposed to execute all instructions in a single cycle, a bit likea RISC,
however there was an on chip cache,and this resulted in a 100% increase in performance

The Pentium is marginally faster probably due to the 64 bit datapath. (60%)

The Pentium Pro introduced speculative processing. (25%)

I have no data on the Pentium III.

RISC
----
I don't have enough knowledge to comment
Pretty similar figures across the board
big jump from Digital 21164 to 21264.
Per MHz, HP PA-RISC is incredible.

presumably it is doing twice as much per clock cycle as
the Intel and most other RISCs.

What is this architecture!!! Shall we use it!


------------------------------------------------------------------------
eGroup home: http://www.eGroups.com/group/f-cpu
http://www.eGroups.com - Simplifying group communications

