🚀 go-pugleaf

In article <30045@castle.ed.ac.uk> ercm20@festival.ed.ac.uk (Sam Wilson) writes:
>I just got a flyer from Intel full of nice (paper) graphics advertising
>their new iCOMP(tm) index (they don't say benchmark!) and giving numbers
>for all(?) the Intel 386 and 486 processors.  You have to send off for
>more info if you want to know how they actually calculate the figures
>and there is no comparison with other manufacturers' products (but if
>Intel push this I expect there will be soon!).

The brochure can be obtained by calling Intel Literature, 1-800-395-7009 ext
12, Monday through Friday 7am-5PM central time. Ask for information on the
iCOMP index; the literature numbers are C5B32 and 76B4.

> They make the point that
>"MHz != performance" but my guess is that these are simply based on
>calculated processor instruction issue rates and are also not simply
>related to the delivered performance of any real system. 

From the brochures distributed in the iCOMP index press kit:
(The brochure is 26 pages long and discusses how the iCOMP index came about,
it's objectives and its technological foundations)

"[The icomp index] is not a system benchmark, but an index of relative
processor performance intended to help non-technical end users decide which
Intel microprocessor best meets their desktop computing needs."

"...[iCOMP index] is intended to suppliment, not replace benchmarks. ...it is a
tool for making comparisons between different Intel processors. This does not
mean that all systems with [the same processor] perform the same--differences
in system design and configuration will effect performance considerably. ...
However, given comparable configuration and design, the one with the higher
iCOMP rating will have more power..."

"The iCOMP index is a weighted average of several measures of processor
performance. "

The area of performance considered are: 16-bit integer, 16-bit floating point,
16-bit graphics, 16-bit video, 32-bit integer, 32-bit floating point, 32-bit
graphics and 32-bit video. The benchmarks currently used to measure these
include: SPEC CINT92, SPEC CFP92, Whetstone and Ziff Davis 7.0.

"...as benchmarks become available and the typical application mix changes,
the iCOMP index weightings will [be adjusted]."

>			The iCOMP(tm) Index
>			-------------------
>
>i486(tm)DX 2-66		297
>i486(tm)DX 50		249
>i486(tm)DX 2-50		231
>i486(tm)DX 33		166	+ OverDrive(tm)	297
>i486(tm)SX 33		136	+ OverDrive(tm)	297
>i486(tm)DX 25[*]	122	+ OverDrive(tm)	231	[= DX 2-50]
>i486(tm)SX 25		100	+ OverDrive(tm)	231
>i486(tm)SX 20		 78	+ OverDrive(tm)	166	[= DX 33]
>i386(tm)DX 33		 68
>i386(tm)DX 25		 49
>i386(tm)SL 25		 41
>i386(tm)SX 25		 39
>i386(tm)SX 20		 32
>
>[*] The Intel486(tm)SL CPU in a mobile configuration performs the same
>    as the i486(tm)DX CPU in a mobile configuration. 
>
>	Intel COmparative Microprocessor Performance
>	--------------------------------------------
>
>The iCOMP(tm) index indicates relative CPU performance.  As the 25MHz
>Intel486(tm)SX is the current entry level member of the Intel family we
>have assigned this an iCOMP(tm) rating of 100. 
>
>iCOMP, intel inside, i486, Intel486, i386 and OverDrive are trademarks of
>Intel Corporation.

Again, the full documentation is close to 26 pages. For details, contact the 
Intel Literature people at the number given above.

Jeff

Jeff Reilly			| "There is something fascinating about
Intel Corporation		| science. One gets such wholesale returns
jwreilly@mipos2.intel.com	| of conjecture out of such a trifling
(408) 765 - 5909		| investment of fact" - M. Twain