🚀 go-pugleaf

In <Bv5EyM.LBD@nntp-sc.Intel.COM> wwitt@adara.intel.com writes:

> Magazines like PC Magazine and Byte have done tons of benchmark comparisons
> of i386DX and i486DX (or i486SX) machines. On the average, I believe the
> i486DX machine were about 2x faster than the i386 machines at the same clock
> frequency. So, given this little bit of information, I find it amazing that
> the Cx486 based Zeos machine is also 2x faster than an i386 since the Cx486
> core is inherently slower than the core of the i486. Then again, perhaps the

I have done a benchmark comparison of the Intel 386DX, Intel RapidCAD, C&T
38600DX and the Cyrix 486DLC, the results of which I am going to post here
in a few days. I used standard benchmarks (Dhrystone, Whetstone, Linpack)
and two project makes (using Microsoft's and Borland's compilers/assemblers/
linkers) to add a bit of real life applications. For the integer benchmarks,
the Cyrix 486DLC (with internal cache enabled) came out the fastest, with an
average performance gain of 35% over the Intel 386DX at the same clock speed.
Maximum speedup for the 486DLC was in software fp arithmetic, with ~80% (this
is due mostly to the HW multiplier). For the floating point benchmarks, the 
Intel RapidCAD came out on top with 60%-90% more performance than the fastest 
386/387 combination. The Cyrix 486DLC + the Cyrix coprocessor were runner-up 
here, providing up to 30% more fp power than the Intel 386DX with the same 
coprocessor. I measured all this on a system with 128 kB of cache, at 33 and 
40 MHz.

Norbert
--------------------------------------------------------------------------------
Norbert Juffa         email: S_JUFFA@IRAVCL.IRA.UKA.DE        Live and let live!