Author: Johan Klockars (as12-5-4.kp.g.bonet.se)
Date: 01-11-2002 00:53
> 30 MIPS for the fastest 486? That's about a quarter of the 68060 :-)
Considering that the 68040 was the Motorola equivalent of the 486 and the 68060 that of the Pentium, this isn't very surprising (but your numbers are suspect, see below).
At http://amigator.free.fr/Atrucs/proces.htm the 40 MHz '040 is listed at 33.3 MIPS (and the 66 MHz '060 at 83.3).
> As you can see, Motorola CPUs have always had about twice the speed of Intels,
They have not.
> but can anybody tell me why exactly this is?
Simple. It is not true.
> Also the 68000 had half of the transistors of the 286 and was running at about half of its speed
>
> but had the same performance.
Dream on. According to BYTE (may 1993), the 286 had a MIPS rating of 1.2-2.66 (6-12.5 MHz). That is about comparable with the 68020 (1.83 at 14 MHz according to the site above).
Also according to BYTE (same issue), the 486 at 66 MHz rated 54 MIPS and the Pentium 112 MIPS at 60 MHz.
Note that MIPS ratings in almost all (reputable) cases are not an actual number of instructions executed. That number will obviously depend a _lot_ on what you are trying to do. Rather, it is a factor that says how much faster the processor runs the Dhrystone benchmark compared to an ancient VAX minicomputer (11/780, IIRC). That old machine is the generic 1 (VAX) MIPS computer.
Naturally, the number you get will also depend a lot on which compiler you used to build the Dhrystone benchmark.
> And is it true that even modern (?) Pentium4 can only access 8 registers due to being compatible with 386s?
There are of course dozens of actual registers (renaming), but since the ISA (instruction set architecture) is the same, there obviously is still only the same set of integer registers. On top of that you have the FPU/MMX and SSE registers, though.
Later this year, AMD will release the Hammer, which has extended the register set to 16 integer register (64 bit at that). This mode will of course be incompatible with old x86 code, though.
> I can't quite understand why the Intel CPUs are such far behind, Intel can't profit from such a monopoly as Microsoft can.
Uhm. Except for one or two seriously expensive workstation/server processors (IBM's Power4 and some HP-PA, IIRC), Intel is at the very top of the performance league these days.
It didn't use to be that way, but no one else has the kind of money to throw at development and manufacturing that Intel does.
A quick look on the net turned up VAX MIPS ratings of 3000-5000 for reasonably current P4/Athlon machines, by the way.
> But perhaps I am totally wrong? (Let's see, I get hardware-informatics next semester...)
You are.
|
