Author: Johan Klockars (as12-5-4.kp.g.bonet.se)
Date: 02-01-2002 01:18
> I'm not sure that the comparison of RISC to
> CISC is that simple. RISC is usually
It is not. See my other message.
> considered faster nowadays, but the chips I'm
At http://www.spec.org/osg/cpu2000/results/cint2000.html the IBM eServer pSeries 690 Turbo (available december 2001) is currently the fastest, but after that you have the Athlon XP and the P4. Sparc, MIPS, HP-PA and Alpha are all behind the x86's.
On the FPU side, the x86 processors are slightly behind the best of the rest, though.
> familiar with (Alpha, Sparc) are all 64-bit
> chips, while x86's and 68k's are 32-bit. So
Which doesn't matter on most code. The larger pointers might hurt a bit, but it is often possible to compile for 32 bit pointers, anyway.
> the 64-bit RISC chips are faster at same
> clock speed, but, of course, it's not as
At the same clock speed, they may indeed be faster. The performance per clock tick is irrelevant, though, and the x86 chips currently run at much higher clock frequencies.
> About the constant x86 bashing. Everyone
> here always says that a 68k will beat out a
> matching generation x86 at the same clock speed.
Not me.
> Does anyone actually have data on this? I
You can find various Dhrystone MIPS numbers on the net. The figures vary a _lot_, though (I've seen nearly 4 times difference for the same CPU in the same list...), so it's difficult to make a real comparison that way.
The x86, especially, is very dependent on having a good compiler, and it obviously is not fair to force it to run 16 bit code.
Anyway, the numbers I've seen seem to place the '386/'030, '486/'040 and Pentium/'060 quite close to eachother (for Dhrystone) at similar clock frequencies.
Of course, both the '486 and the Pentium quickly reached much higher clock frequencies than the '040 and '060, and there was no further generation of the 68k line.
The 233 MIPS at 162 MHz given by Motorola for the ColdFire 5407 is very close to numbers I've seen for a 166 MHz PentiumMMX.
|
