Archived from groups: comp.sys.ibm.pc.hardware.chips,comp.sys.intel (
More info?)
Nate Edel wrote:
> In comp.sys.ibm.pc.hardware.chips Rupert Pigott <roo@try-removing-this.darkboong.demon.co.uk> wrote:
>
>>Nate Edel wrote:
>>[SNIP]
>>
>>>The introduction of MMX with the P-II and earlier Celerons (not to mention
>>
>>MMX turned up with the Pentium. My friend bought a Pentium MMX 200, I
>>bought a Pentium Pro 200 (which lacked MMX, but shredded the MMX on
>>FP and unoptimised 32bit code). IIRC that happened within the same
>>month.
>
>
> Yes, I'm well aware of that with the PPro and P-MMX. Then again, the P-Pro
> underperformed on 16-bit and particularly segmented-model code, which made
Not really. I had the opportunity to benchmark them side by side. Rarely
saw much over 5% hit. I know that won't stop you parroting the wisdom
that originated with the Intel marketdroids, but hell, I benchmarked the
machines side by side and I didn't have any axe to grind either way. For
me the 30-50% improvement on FP in the binaries I ran more than made up
for the odd 5% hit. Quake liked the Pentium Pro a lot too, which was a
bonus.
> it a poor choice for gamers when games were often still for DOS, and a poor
> choice for those folks still on Win 3.1. It was great for NT, though, and a
> mixed but generally good choice for Win95.
The PPro 200 burnt the MMX200 on Quake and Carmageddon to name two games
I cared about at the time. Furthermore Windows 3.1 had ah heck all to do
with games back then, Win95 changed that of course and yes, the PPro was
more than a match for the MMX200 in Win95. YMMV, but I've met very few
people who actually had the machines side by side and compared them like
I did.
>>>the other performance tweaks), and the introduction of SSE2 with the P-III
>>>and later P6 Celerons is IMO significantl.
>>
>>ISA changes don't necessarily turn the core upside down, and I don't
>>think they did in those cases. "tweaks" are generally small changes
>>to an existing design too.
>
>
> They don't turn the core upside down, but they often are more noticeable to
> customers than core redesigns.
Erm, no, not in my experience. Sure, a good tweak in the right place can
have a big impact on specific codes, but in practice a core redesign has
a broader impact.
Compare and contrast the SPEC benchmark profiles of the Pentium II,
Pentium III and Netburst (Pentium IV) cores for example.
Cheers,
Rupert