AMD reinvents x86

[DragonMaster]

http://www.infoworld.com/article/07/02/07/07OPcurve_1.html?source=NLC-CURVE&cgd=2007-02-08

AMD’s next-generation processor line, code-named Torrenza, has gone from a block diagram to living, breathing silicon. The first incarnation of AMD’s redesigned x86 CPU is Barcelona, that which your non-co-readers will call quad-core Opteron. Barcelona is genius, a genuinely new CPU that frees itself entirely of the millstone of the Pentium legacy. It’ll do the same for you.

[Kougar]

That has to be the most positive spin I've heard to date on what's been considered to be "K8L", or Barcelona. 80% improvement over the Opteron in FP performance? If taken at face value, then Intel's Core will become another Prescott in comparison. :tongue:

All in all, it does seem AMD has the potential. I think this article glossed over a few things though... Does Barcelona have anything else to supplement this "SSE 128" instruction in one clock ability? Core can do three separate "SSE 64"'s in one clock by comparison. Secondly a good deal of non-server code is not in 128bits from my understanding, so this would underutilize desktop varients of Barcelona if this was the case.

I'm not attacking Barcelona here, but I think this article needs a large grain of salt. He's picked and choosen to many random things at random... and as far as FP goes AMD themselves stated it was up to 40% so I have no clue where that guy got "80%" from.

I can't wait to see ArsTechnica's article on the breakdown of the uArch, and how it compares to Core though. Their Core uArch article was very informative. :)

[Xero (1)ne]

Who care's if its 10% or 1,000%(that would be nice though lol) the fact still remains that AMD is kicking ass and taking names. And as long as they keep up the inovations we will see great things....

L3 CACHE!!!!!!!:devildance:
:ole::dancingparty:

[Kougar]

No one has proven that, Xero. So far it's all talk... not even "official" AMD benchmarks have been given to the public yet. At least as far as I would know anyway.

The more I comapre this, the more this is either total FUD, or completely true. Core can sustain a total throughput of three 64-bit integer operations per cycle, this is known. Also since "vector math unit" sounds catchy, Core has several vector units as well. ;)


Scalar: single-precision (32-bit), double-precision (64-bit)
Vector: 4x single-precision, 2x double precision