Quad-Core AMD Barcelona Interview

[Nowhere]

Again, I'm perfectly aware of that...
The point is: from recent generations of CPUs (K7, K8, Netburst, C2D) it seems that it is accepted knowledge that if CPU has more pipeline stages it can clock much higher (at given manufacturing process), and vice versa.
So I was just wandering - if K6 had half the pipeline stages of P2, and it still worked at similar frequencies...couldn't that mean that AMD manufacturing technology was actually quite good at the time?

[Technoid]

then again its all up to what you compare to, the K6 was designed to kill Pentium MMX, wich it fairly did, that it could be updated to compete against the p2 shows how good the arcitecture was.

Netburst= Intel showing how faar you could drive the Cyrix way of making cpu's if you only had the layout & fabbing ability (The Cyrix way of making a cpu: small cahce, high FSB, High core clock, lauaghable FPU )

[Hulk]

Again, I'm perfectly aware of that...
The point is: from recent generations of CPUs (K7, K8, Netburst, C2D) it seems that it is accepted knowledge that if CPU has more pipeline stages it can clock much higher (at given manufacturing process), and vice versa.
So I was just wandering - if K6 had half the pipeline stages of P2, and it still worked at similar frequencies...couldn't that mean that AMD manufacturing technology was actually quite good at the time?

Okay I see what you're saying now. AMD got K6-2 to 550MHz on the 25 micron process. Intel got the PIII to 600MHz on the 25 micron process. Given the fact that AMD got pretty close to the PIII clock with a shorter pipeline does imply that towards the end of the K6 they got it worked out pretty well.

But you also have to remember that K6-2's over 500MHz were very hard to find. These were more cherry picked, overclocked factory CPU's rather than something AMD could get out in numbers.

Still even given those mitigating factors I agree with you that AMD eventually did pretty well with the K6-2 process technology.

You also have to keep in mind that AMD released the 550MHz part in February of 2000 while Intel released the PIII 600 in August of 1999. It's not that AMD can't do it. It's just that it seems to take them quite a while to make it happen.

[Jerry Jones]

Cyrix will rise from the ashes and get the brass ring... not.



Jerry Jones

http://www.jonesgroup.net

[Nowhere]

"Cyrix" is still around in some way.

OLPC (100$ laptop) uses CPU design originating from Cyrix.


BTW Hulk, wasn't the poor availability of fastest K6's more because of need to have the capacity to produce Athlon more than anything else?

[Technoid]

"Cyrix" is still around in some way.

OLPC (100$ laptop) uses CPU design originating from Cyrix.


BTW Hulk, wasn't the poor availability of fastest K6's more because of need to have the capacity to produce Athlon more than anything else?

Well, to be more exact Cyrix was bought by Via and their cpu's are alive and kicking

[Nowhere]

Not exactly...Via uses former Winchip designs that were branded for a time by Cyrix name (better name recognition?)

But AMD bought Geode line and is still producing them after modification.


(I made the mistake in my post before edit - I also thought that Via produces Cyrix derived desings, but checked and...)

[Hulk]

"Cyrix" is still around in some way.

OLPC (100$ laptop) uses CPU design originating from Cyrix.


BTW Hulk, wasn't the poor availability of fastest K6's more because of need to have the capacity to produce Athlon more than anything else?

Could be. But the bottom line is still that they couldn't ramp up the speeds fast enough to make a dent in the market. Yields has always been a problem for AMD when competing against Intel because they don't have fabs like Intel.

In the beginning of Netburst it wasn't even a big mistake for Intel. With the huge clockspeed advantage Intel had they remained competitive with Athlon for quite a while in terms of overall performance. It's when Netburst pooped out thermally around 3.8GHz and Athlon clocks kept rising that Intel really started to feel the heat (pun intended).

Go back and look at some benches for K6-2 at anything that involved floating point performance. It was an absolute dog. That coupled with the fact that it was always a little slower than the Pentium III make it unable to really win the market like the Athlon did. Of course this is just my opinion.