so Jobs gets screwed by IBM over game consoles, thus Apple-Intel ?

[Nate Edel]

?
As long as we're being pedantic, I don't believe the K6-III (Chomper-XT?)
ever went to 550MHz. IIRC it topped out at 450MHz (mine's a 400). The
K6-II topped out at 500MHz or perhaps 533MHz, IIRC.

Damn! That's a long time ago! ;-)

My impression was that the > 500 mhz chip was a K6-II+, and thus really a
K6-III (because of the on-board L2 cache) ... but unlike with the Intel
stuff, I was not going from memory and was just taking the consensus of a
couple of sites on Google... see frex:
http://www.tomshardware.com/cpu/20001106/amd-01.html and these guys (
http://www.cpushack.net/K6xID.html ) seem to indicate that AMD switch to
using P-Ratings part way through, perhaps for added confusion.

 

[nobody]

As long as we're being pedantic, I don't believe the K6-III (Chomper-XT?)
ever went to 550MHz. IIRC it topped out at 450MHz (mine's a 400). The
K6-II topped out at 500MHz or perhaps 533MHz, IIRC.

Damn! That's a long time ago! ;-)

I still have an old K6-2+ 500 up and running Win2k at 600 rock-solid.

 

[Nate Edel]

I still have an old K6-2+ 500 up and running Win2k at 600 rock-solid.

Ah, for the days when a 100mhz overclock actually meant something. Did those
have an unlocked multiplier, or are you running it at 120mhz FSB?

 

[nobody]

Ah, for the days when a 100mhz overclock actually meant something. Did those
have an unlocked multiplier, or are you running it at 120mhz FSB?

K6-2 and above interpreted 2x jumper setting as 6x. 600 MHz is more
than enough to run kazaa k++, emule, and browse some iffy web sites I
wouldn't risk my Opty to go to.

 

[keith]

My impression was that the > 500 mhz chip was a K6-II+, and thus really a
K6-III (because of the on-board L2 cache)

Didn't the K6-2+ have half the K6-III's L2 (128K vs. 256K)? It really was
inbetween.

... but unlike with the Intel
stuff, I was not going from memory and was just taking the consensus of a
couple of sites on Google... see frex:

I was (and am). ;-)

http://www.tomshardware.com/cpu/20001106/amd-01.html and these guys (
http://www.cpushack.net/K6xID.html ) seem to indicate that AMD switch to
using P-Ratings part way through, perhaps for added confusion.

Too long ago... I thought AMD went *away* from the P-rating with Chomper.

 

[keith]

I still have an old K6-2+ 500 up and running Win2k at 600 rock-solid.

My Windows (2K) machine is still a 400MHz K6-III on a Tyan 1598C2 (2MB
L3). I'm thinking about upgrading it, but there are so many things to do...

 

[Yousuf Khan]

I assure you, that's not how it happened. AMD was able to match the
price, but, Microsoft, gave Intel a much larger margin, because MSFT
knew that Intel had the manufacturing capacity to guarantee delivery of
CPUs for Xbox, while AMD (unless they dedicated their fac capacity
almost totally to Xbox) did not.

AMD was going to give them a K7-based Duron, with like a 64K L2 cache.
Those things were in the 80 sq.mm size range even in the 180-nm days.
AMD could stamp those out by the thousands per day without breaking a
sweat. And since they were being produced at the 700Mhz speed range,
there wasn't likely going to be a lot of defects.

Intel gave them a P3-based Celeron in the 700Mhz range instead. Neither
the Celerons nor Durons take much of a chunk out of either
manufacturer's capacity.

In fact, MSFT never had any real
intention of going with AMD -- they just leaked the rumours that they
were to scare Intel into giving them a lower price.

Wow good tactic, that was really playing it close to the vest. I mean
they even went so far as to have Nvidia design the chipset around AMD's
Hypertransport specifications, and even work up several preproduction
prototypes with the AMD/Nvidia combination so even Nvidia must've
thought it was going to AMD. Then when it went to Intel Nvidia had to
spend another couple of months redesigning the chipset to interface with
the Intel chip instead.

When Nvidia introduced its Nforce chipsets, which were direct
descendents of the Xbox chipset, the first versions were only meant for
AMD processors. It's as if they were always meant to be used with AMD
processors.

Same thing with
GigaPixel versus nVidia. Notice that *both* Intel and nVidia have
walked away from Xbox 360? MSFT didn't exactly make a lot of friends
on that transaction.

MSFT didn't exactly sell a hell of a lot of the original Xboxes either.
It's price discounting models got screwed. It was pricing those Xboxes
like as if they were selling millions more than actually got sold by the
end.

Maybe MSFT should've stuck to the original design specs?

Yousuf Khan

 

[Yousuf Khan]

2.7 with liquid cooling. And the G5 came out at 1.6-2.0 and the Opteron
at 1.4-1.8.

A pair of 2.6Ghz Opteron 252s will set you back over $1700 at newegg,
so at least for this generation Apple is better off on G5 than AMD.

Why would compare the G5 to the Opterons? They should be compared to
Athlon 64. The price points are completely different.

I think making IBM the bad guy out of this is a mistake, and I don't
know you bad-mouth them so much since AMD is also relying on IBM's
technology for their products (I don't think it's any accident that
Opteron and G5 clockspeeds have been so closely matched these past 2
years).

Opteron and G5 shouldn't have been tracking each other at all. The
Opteron was built with a 12-stage pipeline, while the G5 has a 24-stage
pipeline. The G5 should've been clocking up like a TARDIS.

Yousuf Khan

 

[Yousuf Khan]

Opteron does not use liquid cooling. AMD basically bought fab capacity
from IBM, since they can't fund the purchase of a second fab by
themselves.

Where are you making this stuff up from? AMD is not manufacturing any of
its chips in IBM's fabs. AMD and IBM cooperated together at IBM's
_research_ facilities. There was talk that IBM might fab some chips for
AMD, but that never came to fruition. They came up with various research
breakthroughs, some of which were IBM's (SOI), some of AMD's
(Dual-Stress Liner Strained Silicon).

AMD is using IBM's *fabrication process* (which *is* state of the art),
not their CPU design people.

Not entirely, AMD has a few things like its APM manufacturing process
that IBM doesn't have. This allows AMD to adjust properties of its
fabrication process on the fly to achieve better yields. Without this
process making AMD's processors might be a very low yield affair. AMD
has licensed this to Chartered Semiconductor in preparation to allow it
to make additional AMD64 processors for it, but it hasn't licensed this
to IBM, and therefore IBM won't be fabbing AMD64 processors.

Well actually it *is* a big deal technically. Steve Jobs gave a great
demo of a quick recompile, but Photoshop is clearly a very assembly
language and endian sensitive application. And I suspect a lot of the
media apps on the Mac are in a similar situation. The joke
applications will just be a recompile, but not the serious stuff.

Photoshop is already running on Windows, presumably the source code for
the two versions can be mixed in the new Macintels.

Yousuf Khan

 

[Yousuf Khan]

Ahahahaah! No, I mean *real* benchmarks by independent outsiders. You
know, like Anand's, Tom's, Tech-fest, HARD-OCP, Sharky, FiringSquad,
3D-Mark, SPEC-CPU, SPEC-GL. Oh I forgot, nobody ever benchmarks an
Apple do they? ... Oh wait! Here's one:

http://www.barefeats.com/macvpc.html

When did Tom's become a "real" benchmarks site?

Yousuf Khan

 

[Nate Edel]

Didn't the K6-2+ have half the K6-III's L2 (128K vs. 256K)? It really was
inbetween.

That's what the pages say. I never saw a K6-II+ at the time and am just
familiar with it from various articles... my impression from what I read was
that it was the same updated core.

I'm not sure if I read it or just assumed it but it seems like the sort of
thing where it may well have been the same die, just a way to sell some with
a flaw on one of the cache banks.

I was (and am). ;-)

Too long ago... I thought AMD went *away* from the P-rating with Chomper.

Much too long ago, and super-7 was already enough on the way out at the time
that I didn't follow it very much. I loved the K6-2/300 I had, even if the
VIA-based board I had was kinda flakey, but by the time the -III/II+ were
selling much I was already in the "wait for K7" mode.

 

[Kristoffer Lawson]

Opteron and G5 shouldn't have been tracking each other at all. The
Opteron was built with a 12-stage pipeline, while the G5 has a 24-stage
pipeline. The G5 should've been clocking up like a TARDIS.

*Tries to imagine how a TARDIS clocks up -- is that fast or slow? ;-) *

 

[Yousuf Khan]

*Tries to imagine how a TARDIS clocks up -- is that fast or slow? ;-) *

Well a TARDIS can do anything it likes with its clock, fast or slow,
backwards or forwards. I assume a G5's clock system would only be one
subset o the TARDIS's clock system.

Yousuf Khan

 

[keith]

Where are you making this stuff up from? AMD is not manufacturing any of
its chips in IBM's fabs. AMD and IBM cooperated together at IBM's
_research_ facilities.

^^^^^^^^^^^
East Fishkill is *not* a research facility. Other than that, I won't
comment.

There was talk that IBM might fab some chips for
AMD, but that never came to fruition. They came up with various research
breakthroughs, some of which were IBM's (SOI), some of AMD's
(Dual-Stress Liner Strained Silicon).

Aren't press-releases wunnerful?

Not entirely, AMD has a few things like its APM manufacturing process
that IBM doesn't have. This allows AMD to adjust properties of its
fabrication process on the fly to achieve better yields. Without this
process making AMD's processors might be a very low yield affair. AMD
has licensed this to Chartered Semiconductor in preparation to allow it
to make additional AMD64 processors for it, but it hasn't licensed this
to IBM, and therefore IBM won't be fabbing AMD64 processors.

You think this is novel? Processes are tweaked on the fly to produce
what's needed all the time.

Photoshop is already running on Windows, presumably the source code for
the two versions can be mixed in the new Macintels.

What I wonder is what's going to stio Applex86 applications from running
on Winwhatever (or Linux). Without applications...

 

[keith]

Why would compare the G5 to the Opterons? They should be compared to
Athlon 64. The price points are completely different.

You have the prices of the G5 (aka 970)? The Athlon64 can't do SMP,
right? ...the others do. I think the comparison between the 970 and
Opteron are fair.

 

[keith]

That's what the pages say. I never saw a K6-II+ at the time and am just
familiar with it from various articles... my impression from what I read was
that it was the same updated core.

I think you're right (IIRC codenamed the "Chomper-XT"). Though I believe
the K6-2 was also a CXT core. The CT had to do with cache replacement,
allocation, and cast-out algorithms, or some such. It *has* beeeen a long
time, and I don't write this stuff down.

I'm not sure if I read it or just assumed it but it seems like the sort of
thing where it may well have been the same die, just a way to sell some with
a flaw on one of the cache banks.

I don't think so. With the smaller cache the K6-2+ was a faster sort than
the K6-III. If it were a defective cache sort it wouldn't be any faster.

Much too long ago, and super-7 was already enough on the way out at the time
that I didn't follow it very much. I loved the K6-2/300 I had, even if the
VIA-based board I had was kinda flakey, but by the time the -III/II+ were
selling much I was already in the "wait for K7" mode.

I skipped the K7 entirely (I did my the kid a K7 system). I didn't see
the need to upgrade the K6-III, and when I did the K8 was here, though I
paid dearly for not waiting for S939. Oh well, I paid dearly for my
"first-day-order" IBM PC too. ;-)

 

[Del Cecchi]

^^^^^^^^^^^
East Fishkill is *not* a research facility. Other than that, I won't
comment.

Well, EF has the pilot line and SRC in addition to the 300mm line that
howie got outbid on. He didn't want all those hazardous chemicals in his
state anyway.

snip

del

 

[imouttahere]

Opteron and G5 shouldn't have been tracking each other at all. The
Opteron was built with a 12-stage pipeline, while the G5 has a 24-stage
pipeline. The G5 should've been clocking up like a TARDIS.

Opteron: 12 int/17 fp stages 970FX: 16/21

http://www.anandtech.com/printarticle.aspx?i=2436

The problem with the G5, I gather, is heat dissipation at higher clocks.

 

[Keith R. Williams]

Well, EF has the pilot line

Hasn't the ASTC been closed down?

and SRC in addition to the 300mm line that howie got outbid on.

YEEHAWW Howie wasn't invited to the auction. No water. No
electricity. No land. No workers. Lotsa taxes buys lotsa bureaucracy
though. It's taken forty years to build a two-lane road, and it'll
likely be another forty before it's finished.

He didn't want all those hazardous chemicals in his state anyway.

If people only knew... You shoulda seen people turn white when there
was a b*mb scare on the main site some ten years back (long before
9/11).

 

[YKhan]

^^^^^^^^^^^
East Fishkill is *not* a research facility. Other than that, I won't
comment.

It has a research component. No one is going to do experiments on a
working production line.

You think this is novel? Processes are tweaked on the fly to produce
what's needed all the time.

But each manufacturer has its own history to draw upon when it creates
its own automated process management scheme. AMD's history (including
all of its failures) is particularly relevant to the manufacture of x86
microprocessors. Most of this is learned experience, not theoretical.
AMD has much more experience at producing high-speed processors at good
yields than IBM. This is borne out by IBM's severe yields problem at
Fishkill compared to AMD's relative lack of it using nearly indentical
equipment in Dresden.

AMD needed some help from IBM about integrating new materials into its
processes. But once that knowledge was gained, it would be AMD that
would have the better chance at getting it going on a big scale.

Yousuf Khan