AMD ready to introduce triple-core processors

[Yousuf Khan]

X-bit labs - AMD Thinks Triple-Core Microprocessors – Rumours.
"The new triple-core microprocessors will feature the same silicon as
quad-core chips, but with one core disabled. Nevertheless, the chips
will still include 2MB of shared L3 cache and will take advantage of
other K10 micro-architecture features, such as SSE4A instruction set,
128-bit floating point units (FPU) and so on. Obviously, the chips will
also have advanced power management capabilities."
http://www.xbitlabs.com/news/cpu/display/20070914212726.html

AMD prepares three-core processors
"AMD is probably doing this for two reasons; the lesser being salvage,
the more important one being that Intel can't do it. Intel would have a
far harder time making a tri-core part until Nehalem next September - it
is easy to fuse off a core, far harder to MCM disparate cores."
http://www.theinquirer.net/default.aspx?article=42369

 

[daytripper]

X-bit labs - AMD Thinks Triple-Core Microprocessors – Rumours.
"The new triple-core microprocessors will feature the same silicon as
quad-core chips, but with one core disabled. Nevertheless, the chips
will still include 2MB of shared L3 cache and will take advantage of
other K10 micro-architecture features, such as SSE4A instruction set,
128-bit floating point units (FPU) and so on. Obviously, the chips will
also have advanced power management capabilities."
http://www.xbitlabs.com/news/cpu/display/20070914212726.html

AMD prepares three-core processors
"AMD is probably doing this for two reasons; the lesser being salvage,
the more important one being that Intel can't do it. Intel would have a
far harder time making a tri-core part until Nehalem next September - it
is easy to fuse off a core, far harder to MCM disparate cores."
http://www.theinquirer.net/default.aspx?article=42369

Yeah...well, ok, but a partial is a partial no matter why it exists ;-)

/daytripper
(not that it's necessarily a *bad* thing, but let's call it what it is)

 

[Gary Seven]

: On Sat, 15 Sep 2007 13:40:17 -0400, Yousuf Khan
:
:: X-bit labs - AMD Thinks Triple-Core Microprocessors - Rumours.
:: "The new triple-core microprocessors will feature the same
:: silicon as quad-core chips, but with one core disabled.
:: Nevertheless, the chips will still include 2MB of shared L3
:: cache and will take advantage of other K10 micro-architecture
:: features, such as SSE4A instruction set, 128-bit floating
:: point units (FPU) and so on. Obviously, the chips will also
:: have advanced power management capabilities."
:: http://www.xbitlabs.com/news/cpu/display/20070914212726.html
::
:: AMD prepares three-core processors
:: "AMD is probably doing this for two reasons; the lesser being
:: salvage, the more important one being that Intel can't do it.
:: Intel would have a far harder time making a tri-core part
:: until Nehalem next September - it is easy to fuse off a core,
:: far harder to MCM disparate cores."
:: http://www.theinquirer.net/default.aspx?article=42369
:
:
: Yeah...well, ok, but a partial is a partial no matter why it
: exists ;-)
:
: /daytripper
: (not that it's necessarily a *bad* thing, but let's call it
: what it is)

Heh heh. Isn't this simply called "bining" <sp?>?

G7

 

[daytripper]

: On Sat, 15 Sep 2007 13:40:17 -0400, Yousuf Khan
:
:: X-bit labs - AMD Thinks Triple-Core Microprocessors - Rumours.
:: "The new triple-core microprocessors will feature the same
:: silicon as quad-core chips, but with one core disabled.
:: Nevertheless, the chips will still include 2MB of shared L3
:: cache and will take advantage of other K10 micro-architecture
:: features, such as SSE4A instruction set, 128-bit floating
:: point units (FPU) and so on. Obviously, the chips will also
:: have advanced power management capabilities."
:: http://www.xbitlabs.com/news/cpu/display/20070914212726.html
::
:: AMD prepares three-core processors
:: "AMD is probably doing this for two reasons; the lesser being
:: salvage, the more important one being that Intel can't do it.
:: Intel would have a far harder time making a tri-core part
:: until Nehalem next September - it is easy to fuse off a core,
:: far harder to MCM disparate cores."
:: http://www.theinquirer.net/default.aspx?article=42369
:
:
: Yeah...well, ok, but a partial is a partial no matter why it
: exists ;-)
:
: /daytripper
: (not that it's necessarily a *bad* thing, but let's call it
: what it is)

Heh heh. Isn't this simply called "bining" <sp?>?

Kinda, though as classically used wrt semiconductors, "binning" means
"functionally identical but with differing speed".

This is much more like the old "partially good" semiconductor memories sold by
foundries on the cheap to companies that could find ways to use the good areas
of the arrays and avoid the defective regions (for just one example, the solid
state disks sold by DEC in the '90s, which combined a robust ecc code with
strategically applied array addressing).

As a marketing message, I really don't think "We can make devices that are 75%
functional" is going to sway many Intelophiles...

/daytripper (but then, I could be wrong ;-)

 

[Jan Panteltje]

As a marketing message, I really don't think "We can make devices that are 75%
functional" is going to sway many Intelophiles...

/daytripper (but then, I could be wrong ;-)

IBM does the same with Cell in PS3.
It just differentiates to a more granular price scale.

 

[krw]

100% good ones catching fire wouldn't help their marketing plans
either.

Never!

IBM does the same with Cell in PS3.
It just differentiates to a more granular price scale.

IBM has done it since forever (I saw it first in memories in the
'70s). The Z8 had ten processors. The processors you didn't pay for
were disabled by the crypto hardware. If one of the enabled ones
died another took over. If you needed more MIPS to do your taxes
this month, simply call you friendly salesman and buy/rent another
processor, or eight, and the mothership would download the
appropriate cryptographically protected microcode to enable the
processors for the time needed (a.k.a. "Dial-A-MIP").

 

[Yousuf Khan]

Yeah...well, ok, but a partial is a partial no matter why it exists ;-)

Well, these days the dual-cores are becoming the low-end, so if quads
are the high-end, then something has to fill the middle. You could try
to fill the middle with faster duals, or slower quads, or you could just
as well try selling tri-cores. Voila, Marquetting!

 

[daytripper]

Well, these days the dual-cores are becoming the low-end, so if quads
are the high-end, then something has to fill the middle. You could try
to fill the middle with faster duals, or slower quads, or you could just
as well try selling tri-cores. Voila, Marquetting!

Well, here is a test for you then: assuming there is *some* premium for a
triple core vs an otherwise comparable dual core, would you buy the former?

/daytripper

 

[krw]

Well, here is a test for you then: assuming there is *some* premium for a
triple core vs an otherwise comparable dual core, would you buy the former?

Define "otherwise comparable"? Is there a benefit? Why would I buy
a dual if a single was "otherwise comparable"?

 

[daytripper]

Define "otherwise comparable"? Is there a benefit? Why would I buy
a dual if a single was "otherwise comparable"?

Assume my awkward phrasing referred to cores of equal performance...

/daytripper

 

[krw]

Assume my awkward phrasing referred to cores of equal performance...

Ok, then the answer is, "what premium for what benefit?", or in other
words, a *definite* maybe.

 

[The little lost angel]

Well, here is a test for you then: assuming there is *some* premium for a
triple core vs an otherwise comparable dual core, would you buy the former?

That depends on the premium amount, the definition of comparable and
one's budget wouldn't it?

There must be certain workloads that would scale better with
additional cores such that having triple cores would certainly
outperform a dual core significantly so those on a budget would be
happy paying that premium for that extra performance, no?

 

[Gary Seven]

The little lost angel <[email protected]>
wrote:
: On Sun, 16 Sep 2007 21:09:22 -0400, daytripper
:
:: Well, here is a test for you then: assuming there is *some*
:: premium for a triple core vs an otherwise comparable dual
:: core, would you buy the former?
:
: That depends on the premium amount, the definition of
: comparable and one's budget wouldn't it?
:
: There must be certain workloads that would scale better with
: additional cores such that having triple cores would certainly
: outperform a dual core significantly so those on a budget
: would be happy paying that premium for that extra performance,
: no?

Ahem, may I? I would LOVE that Photoshop CS3 could utilize two of my three
core processors while the other was attending to DVD (re)encoding. Or
better, two cores dedicated to encoding while the other is free just to do
"normal" computer stuff...

G7

 

[daytripper]

Ok, then the answer is, "what premium for what benefit?", or in other
words, a *definite* maybe.

Here's a related article with a couple of interesting twists that would
compound the equation...

http://www.boston.com/business/tech...d_to_sell_chip_with_three_electronics_brains/

/daytripper

 

[Yousuf Khan]

Well, here is a test for you then: assuming there is *some* premium for a
triple core vs an otherwise comparable dual core, would you buy the former?

Assuming that the premium was minimal, then the answer would have to be
yes. Let's say if the difference between the dual- and tri-cores were
less than the price difference between a tri- and quad-cores.

Yousuf Khan

 

[Yousuf Khan]

Here's a related article with a couple of interesting twists that would
compound the equation...

http://www.boston.com/business/tech...d_to_sell_chip_with_three_electronics_brains/

/daytripper

Okay, so what are the interesting twists that you refer to?

 

[daytripper]

Okay, so what are the interesting twists that you refer to?

Oh, ok. I thought this one would jump right out at you:

"By regulating the speed at which each core operates, AMD could conceivably
sell a triple-core chip that has higher performance metrics than one of its
own quad-core chips, said Insight 64 analyst Nathan Brookwood."

So - do I need to explain why that is interesting? ;-)

And then there's this:

"With AMD's Phenom triple-core chip, AMD hopes to speed the adoption of
multi-core chips, since sales of PCs with quad-core chips have been
lackluster."

I dunno. Seems twisted to me - they can't move quads, but somehow triples are
going to bring home the bacon?

There's more. Read carefully for effect ;-)

/daytripper

 

[krw]

Oh, ok. I thought this one would jump right out at you:

"By regulating the speed at which each core operates, AMD could conceivably
sell a triple-core chip that has higher performance metrics than one of its
own quad-core chips, said Insight 64 analyst Nathan Brookwood."

So - do I need to explain why that is interesting? ;-)

Yes. The key word for me is "conceivably". Certainly it's possible
to cripple a quad or "over-clock" a lesser product. Of course such
things are thinkable. They normally don't get a second thought
though.

And then there's this:

"With AMD's Phenom triple-core chip, AMD hopes to speed the adoption of
multi-core chips, since sales of PCs with quad-core chips have been
lackluster."

Perhaps price has something to do with it? Why else would a tipple
make it more palatable to the end-user? Tvo vasn't enuff, but veer
vas too many?

I dunno. Seems twisted to me - they can't move quads, but somehow triples are
going to bring home the bacon?

There's more. Read carefully for effect ;-)

Just looks like the typical clueless-jounalist nonsense to me.

 

[Yousuf Khan]

Oh, ok. I thought this one would jump right out at you:

"By regulating the speed at which each core operates, AMD could conceivably
sell a triple-core chip that has higher performance metrics than one of its
own quad-core chips, said Insight 64 analyst Nathan Brookwood."

So - do I need to explain why that is interesting? ;-)

Well, okay, but how much additional frequency do they expect? With the
cores so close together, one must assume that it doesn't take too much
additional power to power-up an additional core. And vice-versa, it
probably doesn't get back too much power when powering down a core, either.

And then there's this:

"With AMD's Phenom triple-core chip, AMD hopes to speed the adoption of
multi-core chips, since sales of PCs with quad-core chips have been
lackluster."

I dunno. Seems twisted to me - they can't move quads, but somehow triples are
going to bring home the bacon?

There's more. Read carefully for effect ;-)

The triples are likely only heading for the desktop market.

Yousuf Khan

 

[krw]

Well, okay, but how much additional frequency do they expect? With the
cores so close together, one must assume that it doesn't take too much
additional power to power-up an additional core. And vice-versa, it
probably doesn't get back too much power when powering down a core, either.

The third core is going to take, oh, about 50% more power than two.

The triples are likely only heading for the desktop market.

If not the scrap heap.