AMD planning 45nm 12-Core 'Istanbul' Processor ?

  • Thread starter Thread starter AirRaid
  • Start date Start date
A

AirRaid

http://www.dailytech.com/Dodecacore...fficially+the+Multicore+Race/article11531.htm

Hardware Dodeca-core: The Megahertz Race is Now Officially the Multi-
core Race


AMD engineers reveal details about the company's upcoming 45nm
processor roadmap, including plans for 12-core processors

"Shanghai! Shanghai!" the reporters cry during the AMD's financial
analyst day today. Despite the fact that the company will lay off
nearly 5% of its work force this week, followed by another 5% next
month, most employees interviewed by DailyTech continue to convey an
optimistic outlook.

The next major milestone for the CPU engineers comes late this year,
with the debut of 45nm Shanghai. Shanghai, for all intents and
purposes, is nearly identical to the B3 stepping of Socket 1207
Opteron (Barcelona) shipping today. However, where as Barcelona had
its HyperTransport 3.0 clock generator fused off, Shanghai will once
again attempt to get HT3.0 right.

Original roadmaps anticipated that HT3.0 would be used for socket-to-
socket communication, but also for communication to the Southbridge
controllers. Motherboard manufacturers have confirmed that this is no
longer the case, and that HT3.0 will only be used for inter-CPU
communication.

"Don't be disappointed, AMD is making up for it," hints one engineer.
Further conversations revealed that inter-CPU communication is going
to be a big deal with the 45nm refresh. The first breadcrumb comes
with a new "native six-core" Shanghai derivative, currently codenamed
Istanbul. This processor is clearly targeted at Intel's recently
announced six-core, 45nm Dunnington processor.

But sextuple-core processors have been done, or at least we'll see the
first ones this year. The real neat stuff comes a few months after,
where AMD will finally ditch the "native-core" rhetoric. Two separate
reports sent to DailyTech from AMD partners indicate that Shanghai and
its derivatives will also get twin-die per package treatment.

AMD planned twin-die configurations as far back as the K8
architecture, though abandoned those efforts. The company never
explained why those processors were nixed, but just weeks later
"native quad-core" became a major marketing campaign for AMD in
anticipation of Barcelona.

A twin-die Istanbul processor could enable 12 cores in a single
package. Each of these cores will communicate to each other via the
now-enabled HT3.0 interconnect on the processor.

The rabbit hole gets deeper. Since each of these processors will
contain a dual-channel memory controller, a single-core can emulate
quad-channel memory functions by accessing the other dual-channel
memory controller on the same socket. This move is likely a
preemptive strike against Intel's Nehalem tri-channel memory
controller.

Motherboard manufacturers claim Shanghai and its many-core derivatives
will be backwards compatible with existing Socket 1207 motherboards.
However, processor-to-processor communication will downgrade to lower
HyperTransport frequencies on these older motherboards. The newest
1207+ motherboards will officially support the HyperTransport 3.0
frequencies.

Shanghai is currently taped out and running Windows at AMD.



.....................................................................................................................................

http://blogs.zdnet.com/processors/?p=162

AMD can’t get to 45nm fast enough.
A 12-core package?


If there was a bright spot in AMD’s first quarter, it was the new PC
and graphics products that have either recently shipped or are in the
pipeline for 2008. Barcelona, the company’s quad-core design
manufactured using a 65nm process, bombed last year. Meanwhile, Intel
is minting millions of 45nm chips. So it’s no surprise that AMD is
shifting to 45nm processors as quickly as possible.

During the earnings call last week, AMD executives said the 45nm
processor, code-named Shanghai, was on track and would be shipping in
volume in the fourth quarter. By all accounts, Shanghai is largely a
“shrink” of the recently-released B3 version of Barcelona, which
corrected a flaw in the original design. But Daily Tech reports that
AMD has more ambitious plans for the 45nm design.

Citing AMD engineers, the site says AMD will produce a six-core
version, code-named Istanbul, and then, in a reversal of its “native-
quad-core” strategy of putting all cores on a single piece of silicon,
offer a two-die package with a total of 12 cores. The two processors
will be connected using AMD’s HyperTransport 3.0 bus.

To be clear, this platform is designed for the server and workstation
market. Based on recent presentations, AMD will also release a 45nm
enthusiast desktop platform, code-named Leo, later this year that will
consist of triple-core and quad-core chips. Mainstream and budget
desktops, as well as business systems, will continue to use platforms
based on 65nm processors until sometime in 2009.

.....................................................................................................................................

http://www.gadgetell.com/tech/comment/amd-readying-12-core-45nm-processors-for-late-2008/

AMD readying 12-core 45nm processors for late 2008
by David Gonzales on Apr 19, 2008 at 10:29 PM



Forget quad-core, AMD is preparing a dodeca-core chip. If you didn’t
stay in boarding school if you need refreshing with your greek
vocabulary, dodeca means 12, and that’s 12 cores for you right there.
Imagine, a dozen times faster than a normal chip. But how fast does
one’s computer ever really need to be? AMD doesn’t seem to care if
there’s an answer to that question or not (just as its closest
competitor Intel doesn’t), and moves on with their plan to produce a
12-core processor to be released by late 2008. It will reportedly be
called the Shanghai, and will be a 45nm successor to their not-so-
successful Barcelona chip.

The AMD Shanghai will supposedly come with HyperTransport 3.0 and six
cores so it can stand up against Intel’s upcoming Dunnington chips in
a race towards becoming the fastest processors on Earth. And while the
six-core variety of the Shanghai isn’t even close to coming to market,
AMD already has plans for a follow-up: a 12-core version, called
Istanbul, packing 12 cores (what else?) and HyperTransport 3.0
interconnect as well.

Get that? The first version of this product is already 3.0 while some
companies can’t even get 2.0 right. That’s what they meant with all
the numbers, right?


.....................................................................................................................................

http://www.electronista.com/articles/08/04/19/amd.reveals.45nm.shanghai/

AMD talks 45nm Shanghai w/12-cores, HyperTransport 3

AMD engineers this week said that the company plans on introducing new
12-core processors later this year. The first processors based on 45nm
Shanghai platform are due later this year and will be nearly identical
to the B3 variant of the Socket 1207 Opteron (Barcelona) shipping
today, according to DailyTech. The processors will reportedly use the
faster HyperTransport 3.0 for inter-CPU communication and will debut
later this year as a "native six-core" Shanghai derivative, currently
code-named Istanbul. That processor, the report claims, is "clearly
targeted at Intel's recently announced six-core, 45nm Dunnington
processor." A few months later, Shanghai and its derivatives will also
get twin-die per package treatment, allowing for up to 12-cores per
package, the report says.

Each of these processors will contain a dual-channel memory
controller, allowing a single-core to emulate quad-channel memory
functions by accessing the other dual-channel memory controller on the
same socket and offering an alternative to Intel's Nehalem tri-channel
memory controller, DailyTech notes.

Citing motherboard manufacturers, the report says that Shanghai and
its multiple-core derivatives will be backwards compatible with
existing Socket 1207 motherboards, but that processor-to-processor
communication will downgrade to lower HyperTransport frequencies on
older motherboards. The publication also notes that the newer 1207+
motherboards will officially support the HyperTransport 3.0
frequencies.
 
AirRaid said:
AMD planned twin-die configurations as far back as the K8
architecture, though abandoned those efforts. The company never
explained why those processors were nixed, but just weeks later
"native quad-core" became a major marketing campaign for AMD in
anticipation of Barcelona.

A twin-die Istanbul processor could enable 12 cores in a single
package. Each of these cores will communicate to each other via the
now-enabled HT3.0 interconnect on the processor.

The rabbit hole gets deeper. Since each of these processors will
contain a dual-channel memory controller, a single-core can emulate
quad-channel memory functions by accessing the other dual-channel
memory controller on the same socket. This move is likely a
preemptive strike against Intel's Nehalem tri-channel memory
controller.


Well, at least there is an air of practicality entering AMD's designs
now. Twin-die 12-cores will get them to market much sooner.

Yousuf Khan
 
Well, at least there is an air of practicality entering AMD's designs
now. Twin-die 12-cores will get them to market much sooner.

Yousuf Khan

They'd better don't screw up on this one - they already used up all
the margin of error they had, and then some. But then, it's not the
first time AMD is on the brink of extinction, and they always came
back even stronger than anyone, including but not limited to INTC,
could expect. K7 in 1999, K8 in 2003 - both times Intel got their ass
whipped big time for a couple years straight... Why not this time
again? Wish them luck - without AMD the whole CPU world would be
boring and damn expensive.

NNN
 
They'd better don't screw up on this one - they already used up all
the margin of error they had, and then some. But then, it's not the
first time AMD is on the brink of extinction, and they always came
back even stronger than anyone, including but not limited to INTC,
could expect. K7 in 1999, K8 in 2003 - both times Intel got their ass
whipped big time for a couple years straight... Why not this time
again? Wish them luck - without AMD the whole CPU world would be
boring and damn expensive.

NNN

Let's hope they can pull it off. I'm still waiting for a dual core from them
that actually signifigantly ouperforms my S939 Opteron 185 2Mb unit at
3.2Ghz. The performance increases I've seen from my A8N-E system versus the
one I just built for my son, with an X2 5600+ 2Mb unit clocked at 3.2 Ghz
are solely due to the newer chipset and DDR2 clockings. I'm looking at
upgrading my system in the next few months, and I don't see a Phenom or an
X2 6400+ in my future, unfortunately. I swore by the PIII Coppermine and
Tualatin line, but then Intel went to the Willamette then Northwood P4's.
That's when I started with Bartons, A64's and Opterons. The Intel lineup
can't be beat right now, it's been leaps and bounds ahead of AMD for the
last couple of years. Intel needs real competition, and not just in the
bottom feeder lineup. My T7600 4Mb L2 Core2Duo runs circles around both of
my 3.2Ghz AMD units.
 
I upgraded one of my machines to a Core 2 Quad Q6600 a while back;
mencoder
has a limited amount of multithreading support that's available with
different codecs. The highest CPU usage I see with one mencoder process
transcoding a DVD to H.264 is usually somewhere around 200-250% (400%
would
be fully loaded on all 4 cores). To take full advantage of the processor,
I
just do two transcodes at a time. :-)

(With the encoding processes niced, it can still play HD MPEG-2 from
across
the network without any hiccups, too. That slows down encoding a little,
as
decoding and displaying HD takes up a fair chunk of one core.)

I'm using my Opteron 185 box ( as well as an Opteron 180) for digital
mixing and music using Cubase 4 on one and Sonar Producer 7 on the other.
Both are written for multicore. Both choke when adding and mixing effects in
a 10-12 track mix. Badly. I ran Sonar 7 Producer on to the T7600 Dell
laptop, installed the same M-Audio MobilePre USB and mixing hardware, and
believe me, the Intel Core2Duo setup slows down, but never stops and flakes
out. Same memory on each. The Opterons are both running RAID0 setups with
256Kb stripes. The Dell is a 7200RPM laptop HDD. So there's just a bit of
difference in multicore ability there.
 
Wish them luck - without AMD the whole CPU world would be
boring and damn expensive.

Amdroid fantasy. The PC as a platform is pushed on the low end by
game boxes and on the high end by IBM and Sun.

What could be more boring than the x86 domination we have now?
Without AMD, there would be no such monoculture.

As it is, Via is still in the x86 business and will continue in the
x86 business, even if AMD fails.

No matter what, Intel will have price and performance competition.
Things might have progressed more slowly, but Intel would be pushing
the fab envelope, anyway, because that's where it's real strength is.
Take AMD out of the picture completely and there is still no way for
Intel to relax.

Robert.
 
Amdroid fantasy. The PC as a platform is pushed on the low end by
game boxes and on the high end by IBM and Sun.

What could be more boring than the x86 domination we have now?
Without AMD, there would be no such monoculture.

Competition.
Intel would still be with Pentium 2 if not for AMD

Does it have to be crossposted to the whole known universe?
As it is, Via is still in the x86 business and will continue in the
x86 business, even if AMD fails.

I agree other architectures have a possibility here.
But MS, not releasing source of their OS, and only with an x86
binary version, has kept x86 going.

Linux and other Unices allow almost any architecture to succeed.
You can run Linux on PS3 and Cell for example.
And that is not even 'low end'.
 
Does it have to be crossposted to the whole known universe?

That would explain why your reply is the first I've seen of this
thread. People who crosspost should know their messages are filtered
out by many servers and users.

Steve
 
Did you mean "without Intel"?

I suppose if AMD hadn't been around, there would have been a greater
chance of Intel getting their butts whipped by some other architecture,
instead of by AMD.

Of course AMD _did_ come up with "x86-64", which is an improvement over
the x86 (obviously even Intel thinks so).

-Miles
 
Or IA64, with Intel having a complete monopoly, and all that implies.

Without AMD, one of threee things would have happened:

1. Intel would have had the resources to deliver a satisfactory
Itanium product and accompanying compiler on schedule.

2. Intel would have been forced into a partial retreat to x86, anyway.

3. Sparc or Power would be holding much larger market share under any
number of possible licensing and manufacturing arrangements.

I personally believe that what we've got is the worst of all possible
worlds: AMD on death's door, Microsoft holding on to its monopoly
catering to just one ISA, and, to all intents and purposes, zero
diversity in processor architecture.

Robert.
 
(e-mail address removed)>, (e-mail address removed)
says...
Without AMD, one of threee things would have happened:

1. Intel would have had the resources to deliver a satisfactory
Itanium product and accompanying compiler on schedule.

Oh, give me a break! They blew an infinite source of money down
that rat hole as it was.
2. Intel would have been forced into a partial retreat to x86, anyway.

How did you come up with that dream?
3. Sparc or Power would be holding much larger market share under any
number of possible licensing and manufacturing arrangements.

That one too?
I personally believe that what we've got is the worst of all possible
worlds: AMD on death's door, Microsoft holding on to its monopoly
catering to just one ISA, and, to all intents and purposes, zero
diversity in processor architecture.

Of course you do; anything but competition from AMD.
 
I suppose if AMD hadn't been around, there would have been a greater
chance of Intel getting their butts whipped by some other architecture,
instead of by AMD.

What architecture? You grossly underestimate the x86 inertia.
Of course AMD _did_ come up with "x86-64", which is an improvement over
the x86 (obviously even Intel thinks so).

Intel did too, but had no interest in pushing it forward to product.
 
On Mon, 28 Apr 2008 10:22:52 -0700 (PDT), Robert Myers
Without AMD, one of threee things would have happened:

1. Intel would have had the resources to deliver a satisfactory
Itanium product and accompanying compiler on schedule.
Itanium was stillborn, even more so than Netbust. Its only purpose
was to move everyone and their mother-in-law away from x86 and in the
process thereof screw all other chipmakers (chiefly AMD, since others
were, and still are, almost non-entities). While AMD and some other
guys have a license to churn out x86 compatible product, no licenses
were ever planned for IA64. Thanks to AMD and their Opteron product
beating Itanic on performance for a mere fraction of the price, it
didn't happen.
2. Intel would have been forced into a partial retreat to x86, anyway.
It took AMD64 (later renamed x84-64 to make it more digestable to
Intel) to do so. In all Intel roadmaps, x86 was to be relegated to
the low end of the market and then obsoleted in a matter of a few
years, if not months.
3. Sparc or Power would be holding much larger market share under any
number of possible licensing and manufacturing arrangements.
Sparc is big iron stuff, it just doesn't scale down to desktop, let
alone laptop. And Power... It could not even hold on to Apple, the
only desktop/laptop maker ever using it. Ironically, it was dumped to
make way to Intel x86 product.
I personally believe that what we've got is the worst of all possible
worlds: AMD on death's door, Microsoft holding on to its monopoly
catering to just one ISA, and, to all intents and purposes, zero
diversity in processor architecture.
Some prefer divercity, others prefer standard. Looks like you are not
in the business of writing software, otherwise you'd know what a pain
in a$$ is cross-platform compatibility.NNN
 
krw said:
What architecture? You grossly underestimate the x86 inertia.

"greater chance"

Not necessarily a _good_ chance, but it's the only way I can make sense
of the grandparent poster's ranting...

-Miles
 
Not necessarily a _good_ chance, but it's the only way I can make sense
of the grandparent poster's ranting...
Don't try to make sense of something containing none of it whatsoever.

NNN
 
Itanium was stillborn, even more so than Netbust. Its only purpose
was to move everyone and their mother-in-law away from x86 and in the
process thereof screw all other chipmakers (chiefly AMD, since others
were, and still are, almost non-entities). While AMD and some other
guys have a license to churn out x86 compatible product, no licenses
were ever planned for IA64. Thanks to AMD and their Opteron product
beating Itanic on performance for a mere fraction of the price, it
didn't happen.
I've lived a long time now, and I've seen a lot of predictions come
and go. Anyone who wants to make emphatic statements about what was
inevitable should take a good, hard look at all the successful
predictions of the ramifications of the attack of the killer micros.

In all this "I knew all along" talk about Itanium, I've heard a few
insightful comments indicating that people actually understood
something of importance about the actual architecture, and not what
they've heard from others. If anyone *really* understood what went
wrong with Itanium, it would make the case study of all case studies
for business schools interested in the development and management of
technology. As it is, I don't think anyone really knows.

By comparison, it's pretty easy to see what went wrong with Netburst
and, among other things, we have public statements by its principal
architect, who no longer works for Intel. Even so, it's a puzzle as
to why Intel missed the importance of power consumption.
It took AMD64 (later renamed x84-64 to make it more digestable to
Intel) to do so. In all Intel roadmaps, x86 was to be relegated to
the low end of the market and then obsoleted in a matter of a few
years, if not months.

The more fool Intel. The "low end" is where all the action is. To
see that, you have only to look at Blue Gene that was built with "low
end" processors. Low end or high wasn't what mattered. Power
consumption did. Intel has it figured out by now, and they're pouring
resources into low-power processors. If Intel hadn't had a credible
x86 candidate, what would it have done? I have no idea. To give up
on the low-power market is essentially to give up on the future
because the cost of computation is going to be dominated by the cost
of electricity, including the costs of cooling.
Sparc is big iron stuff, it just doesn't scale down to desktop, let
alone laptop. And Power... It could not even hold on to Apple, the
only desktop/laptop maker ever using it. Ironically, it was dumped to
make way to Intel x86 product.
Sparc would have much of the market share for servers now dominated by
x86. The disappearance of Power is simply a matter of money. If the
market is entirely consumed by x86, no one will want to put the
resources into it necessary to compete with Intel's offerings. To be
sure, IBM has never been very much interested in that market, anyway.
Some prefer divercity, others prefer standard. Looks like you are not
in the business of writing software, otherwise you'd know what a pain
in a$$ is cross-platform compatibility.
I have very little sympathy for the concerns of software developers.
We'd be much better off with longer software development cycles so we
had less bad software.

Robert.
 
In comp.sys.ibm.pc.hardware.chips Robert Myers said:
Without AMD, one of threee things would have happened:
1. Intel would have had the resources to deliver a satisfactory
Itanium product and accompanying compiler on schedule.

Are you joking? AFAICS, Intel threw money and people at Itanium
well past the point of diminishing returns into the region of
negative returns (were additional people/resources consumed more
communications and managment than their contribution to the project).

Itanium did not fail for lack of resources. It might have failed
from a surfeit. Dis-economies of scale are real and a constant
peril in large projects.
2. Intel would have been forced into a partial retreat to x86, anyway.

Would it have been any more graceful than P4, ie an overclocked
original Pentium? Perhaps. It could hardly be worse.
Without AMD breathing down Intel's neck on the performance end,
INTC would have developed and released processors much slower.
There would have been no need.
3. Sparc or Power would be holding much larger market share under
any number of possible licensing and manufacturing arrangements.

Add in Alpha. In addition to the non-x86 code hurdle, they
have identical issues with AMD. None can deliver the sheer
massive volume that Intel can and the PC market demands.
I personally believe that what we've got is the worst of all
possible worlds: AMD on death's door, Microsoft holding on to
its monopoly catering to just one ISA, and, to all intents and
purposes, zero diversity in processor architecture.

I do not agree. Linux and to a lesser extent NetBSD have done
wonders to keep alt-arch alive and vital. ARM is far from dead.
I would not be at all surprised my next PC had one (ASUS EEE-like).

MSFT has been seduced down a dead-end. Sic transit gloria mundi.


-- Robert
 
Back
Top