Interesting read about upcoming K9 processors

  • Thread starter Thread starter Yousuf Khan
  • Start date Start date
Y

Yousuf Khan

http://www.digitimes.com/news/a20040726PR202.html

This interview with Tyan president Symon Chang provided the following
quotes:

"Around 2006, when the market moves to AMD's next generation of chips, you
will be able to go over 8-way. What I mean is that with eight sockets, and
dual cores, you then have sixteen processors, but with K9, you'll see it go
over that. I think we'll see a significant increase in the amount of
crossbar switches in the CPU. I'm not up on all the minute details, but you'
ll be able to go over 60 processors without adding any external crossbar
chips. We can do all that within the structure that is being currently
created. The crossbar bar chip is the standard in the mainframe business
whether it is for the Xeon, Opteron or other processors. There are a couple
of versions of the crossbar chip today, but I don't think that anyone is
currently using them for anything in the generic market; these solutions are
really only for the mainframe market. Today's mainframe market with
computers from IBM or Sparc will be using up to and over 128 processors,
with chips such as IBM's 390 microprocessor. These machines are starting
around US$1 million."

That's right over 60 processors without any kind of a special chipset
support!!!

Also he had some opinions about Windows XP64:

"Q: Do you think Microsoft's 64-bit OS will come out on time?

Chang: I hope so. There are delays, but I believe it will. Interestingly
enough, a couple of significant things have happened this year; for example,
Intel's Xeon processor with 64-bit extensions is a reaction to the
unexpected popularity of AMD's Opteron, which put Intel under pressure to
provide a similar solution for the OEM market. If Intel had not reacted, it
would have lost out. Their response was to come out with a 64-bit CPU that
is not optimal, but at least they have it, and I would compare that with
what Microsoft is doing now in the realm of the 64-bit operating system."

Doesn't sound like Chang believes that Microsoft is trying all that hard to
build a 64-bit OS. It's getting something out to show that it isn't behind
the times.

Yousuf Khan
 
Yousuf Khan said:
"Q: Do you think Microsoft's 64-bit OS will come out on time?

Are we still supposed to be excited about a 64-bit desktop OS from MS after
all these years? I heard once it was going to be a slam dunk. Guess
not... :-)

Regards,
Dean
 
http://www.digitimes.com/news/a20040726PR202.html

This interview with Tyan president Symon Chang provided the following
quotes:

"Around 2006, when the market moves to AMD's next generation of chips, you
will be able to go over 8-way. What I mean is that with eight sockets, and
dual cores, you then have sixteen processors, but with K9, you'll see it go
over that. I think we'll see a significant increase in the amount of
crossbar switches in the CPU. I'm not up on all the minute details, but you'
ll be able to go over 60 processors without adding any external crossbar
chips. We can do all that within the structure that is being currently
created. The crossbar bar chip is the standard in the mainframe business
whether it is for the Xeon, Opteron or other processors. There are a couple
of versions of the crossbar chip today, but I don't think that anyone is
currently using them for anything in the generic market; these solutions are
really only for the mainframe market. Today's mainframe market with
computers from IBM or Sparc will be using up to and over 128 processors,
with chips such as IBM's 390 microprocessor. These machines are starting
around US$1 million."

That's right over 60 processors without any kind of a special chipset
support!!!

Also he had some opinions about Windows XP64:

"Q: Do you think Microsoft's 64-bit OS will come out on time?

Chang: I hope so. There are delays, but I believe it will. Interestingly
enough, a couple of significant things have happened this year; for example,
Intel's Xeon processor with 64-bit extensions is a reaction to the
unexpected popularity of AMD's Opteron, which put Intel under pressure to
provide a similar solution for the OEM market. If Intel had not reacted, it
would have lost out. Their response was to come out with a 64-bit CPU that
is not optimal, but at least they have it, and I would compare that with
what Microsoft is doing now in the realm of the 64-bit operating system."

Hmmm, "not optimal"?? How much can we read into that on top of the
deafening EM64T silence on WWW? As for "unexpected popularity of AMD's
Opteron"... must be a riddle... too difficult for me.:-)
Doesn't sound like Chang believes that Microsoft is trying all that hard to
build a 64-bit OS. It's getting something out to show that it isn't behind
the times.

.... or that we'll get it when Intel is good and ready for us to have it.:-(

Rgds, George Macdonald

"Just because they're paranoid doesn't mean you're not psychotic" - Who, me??
 
Yousuf said:
Doesn't sound like Chang believes that Microsoft is trying all that hard to
build a 64-bit OS. It's getting something out to show that it isn't behind
the times.

Going to 64 bits will be trivial compared to going to 64 way for Microsoft.

Joe Seigh
 
George said:
[SNIP]
Doesn't sound like Chang believes that Microsoft is trying all that hard to
build a 64-bit OS. It's getting something out to show that it isn't behind
the times.


... or that we'll get it when Intel is good and ready for us to have it.:-(

That just means MS loses market & mindshare to other more capable
operating systems.

It also weakens their "Enterprise Class" claims. 64bit Windows has
very few production machine hours compared to Linux (for example).

64 bit Windows is hardly what I would call "Enterprise Ready", but
there are plenty of alternatives that are.

Cheers,
Rupert
 
Going to 64 bits will be trivial compared to going to 64 way for Microsoft.
Joe Seigh

why do you say that? Maybe there will be particular issues for
applications to make use of all those CPUs but I don't see why it would
be such a big deal for the OS kernel scheduler.

IMHO, 64-bits is much harder considering the Win32 API.
 
|>
|> why do you say that? Maybe there will be particular issues for
|> applications to make use of all those CPUs but I don't see why it would
|> be such a big deal for the OS kernel scheduler.

Hmm. Do you manage any large CPU-count SMP systems?

|> IMHO, 64-bits is much harder considering the Win32 API.

No, Joe Seigh is right.


Regards,
Nick Maclaren.
 
Pleasant said:
why do you say that? Maybe there will be particular issues for
applications to make use of all those CPUs but I don't see why it would
be such a big deal for the OS kernel scheduler.

IMHO, 64-bits is much harder considering the Win32 API.

Scalability primarily judging from the experience of the other OS vendors.

Joe Seigh
 
Rupert said:
That just means MS loses market & mindshare to other more capable
operating systems.

It also weakens their "Enterprise Class" claims. 64bit Windows has
very few production machine hours compared to Linux (for example).

64 bit Windows is hardly what I would call "Enterprise Ready", but
there are plenty of alternatives that are.

Microsoft should have just released Windows 64, despite not having enough
optimized drivers for it. There's nothing like a shipping product to drive
driver development.

Yousuf Khan
 
Pleasant said:
why do you say that? Maybe there will be particular issues for
applications to make use of all those CPUs but I don't see why it
would be such a big deal for the OS kernel scheduler.

The scheduler would have to be quite non-trivial for such a large-scale
machine. The scheduler would have to not only take into account the number
of processors involved (that's the easy part), but it will have to take into
account metrics like how much latency there is between processors talking to
each other, how much latency there is in processors talking to various
sections of memory, etc.

For example with the processor-processor latencies, Hypertransport itself
creates a NUMA architecture, a relatively quick NUMA architecture that can
almost be treated as SMP, but NUMA nonetheless. Then if it goes above 4 or 8
processors, a second level of interconnect will need to be introduced which
might make things even slower. So certain groups of processors could talk to
each other at the highest speed, through Hypertransport, those groups would
likely be located on the same system boards. Then processors within each
group would have to talk to processors in other groups through a different
interconnect. So you have at least two levels of NUMA to take into account,
and get the timings right, etc.

Yousuf Khan
 
Pleasant Thrip said:
Microsoft.

why do you say that?

Er, NUMA. Processor affinity in order to leverage local cache contents.
Synchronization mechanisms that scale easily to 4 - 8 processors but fall
flat on their face at 64. Just for a start.

Maybe there will be particular issues for
applications to make use of all those CPUs but I don't see why it would
be such a big deal for the OS kernel scheduler.

IMHO, 64-bits is much harder considering the Win32 API.

Considering that they had a beta 64-bit version out in the field on Alpha 5
years ago, I'd suspect that they had that pretty well under control by now.

- bill
 
Bill said:
Considering that they had a beta 64-bit version out in the field on
Alpha 5 years ago, I'd suspect that they had that pretty well under
control by now.

They may have had an OS in 64-bit for Alpha 5 years ago, but did they have
any applications or drivers? That seems to be where they're stumbling right
now: on a bit of application support issues, and a lot of driver support
issues.

Yousuf Khan
 
IMHO, 64-bits is much harder considering the Win32 API.
Considering that they had a beta 64-bit version out in the field on Alpha 5
years ago, I'd suspect that they had that pretty well under control by now.

- bill

Did they? I thought WinNT on Alpha was really and truly 32 bit, even
though it was on a 64b processor. Besides, they threw all that code out
and made so many thousands of changes since then that there probably
isn't any legacy of Alpha's NT left in XP.

Alex
 
Alex Johnson said:
Did they?

Yes.

I thought WinNT on Alpha was really and truly 32 bit, even
though it was on a 64b processor.

The versions that actually shipped were. The field-test version in 1999
that I referred to (being tested concurrently with its 32-bit Alpha
counterpart, until both were canned in August of that year courtesy of Curly
Capellas, bless his incompetent shiny head) was 64 bits.
Besides, they threw all that code out

Well, not exactly: rumor has it that it's *still* running on Alphas inside
Microsoft (Dave Cutler reportedly isn't a great Itanic fan), and has even
been kept moderately up to date in the interim.
and made so many thousands of changes since then that there probably
isn't any legacy of Alpha's NT left in XP.

That would be stupid even for Microsoft: it was the first 64-bit Windows
version they had working, and they continued to use it for 64-bit Windows
development not only until they were able to get usable Itanic systems
(i.e., McKinleys) but significantly thereafter. To suggest that at some
point they then scrapped it and started over is, well, ridiculous.

Since Windows XP code is largely Win2K code underneath, and since 64-bit
Win2K was developed on Alpha, a large percentage of the 64-bit code in
current 64-bit Windows products almost certainly originated with the 64-bit
Alpha version.

- bill
 
Yousuf Khan said:
They may have had an OS in 64-bit for Alpha 5 years ago, but did they have
any applications or drivers?

Who cares (at least in the current discussion context)? The point was that
they had the API worked out sufficiently back in 1999 to give it to outside
developers.

- bill
 
How can you doubt it? ;-)


--
Ed Light

Smiley :-/
MS Smiley :-\

Send spam to the FTC at
(e-mail address removed)
Thanks, robots.
 
George said:
[SNIP]
Doesn't sound like Chang believes that Microsoft is trying all that hard to
build a 64-bit OS. It's getting something out to show that it isn't behind
the times.


... or that we'll get it when Intel is good and ready for us to have it.:-(

That just means MS loses market & mindshare to other more capable
operating systems.

True but those people are among the most arrogant on the planet. They
think they can get away with it and maybe they can. There are signs that
AMD64 supply is tightening up and prices are staying relatively high. It's
probable that the bottom line is that M$ figures AMD64 volume can never
reach what they call "volume".
It also weakens their "Enterprise Class" claims. 64bit Windows has
very few production machine hours compared to Linux (for example).

64 bit Windows is hardly what I would call "Enterprise Ready", but
there are plenty of alternatives that are.

Just a couple of months ago, Oracle and IBM announced availability of
Oracle and DB2 x86-64 for Linux - as in you can buy it. As part of the
same announcement, they offered a "development version" for Windows
Server... development because the OS is err, late and itself in
"development"... Beta I suppose. That seems like a pretty strong prod to
me and yet, still, M$ just dawdles along.<shrug>

Rgds, George Macdonald

"Just because they're paranoid doesn't mean you're not psychotic" - Who, me??
 
Bill Todd said:
Who cares (at least in the current discussion context)? The point
was that they had the API worked out sufficiently back in 1999 to
give it to outside developers.

I thought the point was getting a working 64-bit Microsoft system? That
would mean not just the OS, but also the apps and drivers. If it's just the
OS, then Microsoft is already done, the OS is already ready for Opteron. But
Microsoft has said that the only thing holding them back from releasing the
OS is the drivers, and a few apps which might do things and get away with in
the 32-bit OS which they won't be allowed to get away with in 64-bit.

Yousuf Khan
 
Back
Top