Intel ready to put out desktop 64-bit now

[YKhan]

Intel is saying that it will put the 64-bit extensions into all of its
desktop processors by 2005, including Celeron. This 2005 date coincides
with the date that Microsoft is expected to release Windows for x64! Of
course that's just a coincidence, Microsoft couldn't possibly be giving
Intel breathing room to catch up, afterall?

Intel to put 64 bits in desktops in 2005 | CNET News.com
http://news.com.com/Intel+to+put+64+bits+in+desktops+in+2005/2100-1006_3-5482345.html?tag=html.alert

Most interesting part of announcement is that it'll be offering it even
in Celeron. Which would mean that the 64-bit capabilities in S754
Sempron would need to be turned back on to compete, of course.
Yousuf Khan

 

[George Macdonald]

Intel is saying that it will put the 64-bit extensions into all of its
desktop processors by 2005, including Celeron. This 2005 date coincides
with the date that Microsoft is expected to release Windows for x64! Of
course that's just a coincidence, Microsoft couldn't possibly be giving
Intel breathing room to catch up, afterall?

Intel to put 64 bits in desktops in 2005 | CNET News.com
http://news.com.com/Intel+to+put+64+bits+in+desktops+in+2005/2100-1006_3-5482345.html?tag=html.alert

Most interesting part of announcement is that it'll be offering it even
in Celeron. Which would mean that the 64-bit capabilities in S754
Sempron would need to be turned back on to compete, of course.
Yousuf Khan

Do they have a new desktop chipset coming which will allow addressing

4GB?... not the kind of thing your average analyst would notice of course

but current offerings fall short.

Rgds, George Macdonald

"Just because they're paranoid doesn't mean you're not psychotic" - Who, me??

 

[YKhan]

Do they have a new desktop chipset coming which will allow addressing course
but current offerings fall short.

I don't know, wouldn't their existing chipsets have this capability? I
mean Pentiums have had the PAE feature since the PentiumPro days, which
allows them to address at least upto 36 bits of address space through
segments.

Yousuf Khan

 

[John Savard]

Intel is saying that it will put the 64-bit extensions into all of its
desktop processors by 2005, including Celeron.

This is good, but what they should *really* be doing is including the
ability to decode Itanium-format instructions *too*. Then, once people
discover that these allow the chip to work even *better*, this
proprietary architecture becomes the new standard, and AMD is locked
out!

Or would this get them into antitrust trouble...

John Savard
http://home.ecn.ab.ca/~jsavard/index.html

 

[Tony Hill]

Intel is saying that it will put the 64-bit extensions into all of its
desktop processors by 2005, including Celeron. This 2005 date coincides
with the date that Microsoft is expected to release Windows for x64! Of
course that's just a coincidence, Microsoft couldn't possibly be giving
Intel breathing room to catch up, afterall?

I still doubt it. Microsoft has never needed any excuse to be
horribly late in delivering a product before, why start now?

More likely it's the other way around, ie Intel will release 64-bit
desktop chips once MS finally gets off their collective asses and
finally gets WinXP for x64 out the door.

Of course, I'm still expecting the release date of 64-bit Windows to
be pushed back at least once more... after all, they're only running a
year and a half late now, why not just make it a full 2 years late?

Intel to put 64 bits in desktops in 2005 | CNET News.com
http://news.com.com/Intel+to+put+64+bits+in+desktops+in+2005/2100-1006_3-5482345.html?tag=html.alert

Most interesting part of announcement is that it'll be offering it even
in Celeron. Which would mean that the 64-bit capabilities in S754
Sempron would need to be turned back on to compete, of course.

Queue corresponding AMD announcement in 5... 4... 3... :>

 

[George Macdonald]

I don't know, wouldn't their existing chipsets have this capability? I
mean Pentiums have had the PAE feature since the PentiumPro days, which
allows them to address at least upto 36 bits of address space through
segments.

Nope, so far, no desktop chipset, including i925, has had 36 address lines
on the FSB. You have to go to a workstation chipset, like i7505(?) to get
that. Seems like AMD64 has been a real wrench in the works for Intel's
market segmentation strategy.ô_ô

Rgds, George Macdonald

"Just because they're paranoid doesn't mean you're not psychotic" - Who, me??

 

[Tony Hill]

I don't know, wouldn't their existing chipsets have this capability? I
mean Pentiums have had the PAE feature since the PentiumPro days, which
allows them to address at least upto 36 bits of address space through
segments.

The E7xxx series of chipsets, targeting servers and workstations, can
mostly handle up to 16GB of memory. However their desktop chips,
including the new i915 and i925 are limited to only 4GB of memory.

Intel does have some new chipsets planned though which should arrive
around the same time as the 64-bit desktop chips. I haven't heard how
much memory they will support though.

Here's a link to the chipset comparison chart:

http://www.intel.com/design/chipsets/linecard/svr_wkstn.htm

 

[Tony Hill]

This is good, but what they should *really* be doing is including the
ability to decode Itanium-format instructions *too*.

And just how do you propose to do this?! Are you going to just hide a
300mm^2+ Itanium core in the corner of the chip or something? And
that's not even trying to touch on all the complexities of running two
completely unrelated processors on the same piece of silicon.
Definitely NOT an easy task.

Then, once people
discover that these allow the chip to work even *better*, this
proprietary architecture becomes the new standard, and AMD is locked
out!

The Itanium2 with only 1.5MB of cache is easily outperformed by both
Intel's Pentium4 and AMD's Athlon64/Opteron chips with 1MB of cache.
If you want to make the Itanium perform well you need a LOT of cache,
this is why Itaniums come with 6MB or even 9MB of cache. That cache
doesn't come for free though, the Itanium2 chips with said cache have
a HUGE number of transistors (pushing half a billion for cache alone)
and significantly larger die sizes than the P4 of Athlon64. Part of
their performance also comes from the fact that their system bus is
twice as wide as that of the P4's, again adding to the cost.

There are very good reasons why the Itanium chips that offer high
performance cost $5000 and the motherboards cost $2000.

Or would this get them into antitrust trouble...

Or it would destroy their profits and AMD would take their winnings to
the bank.

 

[YKhan]

This is good, but what they should *really* be doing is including the
ability to decode Itanium-format instructions *too*. Then, once people
discover that these allow the chip to work even *better*, this
proprietary architecture becomes the new standard, and AMD is locked
out!

I know you keep coming back to that, but how exactly do you expect to
hide an Itanium core inside a Pentium core, let alone a Celeron core?
The size increase in the core will kill its profits.

Worse still how do you expect to make an Itanium core, which currently
doesn't go much over 1.5Ghz run at the pace of a Pentium/Celeron which
are all well over 2.0Ghz?

Or would this get them into antitrust trouble...

No anti-trust problems I can think of there, Intel is perfectly within
its rights to do such a thing if it wishes. However, it is already too
late in the game to get the popularity of the IA64 codebase to the
levels where it would pose a threat to the AMD64/EM64T codebase.

If anything AMD would be praying to god that Intel tried something so
stupid. It would automatically make its own chips the outright winners
here. They'd be cheaper to produce and perform better in most cases
too.

Yousuf Khan

 

[YKhan]

I still doubt it. Microsoft has never needed any excuse to be
horribly late in delivering a product before, why start now?

Because I really don't believe there were ever any showstoppers to
porting to AMD64. It is probably the most trivial port in their history
they have ever done.

During the entire time you heard all kinds of showstopper excuses
(brand new filesystems, user interfaces, etc.) about why Longhorn was
being delayed, but there was a decided lack of detail about why XP64
was being delayed. The only excuse I heard was originally about
programmers being diverted to XP32's SP2 introduction -- but that
excuse ceased being an excuse months ago.

It was one thing to have Linux come out for x64 exactly on the same day
as the AMD processor itself came out, and several years ahead of
Windows. But it is quite another thing to have Solaris 10 come out at
least six months ahead of Windows too, despite having started the port
several years later. Remember Windows had access to prerelease Opteron
hardware and simulators, years before release. On the other hand,
Solaris 10's future on x86 hardware was in doubt until recently and
only once its future was secured did the port go through in earnest, at
which point it also finished its port months earlier.

Of course, I'm still expecting the release date of 64-bit Windows to
be pushed back at least once more... after all, they're only running a
year and a half late now, why not just make it a full 2 years late?

Well, actually by the point they are now scheduled to release it
(May'ish) that will be about 2 years after the original release of the
Opteron, so they've made their even-steven target.

Yousuf Khan