Prescott with 64-bit extensions coming in June

[Judd]

http://www.infoworld.com/article/04/05/13/HNprescott_1.html

So what's all expected for Q3 now?
- 64-bit extensions
- 1066 MHz Bus
- 3.73 MHz Prescott/Dothan?
- Grantsdale/Alderwood chipset with PCI Express

How about on-board Firewire 800 and SATA-II?

Still no XP2...

 

[Yousuf Khan]

http://www.infoworld.com/article/04/05/13/HNprescott_1.html

So what's all expected for Q3 now?
- 64-bit extensions
- 1066 MHz Bus
- 3.73 MHz Prescott/Dothan?
- Grantsdale/Alderwood chipset with PCI Express

How about on-board Firewire 800 and SATA-II?

Still no XP2...

That's funny, according to the article:

<quote>
Prescott supports the NX (no execute) feature that will prevent worms and
viruses from executing dangerous code through the exploitation of buffer
overflows, Otellini said during a Webcast of the event. Advanced Micro
Devices Inc.'s Athlon 64 and Opteron processors also come with this feature,
which requires software support from Microsoft Corp.'s Windows XP Service
Pack 2 expected later this year.
</quote>

As of the original release of the EM64T documentation, Intel didn't yet
support the NX bit. That was one of the most glaring ommisions from an
otherwise perfect copy job of the AMD64 specs. Has this oversight now been
corrected, or is the article writer just assuming things here?

Yousuf Khan

 

[Judd]

That's funny, according to the article:

<quote>
Prescott supports the NX (no execute) feature that will prevent worms and
viruses from executing dangerous code through the exploitation of buffer
overflows, Otellini said during a Webcast of the event. Advanced Micro
Devices Inc.'s Athlon 64 and Opteron processors also come with this feature,
which requires software support from Microsoft Corp.'s Windows XP Service
Pack 2 expected later this year.
</quote>

As of the original release of the EM64T documentation, Intel didn't yet
support the NX bit. That was one of the most glaring ommisions from an
otherwise perfect copy job of the AMD64 specs. Has this oversight now been
corrected, or is the article writer just assuming things here?

I was hoping you saw that. I assumed Otellini told them it supported NX and
that, yes, the new Prescott steppings would support the addition (couldn't
have been that difficult to implement). I assume Intel doesn't want to get
bullet pointed out on a feature that analyst and ragazines would promote as
the next best thing since sliced bread. They've already given in this far.
Still, there seems to be a rather large number of near future offerings from
Intel that really weren't expected at this point. This might be a much more
interesting year for PCs than previously anticipated.

 

[Yousuf Khan]

I was hoping you saw that. I assumed Otellini told them it supported
NX and that, yes, the new Prescott steppings would support the
addition (couldn't have been that difficult to implement). I assume
Intel doesn't want to get bullet pointed out on a feature that
analyst and ragazines would promote as the next best thing since
sliced bread. They've already given in this far. Still, there seems
to be a rather large number of near future offerings from Intel that
really weren't expected at this point. This might be a much more
interesting year for PCs than previously anticipated.

Yeah, it can't be that difficult to implement. Just wonder if Intel's
documentation writers have updated their PDF's yet. I'm going to have to
maintain a CVS repository of Intel PDFs at this rate.

Yousuf Khan

 

[Carlo Razzeto]

Judd wrote:
Yeah, it can't be that difficult to implement. Just wonder if Intel's
documentation writers have updated their PDF's yet. I'm going to have to
maintain a CVS repository of Intel PDFs at this rate.

Yousuf Khan

Are you sure that will help? Having delt with CVS quite a bit at work I know
how easy it is to make things a mess in CVS, especially when you need to
start branching as much as you would with Intels road maps as of late.

Carlo

 

[Grumble]

So what's all expected for Q3 now?
- 64-bit extensions
- 1066 MHz Bus
- 3.73 MHz Prescott/Dothan?

How can you even think that Dothan would scale to > 3 GHz in 2004?

 

[Grumble]

I was hoping you saw that. I assumed Otellini told them it
supported NX and that, yes, the new Prescott steppings would
support the addition (couldn't have been that difficult to
implement). I assume Intel doesn't want to get bullet pointed out
on a feature that analyst and ragazines would promote as the next
best thing since sliced bread. They've already given in this far.

Apparently, the OpenBSD crew was given IA32e processors to play with,
which did not include the NX bit.

http://www.openbsd.org/35.html

Note: The upcoming Intel "ia32e" AMD64-compatible cpus have
also been tested, and work, even though they lack the NX bit.

As you say, maybe it's been corrected with a new core stepping.

 

[Alex Johnson]

As of the original release of the EM64T documentation, Intel didn't yet
support the NX bit. That was one of the most glaring ommisions from an
otherwise perfect copy job of the AMD64 specs. Has this oversight now been
corrected, or is the article writer just assuming things here?

My understanding (I am not near to this at all) is that the feature was
always there but not adequately validated. Harken back to the
Willamette and Northwood and Hyperthreading. Intel does put in optional
features it wants and if they can't be tested by the deadline, "ship it
anyway, we want a product on the market." In the next stepping they'll
have had time to test thoroughly and will enable it. Remember the other
thread: "Intel follows the margin"

Alex

 

[Yousuf Khan]

My understanding (I am not near to this at all) is that the feature
was always there but not adequately validated. Harken back to the
Willamette and Northwood and Hyperthreading. Intel does put in
optional features it wants and if they can't be tested by the
deadline, "ship it anyway, we want a product on the market." In the
next stepping they'll have had time to test thoroughly and will
enable it. Remember the other thread: "Intel follows the margin"

Yeah, but the whole EM64T stuff is not validated yet. It's an announcement
about a future product that hasn't come out yet, and until recently they
hadn't even put out an official date of arrival yet.

What's the point in shipping something that is half-finished, when the
fully-finished thing (AMD's product) already exists? It's like saying buy
our car, even though it doesn't have windshield wipers. Sure it will work
fine most of the time, except on the odd days when it rains.

Yousuf Khan

 

[Yousuf Khan]

http://www.infoworld.com/article/04/05/13/HNprescott_1.html

So what's all expected for Q3 now?
- 3.73 MHz Prescott/Dothan?

Oh, and no, Dothan obviously wouldn't run at 3.73 Ghz, but it might touch
2.2 Ghz by Q3.

Yousuf Khan

 

[Judd]

Oh, and no, Dothan obviously wouldn't run at 3.73 Ghz, but it might touch
2.2 Ghz by Q3.

So one would think, but recall the article I posted earlier about a 3.73
Prescott CPU with a Dothan core? My guess is that the article is wrong and
it's not Dothan, but it did raise an eyebrow

 

[Judd]

Yeah, but the whole EM64T stuff is not validated yet. It's an announcement
about a future product that hasn't come out yet, and until recently they
hadn't even put out an official date of arrival yet.

What's the point in shipping something that is half-finished, when the
fully-finished thing (AMD's product) already exists? It's like saying buy
our car, even though it doesn't have windshield wipers. Sure it will work
fine most of the time, except on the odd days when it rains.

The point is to make $$$. Happens all the time. Software companies ship
products 3/4 finished all the time. Heck, Microsoft does it too! Engineers
must be paid you know. My guess is that they have finally validated all of
those parts (64 bit extensions and Nx bit) and are ready to ship something
besides vaporware. On a site note, they are likely getting their tools
ready for this transition as well (icc, itune, etc.). It's all a part of
the process. Since almost nothing has been built for 64-bit on the x86
side, it doesn't matter anyway! Until MS builds it's OS for 64-bit ext and
adds a VisualStudio suite with the capabilities, rapid adoption is a long
ways away.

 

[Eric]

That's funny, according to the article:

<quote>
Prescott supports the NX (no execute) feature that will prevent worms and
viruses from executing dangerous code through the exploitation of buffer
overflows, Otellini said during a Webcast of the event. Advanced Micro
Devices Inc.'s Athlon 64 and Opteron processors also come with this
feature, which requires software support from Microsoft Corp.'s Windows XP
Service Pack 2 expected later this year.
</quote>

As of the original release of the EM64T documentation, Intel didn't yet
support the NX bit. That was one of the most glaring ommisions from an
otherwise perfect copy job of the AMD64 specs. Has this oversight now been
corrected, or is the article writer just assuming things here?

Yousuf Khan

So, do you mean to say that Intel & AMD are fixing Windows' Virus problems
in the CPU hardware? I thought AMD and Intel were smarter than that. This
reminds me of Compaq's attempt to fix an application bug by modifying their
BIOS.
Eric

 

[Eric]

That's funny, according to the article:

<quote>
Prescott supports the NX (no execute) feature that will prevent worms and
viruses from executing dangerous code through the exploitation of buffer
overflows, Otellini said during a Webcast of the event. Advanced Micro
Devices Inc.'s Athlon 64 and Opteron processors also come with this
feature, which requires software support from Microsoft Corp.'s Windows XP
Service Pack 2 expected later this year.
</quote>

As of the original release of the EM64T documentation, Intel didn't yet
support the NX bit. That was one of the most glaring ommisions from an
otherwise perfect copy job of the AMD64 specs. Has this oversight now been
corrected, or is the article writer just assuming things here?

Yousuf Khan

After reading up on the NX bit I have this thought: Isnt that what
segmentation is for? Cripe, start using Data segments and the buffer
overflows will just fault.
Eric

 

[Yousuf Khan]

So one would think, but recall the article I posted earlier about a
3.73 Prescott CPU with a Dothan core? My guess is that the article
is wrong and it's not Dothan, but it did raise an eyebrow

Which article? The one in the link above doesn't mention anything about the
Dothan.

Yousuf Khan

 

[Yousuf Khan]

So, do you mean to say that Intel & AMD are fixing Windows' Virus
problems in the CPU hardware? I thought AMD and Intel were smarter
than that. This reminds me of Compaq's attempt to fix an application
bug by modifying their BIOS.
Eric

Well, we're not exactly sure about Intel yet, but AMD is attempting to
nullify one of the more common methods worms use against Windows. It's a
relatively easy solution too, mostly transparent to the applications
software; they just mark certain memory pages as data-only, code cannot be
executed from those pages. That way if a buffer overflow exploit is employed
against that program, the malicious code inserted sits in non-executable
memory and goes nowhere.

When Intel copied AMD's 64-bit specifications, it is assumed that they
copied an older version of that spec, and that some of the more recent
features that AMD put in, didn't get put in by Intel. The NX (no execute)
bit in the page table was one of the ones not mentioned in Intel's
documentation.

Yousuf Khan

 

[Yousuf Khan]

After reading up on the NX bit I have this thought: Isnt that what
segmentation is for? Cripe, start using Data segments and the buffer
overflows will just fault.
Eric

Yeah, that was my original thought too, but for some reason segmentation has
gotten a nasty reputation from its Real mode days, and OS companies from the
beginning of the 32-bit era couldn't wait to stop using it, despite the fact
that it had all of those advantages. Hell it could've even delayed the
introduction of 64-bit computing another couple of years, if applications
started to use multiple 4GB data segments. But that's only for historical
interest now, segmentation has been removed from 64-bit mode.

Yousuf Khan

 

[Yousuf Khan]

Are you sure that will help? Having delt with CVS quite a bit at work
I know how easy it is to make things a mess in CVS, especially when
you need to start branching as much as you would with Intels road
maps as of late.

I just downloaded those EM64T PDFs again -- they are exactly the same PDFs
as the ones they released in February. They even still call them IA32E
inside. So the PDFs haven't be updated. (Of course, that's not entirely
surprising, documentation lags behind implementation in most places I've
worked too.)

Yousuf Khan

 

[Carlo Razzeto]

I just downloaded those EM64T PDFs again -- they are exactly the same PDFs
as the ones they released in February. They even still call them IA32E
inside. So the PDFs haven't be updated. (Of course, that's not entirely
surprising, documentation lags behind implementation in most places I've
worked too.)

Yousuf Khan

You mean you actually had documentation at the places you worked? Must be
nice, where I am it's do this project and figure out where you need to pull
the data from on your own. Very annoying, getting easier since I've been
with this company for a few months now, I'm actually getting familiar with
some of their DBs, but it's still just a huge mess. Very fustrating
experience.

Carlo

 

[Tony Hill]

So, do you mean to say that Intel & AMD are fixing Windows' Virus problems
in the CPU hardware? I thought AMD and Intel were smarter than that. This
reminds me of Compaq's attempt to fix an application bug by modifying their
BIOS.

It's not exactly that, it's more adding in a feature to the processor
that SHOULD have been there ages ago. Such functionality is common in
many non-x86 chips, but was implemented in a rather bass-ackwards way
in current x86 chips.

It's not going to fix the Windows virus problem. In fact, it won't
affect actual *virus* programs at all, though it should really help
with *worm* programs, the difference between the two is often lost in
the mainstream media these days. It offers Microsoft (and others)
another tool to try and mitigate (though not eliminate) security
risks.

FWIW while it is popular to attack Microsoft for their rather, umm..
questionable decisions regarding security, they do seem to have seen
the light of day. Take a look at WinXP SP2 sometime, and you'll see
that they are making a LOT of changes that should significantly
improve security. I'm not just talking about patching bugs here or
anything, this is a fundamental change in philosophy.