Whatever happened to x86-64?

[YKhan]

Adorable little Ed Stroglio rant:

Hatching Eggs . . .
"Remember x86-64? The stick AMD was supposed to beat Intel to death
with?

Then Intel got themselves the same stick, and the would-be beaters
found something else to talk about. Yes, AMD is doing better now, but
not because of x86-64. Indeed, AMD's increased fortunes have come well
after Intel switched.

Microsoft came out with a Windows XP for x86-64, but the world hasn't
exactly stampeded to get it."
http://www.overclockers.com/tips00910/

I always thought the reason nobody flocked to XP x64 was because
Microsoft delayed and delayed so much to let Intel catch up that people
lost interest in it? I mean you know Microsoft was delaying here, even
Solaris 10 came out for x64 before Windows did, despite starting work
on it 2 years later.

Yousuf Khan

 

[George Macdonald]

Adorable little Ed Stroglio rant:

Hatching Eggs . . .
"Remember x86-64? The stick AMD was supposed to beat Intel to death
with?

Then Intel got themselves the same stick, and the would-be beaters
found something else to talk about. Yes, AMD is doing better now, but
not because of x86-64. Indeed, AMD's increased fortunes have come well
after Intel switched.

Microsoft came out with a Windows XP for x86-64, but the world hasn't
exactly stampeded to get it."
http://www.overclockers.com/tips00910/

"How many of you bought CPUs back then at a premium primarily because of
x86-64?"?? I got my Athlon64 ~15months ago and I don't recall paying much
of a "premium" for it. Hmm, must have been a slow day chez Ed!

I always thought the reason nobody flocked to XP x64 was because
Microsoft delayed and delayed so much to let Intel catch up that people
lost interest in it? I mean you know Microsoft was delaying here, even
Solaris 10 came out for x64 before Windows did, despite starting work
on it 2 years later.

I'm still not 100% sure about that. Its also been suggested that M$ was
just somewhat less than competent. They also knew that AMD's supply was
not going to carry the market so apply a bit of Parkinson's Law as well and
a new-found(?) desire to provide "security" and that accounts for a fair
part of it. As Tony has remarked, it *did* take them a while to get WinXP
SP2 out the door.

 

[Keith]

"How many of you bought CPUs back then at a premium primarily because of
x86-64?"?? I got my Athlon64 ~15months ago and I don't recall paying much
of a "premium" for it. Hmm, must have been a slow day chez Ed!

I did. I bought my socket-940 Opterpon about 20months ago and paid a
rather large premium for it (including the registered memory). I decided
that I was ready for a new system and that since I was still using the
(then) four-year-old K6-III I was goign to skip the K7 generation entirely
and move on to AMD64. I'll quite likely have this system at least another
two years, perhaps far longer.

I'm still not 100% sure about that. Its also been suggested that M$ was
just somewhat less than competent. They also knew that AMD's supply was
not going to carry the market so apply a bit of Parkinson's Law as well
and a new-found(?) desire to provide "security" and that accounts for a
fair part of it. As Tony has remarked, it *did* take them a while to
get WinXP SP2 out the door.

....a small part of the reason I moved onto Linux.

 

[EdG]

Adorable little Ed Stroglio rant:

Hatching Eggs . . .
"Remember x86-64? The stick AMD was supposed to beat Intel to death
with?

Then Intel got themselves the same stick, and the would-be beaters
found something else to talk about. Yes, AMD is doing better now, but
not because of x86-64. Indeed, AMD's increased fortunes have come well
after Intel switched.

Microsoft came out with a Windows XP for x86-64, but the world hasn't
exactly stampeded to get it."
http://www.overclockers.com/tips00910/

I always thought the reason nobody flocked to XP x64 was because
Microsoft delayed and delayed so much to let Intel catch up that people
lost interest in it? I mean you know Microsoft was delaying here, even
Solaris 10 came out for x64 before Windows did, despite starting work
on it 2 years later.

Yousuf Khan

that Ed is ... I don't know,... he must own lots of Intel stock or sold
AMD at the wrong time and lost his ass on it, I've never seen anyone on
the net try and twist so much negative spin on AMD.

Well .... I didn't buy my two AMD64's for their 64-bit, it was basically
icing on the cake, I always upgrade and use one of the older ones for a
server and thought... ya maybe in 3 years this AMD will be my home
server anyway and if it runs 64bit that can't be a bad thing, and I
thought the price/performance was fair.

EdG

 

[EdG]

I always thought the reason nobody flocked to XP x64 was because
Microsoft delayed and delayed so much to let Intel catch up that people
lost interest in it? I mean you know Microsoft was delaying here, even
Solaris 10 came out for x64 before Windows did, despite starting work
on it 2 years later.

Yousuf Khan

I tried XP-64 beta and the final when it came out... long story
short.... no HP printer drivers for me, no real working anti-virus
software at the time to choose from, my video, sound, LAN drivers were
all still in beta versions, I let someone else do the debugging. ;p

 

[Yousuf Khan]

I'm still not 100% sure about that. Its also been suggested that M$ was
just somewhat less than competent. They also knew that AMD's supply was
not going to carry the market so apply a bit of Parkinson's Law as well and
a new-found(?) desire to provide "security" and that accounts for a fair
part of it. As Tony has remarked, it *did* take them a while to get WinXP
SP2 out the door.

It's an ironic world we live in, when people find you more acceptable
because you're incompetant.

There were good reasons for it to delay introduction of SP2, as they
were trying to fix a leaky boat. At that time security on Windows seemed
like swiss cheese. However, they had x64 finished much earlier, even
though they kept releasing release candidates. They were hoping for
device drivers to be released, but nobody would release device drivers
until Microsoft released the OS. They could've released x64 and still
released SP2 for 32-bit.

Really the most telling sign that they were delaying was the fact that
Sun Microsystems started development work on their Solaris 10 x64 about
two years after Microsoft, and still released it six months before
Microsoft! One could say that Sun has years of experience with 64-bit
operating systems, and Microsoft doesn't. But Microsoft already had
64-bit source code for Windows in their Itanium port, and other than
low-level assembly language differences, most of it would be the same
between the two architectures, just a recompile away.

If Windows x64 had been released near to when AMD had released their
hardware, even if AMD didn't sell as many chips as Intel, the fact that
a released OS was sitting there it would've been gathering device driver
support ready for the time when Intel came on board. By the time Intel
was ready with 64-bit, you'd have a mature operating system already with
lots of device driver support. And people coming in with Intel would
have had much easier time. AMD users could've acted as the beta testers
for Intel users.

Yousuf Khan

 

[nobody]

Adorable little Ed Stroglio rant:

Hatching Eggs . . .
"Remember x86-64? The stick AMD was supposed to beat Intel to death
with?

Then Intel got themselves the same stick, and the would-be beaters
found something else to talk about. Yes, AMD is doing better now, but
not because of x86-64. Indeed, AMD's increased fortunes have come well
after Intel switched.

Microsoft came out with a Windows XP for x86-64, but the world hasn't
exactly stampeded to get it."
http://www.overclockers.com/tips00910/

I always thought the reason nobody flocked to XP x64 was because
Microsoft delayed and delayed so much to let Intel catch up that people
lost interest in it? I mean you know Microsoft was delaying here, even
Solaris 10 came out for x64 before Windows did, despite starting work
on it 2 years later.

Yousuf Khan

My Opteron system is close to 2 year old, and it never ran a single
line of 64 bit code. Back then when I built it, the reason behind it
was a dual processor system for just over a half of equal Xeon rig
price. Will it ever do 64? Only when there will be demand to do some
work that can't possibly be done under 32 bit OS. So far, never heard
of any requirement of 64 bit code in the section of software market in
which I make a living.

NNN

 

[Robert Redelmeier]

Well .... I didn't buy my two AMD64's for their 64-bit,
it was basically icing on the cake,

Sorta like the 386s? I consider the uptake of AMD64 to be
nothing short of phenomenal in comparison. In a far more
entrenched and less geeky market, we have day one OSes and a
remarkable amount of support. Even from the known laggard, MS.

IIRC the 386 had to wait 5 years for runnable OSes (IIRC, OS/2
v2 and non-beta Linux). Apps followed, and will this time too.

The case for 64bit over 32b is less compelling than 32 over 16.
So uptake will be somewhat slower in the middle part of the
life-cycle curve. But the start has been very quick.

-- Robert

 

[Evgenij Barsukov]

My Opteron system is close to 2 year old, and it never ran a single
line of 64 bit code. Back then when I built it, the reason behind it
was a dual processor system for just over a half of equal Xeon rig
price. Will it ever do 64? Only when there will be demand to do some
work that can't possibly be done under 32 bit OS. So far, never heard
of any requirement of 64 bit code in the section of software market in
which I make a living.

There are still very few softw. that support 64, not to mention
_require_ it. I was surprized recently that Half Life II was released
in 64 bit version. However I am not going to run and by win xp-64 just
because of that.

Now if you need to address 8 GB of memory for some purposes (DB
server?), 64 bit is the only game in town.

Regards,
Yevgen

 

[Kai Harrekilde-Petersen]

There are still very few softw. that support 64, not to mention
_require_ it. I was surprized recently that Half Life II was released
in 64 bit version. However I am not going to run and by win xp-64 just
because of that.

Now if you need to address 8 GB of memory for some purposes (DB
server?), 64 bit is the only game in town.

Yup. I did some chip-level gatelevel simulations on an 98Mtransistor
chip, and just the SDF file with timing information is 8GB. In-memory
image size was ~4GB, so I had to go 64bit.

Starting the simulation up (annotating timing to all nodes etc) took
around 5 minutes on our Sun v20z dual-opty - I gave up trying after I
had waited 1 hour on our old SunBlade 1000.


Kai

 

[nobody]

Now if you need to address 8 GB of memory for some purposes (DB
server?), 64 bit is the only game in town.

I might've lost track of releases, but IIRC Server'03 has not been
released for AMD64 yet (I mean production, not betas and previews).
SQL Server 2000 (the current corporate work horse) will never go 64
bit. SQL Server 2005, while released, is still exotic. Besides,
lacking Windows Server to run on, the 64 bit version may run only as
developer edition on WinXP64, and in this environment you hardly ever
need more than 4 GB - I've never felt constrained by 1 GB yet, with
all multiple instances I'm usually running. Maybe in Solaris/Oracle
world you can get advantage of extra memory, but, not being a
professional in that area, I can't comment on it.

NNN

 

[David Kanter]

Now if you need to address 8 GB of memory for some purposes (DB

I might've lost track of releases, but IIRC Server'03 has not been
released for AMD64 yet (I mean production, not betas and previews).

It has actually. I have a full copy sitting next to me. Check MS's
website as well, it is no longer in beta. Some people have said that
MS just waited till Intel had x86-64 products out, but I rather doubt
that's the only reason. Certainly with AMD's volume alone, there was
less impetus to do so, but MS does also take a while to port and QA
their products.

SQL Server 2000 (the current corporate work horse) will never go 64
bit. SQL Server 2005, while released, is still exotic. Besides,
lacking Windows Server to run on, the 64 bit version may run only as
developer edition on WinXP64, and in this environment you hardly ever
need more than 4 GB - I've never felt constrained by 1 GB yet, with
all multiple instances I'm usually running. Maybe in Solaris/Oracle
world you can get advantage of extra memory, but, not being a
professional in that area, I can't comment on it.

You can get win64...I don't much about MSSQL, but everything you've
said sounds reasonable (won't port 2000, 2005 is still too new). I
think people will migrate to '05 if they see rather substantial
benefits.

DK

 

[Keith]

There are still very few softw. that support 64, not to mention
_require_ it. I was surprized recently that Half Life II was released
in 64 bit version. However I am not going to run and by win xp-64 just
because of that.

Support it? There is a ton (SuSE is all 64bit). Require, perhaps not.
IMO when we went to GB real memory two or three GB virtual gets a little
close. I've had my laptop shut down short of memory (2GB real).

Now if you need to address 8 GB of memory for some purposes (DB
server?), 64 bit is the only game in town.

Real memory <> virtual memory. 32bit machiens can only address 4GB (2GB
in most cases) virtual memory, no matter how much real memory they have.
No, we're well in shootin'-range of 64b being necessary.

 

[George Macdonald]

Sorta like the 386s? I consider the uptake of AMD64 to be
nothing short of phenomenal in comparison. In a far more
entrenched and less geeky market, we have day one OSes and a
remarkable amount of support. Even from the known laggard, MS.

IIRC the 386 had to wait 5 years for runnable OSes (IIRC, OS/2
v2 and non-beta Linux). Apps followed, and will this time too.

There *were* the 386 DOS Extenders and Desqview386 which was a very solid
multitasking environment, long before Windows 3.0 came along or was even
worth looking at. With a coupla nudges here & there, things could have
turned out very differently. In the corporate market, the IBM/M$ "promise"
of OS/2 V1 had a serious effect in slowing down 32-bit Protected Mode
uptake.

I'm sure we still have the "notices" from our big clients warning us of the
requirements for OS/2 V1 versions of our software to fit their scheduled
transition; fortunately we ignored them... never happened. I'd attended
the OS/2 V1 dog 'n' pony show in NYC and I was not impressed... though it
did take some serious "discussions" with our guys "out in the trenches" to
stay away from it. Besides, Phar Lap's DOS Extender was so simple and
DesqView386 was no effort at all - it just worked.

 

[George Macdonald]

It's an ironic world we live in, when people find you more acceptable
because you're incompetant.

There were good reasons for it to delay introduction of SP2, as they
were trying to fix a leaky boat. At that time security on Windows seemed
like swiss cheese. However, they had x64 finished much earlier, even
though they kept releasing release candidates. They were hoping for
device drivers to be released, but nobody would release device drivers
until Microsoft released the OS. They could've released x64 and still
released SP2 for 32-bit.

Whether they could have or not, is a matter which can only be known inside
M$ - anything else is speculation. They publicly stated that SP2 had
priority and x-64 would have to come after all the security stuff had been

Really the most telling sign that they were delaying was the fact that
Sun Microsystems started development work on their Solaris 10 x64 about
two years after Microsoft, and still released it six months before
Microsoft! One could say that Sun has years of experience with 64-bit
operating systems, and Microsoft doesn't. But Microsoft already had
64-bit source code for Windows in their Itanium port, and other than
low-level assembly language differences, most of it would be the same
between the two architectures, just a recompile away.

If Windows x64 had been released near to when AMD had released their
hardware, even if AMD didn't sell as many chips as Intel, the fact that
a released OS was sitting there it would've been gathering device driver
support ready for the time when Intel came on board. By the time Intel
was ready with 64-bit, you'd have a mature operating system already with
lots of device driver support. And people coming in with Intel would
have had much easier time. AMD users could've acted as the beta testers
for Intel users.

Debatable IMO - I have my suspicions, just as you do, but the "situation"
does have *some* ring of truth to it. AMD still had no momentum and
certainly not in servers at the time. Would the drivers have materialized
for a platform with such a small market footprint?... hard to be certain
from my POV. Even now there are many x64 drivers which are lacking,
extremely buggy or for devices like printers which will never appear.

 

[Del Cecchi]

There *were* the 386 DOS Extenders and Desqview386 which was a very
solid
multitasking environment, long before Windows 3.0 came along or was
even
worth looking at. With a coupla nudges here & there, things could have
turned out very differently. In the corporate market, the IBM/M$
"promise"
of OS/2 V1 had a serious effect in slowing down 32-bit Protected Mode
uptake.

I'm sure we still have the "notices" from our big clients warning us of
the
requirements for OS/2 V1 versions of our software to fit their
scheduled
transition; fortunately we ignored them... never happened. I'd
attended
the OS/2 V1 dog 'n' pony show in NYC and I was not impressed... though
it
did take some serious "discussions" with our guys "out in the trenches"
to
stay away from it. Besides, Phar Lap's DOS Extender was so simple and
DesqView386 was no effort at all - it just worked.

I'm sure you recall that OS/2 was supposed to do all that stuff while
running on a 286 PC/AT.

del

 

[Tony Hill]

Adorable little Ed Stroglio rant:

Hatching Eggs . . .
"Remember x86-64? The stick AMD was supposed to beat Intel to death
with?

Then Intel got themselves the same stick, and the would-be beaters
found something else to talk about. Yes, AMD is doing better now, but
not because of x86-64. Indeed, AMD's increased fortunes have come well
after Intel switched.

Microsoft came out with a Windows XP for x86-64, but the world hasn't
exactly stampeded to get it."
http://www.overclockers.com/tips00910/

I always thought the reason nobody flocked to XP x64 was because
Microsoft delayed and delayed so much to let Intel catch up that people
lost interest in it? I mean you know Microsoft was delaying here, even
Solaris 10 came out for x64 before Windows did, despite starting work
on it 2 years later.

The reason most people I talk to are still running 32-bit WinXP on
their 64-bit workstations is that the driver support *STILL* isn't
there for a lot of their peripherals. Sure, the big name stuff like
nVidia, Broadcom, etc. is fine. However when you start throwing some
obscure hardware in the system you're often SOL.

Of course, that's not to say that Microsoft is totally without blame
here, other OSes seem to have had a much easier time in their
transition to 64-bit drivers. For most it was just a recompile, or at
worst a few patches here and there to make the code 64-bit safe. On
Windows it seems like it's a rather more complicated process.

 

[Tony Hill]

It's an ironic world we live in, when people find you more acceptable
because you're incompetant.

There were good reasons for it to delay introduction of SP2, as they
were trying to fix a leaky boat. At that time security on Windows seemed
like swiss cheese. However, they had x64 finished much earlier, even
though they kept releasing release candidates. They were hoping for
device drivers to be released, but nobody would release device drivers
until Microsoft released the OS. They could've released x64 and still
released SP2 for 32-bit.

My outsider's view of it was that there was a big effort on-going at
Microsoft to link up the code-bases from a security standpoint. As
such they wanted to have roughly the same base code (as far as service
packs go at least) for WinXP Home, XP Pro, XP x64 and Win2K3 Server.
Since XP is BY FAR the biggest ticket item/biggest money maker of the
lot, they needed to get that one solidified first. Everything else
could be pushed back behind that.

Again though, this is just the outsider's perspective. What was going
on in the inside could very well have just been complete chaos! :>

 

[George Macdonald]

I'm sure you recall that OS/2 was supposed to do all that stuff while
running on a 286 PC/AT.

Yeah well that was the big mistake. Back then even Intel took the stance
that the 386 was only for those "who really needed it" - they thought the
same with 486... and even had a slew of 386 support chips still in the
pipeline when 486 came out.

But "all that stuff" I mentioned included the 32-bit flat memory model -
that was the killer for doing real computing on a desktop. A lot of
mainframe code got converted to Phar Lap's DOS Extender.

 

[Henry Nettles]

message


I'm sure you recall that OS/2 was supposed to do all that stuff while
running on a 286 PC/AT.

del

I still remember the PS/2 rollout in 1987. I was working for a major
oil company, who was a big time IBM client (millions and millions of
dollars each and every year). One of the big conference rooms was
turned over to IBM, and they had several sharply dressed sales-droids
there to "present" the new PS/2 computers. I remember being totally
underwhelmed that they were trying to sell us 286 based computers.

It still seems totally insane to me that anyone would have spent
serious money to engineer a new 286 computer a year after the 386
computers had reached the market.

Looking back on it all now, it still seems to me that the PS/2 was the
biggest fiasco of the entire computer revolution. I used OS/2 for a
while, and I still think OS/2 was better than Windows 95. But the
PS/2, now that was a fiasco from beginning to end.