64-bit or 32-bit: When will it matter?

avid Schwartz wrote:
"Bill Davidsen" wrote in message
. ..


David Schwartz wrote:


The need will be the things that will be possible on 64-bit platforms
that we don't even think about on 32-bit platforms. For example, you can
'mmap' an entire partition. You can write code that reserves large chunks
of address space for sparse data.


Sure they can, but what application will benefit from that?


All of the high-end ones. As soon as it's practical to make an app that
only works on a 64-bit machine, people will do it.

Hate to keep asking, but what "high end ones" are those? Outside of BIG
database apps, which are usually server functions rather than user apps,
and high end graphics, which are usually done on dedicated workstations,
I don't see where the benefit will come. And to use all that address
space will require rewriting the applications which now use the 64 bit
file addressing which has been around for about a decade.


I keep hearing the old song "things we haven't even thought of yet," but
we have had 64 bit Sparc and Power for a decade, and I haven't seen the
common desktop application which is pushing the limits of the memory we
have now. Where's the killer app?


It's not about pushing the limits of the memory, it's about having lots
of address space and using it.

I keep hearing that, too, what's the app which benefits?


My premise is that somewhere between 75-90% of home and office users will
not need the address space, because the common things are bounded by the
human not the computer. That limits the size of a useful web page,
graphic, etc. Email isn't going to be 4GB, or manuals, and even in the USA
the tax forms won't be that large. You can propose a spreadsheet of huge
size, but it's not easily human usable and generally better served by a
database.


How much data fits on a DVD? How much will fit on a blu-ray or the next
high density storage medium? How much fits on a hard drive today?

Which applications read an entire DVD into memory, and why? And other
than a tiny reduction in code complexity, where's the gain? That's a
real gain for a "from scratch" new program, but what vendor will lock
out the 32 bit market by doing it? The gain is not in performance, if
the data doesn't fit in physical memory performance will be limited by
disk speed.


Probably so. If you assume memory sizes will want to continue to grow
at the same rate, the average machine will be hitting the 4GB limit
pretty soon now, in about three years if memory serves me.

It won't be long before more new x86 machines have 64-bit capable
processors than don't.

New machines? That sounds possible, the 64 bit capable CPUs are
inexpensive, it's cheap insurance. But people won't be rushing to get
upgrades to 64 bit, their stuff works now.


I agree. Nothing will dramatically change until it makes commercial
sense to release software that only works on 64-bit. Then people will still
only do that if they get some significant benefit or the 32-bit market is
really small. The more the 64-bit benefit, and the higher percentage of the
target machines that are 64-bit capable, the more likely that 64-bit only
software will be released.

You express my thoughts exactly in that paragraph! I just don't see many
places where there will be a co$t advantage or a performance advantage
with applications which sell in any significant volume. I talked to an
archetect about 64 bit when the UltraSPARC came out, and he said that
the drawing programs were all vector, and took almost no space at all.
That was one place I thought 64 bits would help, and it doesn't.

One thing that hasn't been mentioned is the larger register set in
64-bit mode.

I doubt that make a difference. There's so much register aliasing in
modern processors that (from what I read) the program logic is more of a
limiting factor. That and the quality of the compilers.

--
bill davidsen )
SBC/Prodigy Yorktown Heights NY data center
Project Leader, USENET news
http://newsgroups.news.prodigy.com

"Bill Davidsen" wrote in message
m...

Sure they can, but what application will benefit from that?

All of the high-end ones. As soon as it's practical to make an app
that only works on a 64-bit machine, people will do it.

Hate to keep asking, but what "high end ones" are those?

Physics engines such as are used in 3D games. Graphics applications.
Development tools. Encryption. All the applications that don't exist yet
because today's computers can't run them efficiently.

Outside of BIG database apps, which are usually server functions rather
than user apps, and high end graphics, which are usually done on dedicated
workstations,

What about the high end graphics, where by high end I mean high end for
what's done on PCs? Things are done on dedicated workstations only when they
can't be done on PCs.

Don't forget, when the '486 first came out, everyone was saying that
nobody would ever need that much power on the desktop. If you give people
the ability to do things, they will want to do them.

I don't see where the benefit will come. And to use all that address space
will require rewriting the applications which now use the 64 bit file
addressing which has been around for about a decade.

Not necessarily. A lot of work can be done by the OS. And the 80/20
rules means that only 20% of the application has to be rewritten to get the
benefit.

I keep hearing the old song "things we haven't even thought of yet," but
we have had 64 bit Sparc and Power for a decade, and I haven't seen the
common desktop application which is pushing the limits of the memory we
have now. Where's the killer app?

It's not about pushing the limits of the memory, it's about having
lots of address space and using it.

I keep hearing that, too, what's the app which benefits?

All applications that use disks and files. Games. Document preparation
tools. Encryption. Once people can do things, they will insist on being able
to.

My premise is that somewhere between 75-90% of home and office users will
not need the address space, because the common things are bounded by the
human not the computer. That limits the size of a useful web page,
graphic, etc. Email isn't going to be 4GB, or manuals, and even in the
USA the tax forms won't be that large. You can propose a spreadsheet of
huge size, but it's not easily human usable and generally better served
by a database.

How much data fits on a DVD? How much will fit on a blu-ray or the
next high density storage medium? How much fits on a hard drive today?

Which applications read an entire DVD into memory, and why?

That's not the point. The point is that the application may need any
part of the DVD at any time.

And other than a tiny reduction in code complexity, where's the gain?

It's much more than a tiny reduction in code complexity. Managing a
limited address space is a major PITA.

That's a real gain for a "from scratch" new program, but what vendor will
lock out the 32 bit market by doing it? The gain is not in performance, if
the data doesn't fit in physical memory performance will be limited by
disk speed.

The gain is also in performance. Not having to view the world through a
narrow window will improve performance as the OS manages the caching rather
than the application, which has a more limited view of system resources and
priorities.

I agree. Nothing will dramatically change until it makes commercial
sense to release software that only works on 64-bit. Then people will
still only do that if they get some significant benefit or the 32-bit
market is really small. The more the 64-bit benefit, and the higher
percentage of the target machines that are 64-bit capable, the more
likely that 64-bit only software will be released.

You express my thoughts exactly in that paragraph! I just don't see many
places where there will be a co$t advantage or a performance advantage
with applications which sell in any significant volume. I talked to an
archetect about 64 bit when the UltraSPARC came out, and he said that the
drawing programs were all vector, and took almost no space at all. That
was one place I thought 64 bits would help, and it doesn't.

The issue is not what current applications will benefit, but what new
applications will be possible.

One thing that hasn't been mentioned is the larger register set in
64-bit mode.

I doubt that make a difference. There's so much register aliasing in
modern processors that (from what I read) the program logic is more of a
limiting factor. That and the quality of the compilers.

More registers will mean you won't need as high a compiler quality to
get the performance. Or, to put it another way, for the same compiler
quality, you can get more performance with more registers.

DS

David Schwartz wrote:

This has become pointless, I keep asking for the name or type of
application which any significant postion of desktop users (personal or
business) would find useful enough to justify buying a 64 bit computer
before the normal lifespan of the existing 32 but unit. Haven't gotten
it yet.

I don't expect any rush to 64 bit, when it's cheap and time for a new
computer, then people will go 64 bit. I would guess that means 90% of
the office desktops and 75% of the home units will be 32 bits until they
hit end of useful life.

"Bill Davidsen" wrote in message
m...


Sure they can, but what application will benefit from that?


All of the high-end ones. As soon as it's practical to make an app
that only works on a 64-bit machine, people will do it.


Hate to keep asking, but what "high end ones" are those?


Physics engines such as are used in 3D games. Graphics applications.
Development tools. Encryption. All the applications that don't exist yet
because today's computers can't run them efficiently.


Outside of BIG database apps, which are usually server functions rather
than user apps, and high end graphics, which are usually done on dedicated
workstations,


What about the high end graphics, where by high end I mean high end for
what's done on PCs? Things are done on dedicated workstations only when they
can't be done on PCs.

Don't forget, when the '486 first came out, everyone was saying that
nobody would ever need that much power on the desktop. If you give people
the ability to do things, they will want to do them.


I don't see where the benefit will come. And to use all that address space
will require rewriting the applications which now use the 64 bit file
addressing which has been around for about a decade.


Not necessarily. A lot of work can be done by the OS. And the 80/20
rules means that only 20% of the application has to be rewritten to get the
benefit.


I keep hearing the old song "things we haven't even thought of yet," but
we have had 64 bit Sparc and Power for a decade, and I haven't seen the
common desktop application which is pushing the limits of the memory we
have now. Where's the killer app?


It's not about pushing the limits of the memory, it's about having
lots of address space and using it.


I keep hearing that, too, what's the app which benefits?


All applications that use disks and files. Games. Document preparation
tools. Encryption. Once people can do things, they will insist on being able
to.


My premise is that somewhere between 75-90% of home and office users will
not need the address space, because the common things are bounded by the
human not the computer. That limits the size of a useful web page,
graphic, etc. Email isn't going to be 4GB, or manuals, and even in the
USA the tax forms won't be that large. You can propose a spreadsheet of
huge size, but it's not easily human usable and generally better served
by a database.


How much data fits on a DVD? How much will fit on a blu-ray or the
next high density storage medium? How much fits on a hard drive today?


Which applications read an entire DVD into memory, and why?


That's not the point. The point is that the application may need any
part of the DVD at any time.


And other than a tiny reduction in code complexity, where's the gain?


It's much more than a tiny reduction in code complexity. Managing a
limited address space is a major PITA.


That's a real gain for a "from scratch" new program, but what vendor will
lock out the 32 bit market by doing it? The gain is not in performance, if
the data doesn't fit in physical memory performance will be limited by
disk speed.


The gain is also in performance. Not having to view the world through a
narrow window will improve performance as the OS manages the caching rather
than the application, which has a more limited view of system resources and
priorities.


I agree. Nothing will dramatically change until it makes commercial
sense to release software that only works on 64-bit. Then people will
still only do that if they get some significant benefit or the 32-bit
market is really small. The more the 64-bit benefit, and the higher
percentage of the target machines that are 64-bit capable, the more
likely that 64-bit only software will be released.


You express my thoughts exactly in that paragraph! I just don't see many
places where there will be a co$t advantage or a performance advantage
with applications which sell in any significant volume. I talked to an
archetect about 64 bit when the UltraSPARC came out, and he said that the
drawing programs were all vector, and took almost no space at all. That
was one place I thought 64 bits would help, and it doesn't.


The issue is not what current applications will benefit, but what new
applications will be possible.


One thing that hasn't been mentioned is the larger register set in
64-bit mode.


I doubt that make a difference. There's so much register aliasing in
modern processors that (from what I read) the program logic is more of a
limiting factor. That and the quality of the compilers.


More registers will mean you won't need as high a compiler quality to
get the performance. Or, to put it another way, for the same compiler
quality, you can get more performance with more registers.

DS




--
bill davidsen
SBC/Prodigy Yorktown Heights NY data center
http://newsgroups.news.prodigy.com

"Bill Davidsen" wrote in message
m...

This has become pointless, I keep asking for the name or type of
application which any significant postion of desktop users (personal or
business) would find useful enough to justify buying a 64 bit computer
before the normal lifespan of the existing 32 but unit. Haven't gotten it
yet.

Right, because that's a pointless question.

Before the WWW was invented, you would ask for the application that
would make the ordinary person want to, and be able to, use the Internet.
Nobody could name it.

When the '486 was released, articles said it was so powerful it would
forever be in the server room, never on the desktop. Where were the
applications that would need it?

Of course people aren't using the applications today that require
64-bits to work smoothly because they don't *have* 64-bits. Perhaps in the
near future, people will use 3D rendering software on their desktops that
rival what's now done only in Hollywood. Crazy? Who would have predicted in
1990 that 15 years later people would do desktop publishing on their
desktops comparable to what's done in publishing houses?

Anything that can be done, people will want to do it on their desktops.
And as soon as it's possible, they'll start doing it.

I don't expect any rush to 64 bit, when it's cheap and time for a new
computer, then people will go 64 bit. I would guess that means 90% of the
office desktops and 75% of the home units will be 32 bits until they hit
end of useful life.

I predict that it will be comparable to the switch from 16-bits to
32-bits. Perhaps slightly slower.

DS

David Schwartz wrote:
"Bill Davidsen" wrote in message
m...


This has become pointless, I keep asking for the name or type of
application which any significant postion of desktop users (personal or
business) would find useful enough to justify buying a 64 bit computer
before the normal lifespan of the existing 32 but unit. Haven't gotten it
yet.


Right, because that's a pointless question.

Before the WWW was invented, you would ask for the application that
would make the ordinary person want to, and be able to, use the Internet.
Nobody could name it.

When the '486 was released, articles said it was so powerful it would
forever be in the server room, never on the desktop. Where were the
applications that would need it?

Of course people aren't using the applications today that require
64-bits to work smoothly because they don't *have* 64-bits. Perhaps in the
near future, people will use 3D rendering software on their desktops that
rival what's now done only in Hollywood. Crazy? Who would have predicted in
1990 that 15 years later people would do desktop publishing on their
desktops comparable to what's done in publishing houses?

Anything that can be done, people will want to do it on their desktops.
And as soon as it's possible, they'll start doing it.

We've had 64 bits for a decade, and the killer desktop apps are somewhat
missing. I'm not holding my breath, the people who do mail and browsing,
maybe their taxes and a "fits on the screen" spreadsheet are the
majority of both personal and business users. Documents, spreadsheets,
E-mail and web pages are limited by people now. Only taxes could be
complex enough to require 64 bits ;-(


I don't expect any rush to 64 bit, when it's cheap and time for a new
computer, then people will go 64 bit. I would guess that means 90% of the
office desktops and 75% of the home units will be 32 bits until they hit
end of useful life.


I predict that it will be comparable to the switch from 16-bits to
32-bits. Perhaps slightly slower.

Were you around as a participant at that time? The 16 bit systems were
so limited by what you could use and address and do that people jumped
from 16 to 32 bits as soon as they could, and that was "way too long"
for DOS users, much less for the UNIX users of us.

It takes some level of dissatisfaction with what you have to make you
spend, "you" being personal or corporate. Maybe when that killer 64 bit
app comes out. When people are ready to buy a new system anyway they may
go 64 bit if the price is similar. It is at the bottom end and
definitely isn't at the server end.

Save this post, if a killer app comes out which needs 64 bits in two
years, you can reply.
--
bill davidsen )
SBC/Prodigy Yorktown Heights NY data center
Project Leader, USENET news
http://newsgroups.news.prodigy.com

"Bill Davidsen" wrote in message
. com...

Were you around as a participant at that time?

Yes. Do you remember how quickly the requirements on typical games went
up? How one year most games required only a 386 and a year later they
required a 486? How long was it after that before 133Mhz or better was
needed?

I just grabbed a random program off my shelf and the box requirements
say "800Mhz Pentium 3 or faster". This would mean your CPU can be no more
than 5 years old *if* it was the very fastest CPU purchased at the time of
purchase! More realistically (few typical users buy the very fastest CPU
available), it means your CPU can be no more than 3 years old.

The 16 bit systems were so limited by what you could use and address and
do that people jumped from 16 to 32 bits as soon as they could, and that
was "way too long" for DOS users, much less for the UNIX users of us.

Right, and that will be the case with 32-bit systems in about three
years as people want to put more than 2Gb (and then more than 4Gb) of memory
in their systems.

It takes some level of dissatisfaction with what you have to make you
spend, "you" being personal or corporate. Maybe when that killer 64 bit
app comes out. When people are ready to buy a new system anyway they may
go 64 bit if the price is similar. It is at the bottom end and definitely
isn't at the server end.

All new x86 CPUs except perhaps the very low end will be 64-bit capable
in less than two years. How long do you think it will be before new PCs ship
with a 64-bit operating system by default?

Save this post, if a killer app comes out which needs 64 bits in two
years, you can reply.

There were ways to handle larger amounts of memory on 16-bit systems
too. It is my bet that at least 15% of commodity software will require a
64-bit CPU and OS within five years.

You are essentially predicting that software requirements will lag
behind hardware availability by an amount that they have never lagged
before. Ever.

DS

Hi,

"David Schwartz" wrote in message
...
Right, and that will be the case with 32-bit systems in about three
years as people want to put more than 2Gb (and then more than 4Gb) of
memory
in their systems.

I think you're missing the difference between "physical addresses"
and "virtual addresses", and the difference between architectural
design and CPU implementation.

The physical address size determines how much RAM (and ROM,
memory mapped devices, etc) a CPU can access. The OS uses this
RAM to create "virtual address spaces" (or fake address spaces)
that other software uses.

For about a decade, 80x86 CPUs have had "physical address extensions"
which allows for 36 bit physical addresses. This limits the amount of
RAM (and ROM, memory mapped devices, etc) to 64 GB (not 4 GB).
The virtual address size is still 32 bit though (or 4 GB), and the OS
usually uses 1 GB or 2 GB for itself, which means applications can
only have up to 2 GB or 3 GB each. This means you can have a
computer with 64 GB of RAM running 32 applications where
each application uses 2 GB (without anything being 64 bit).

All of that is "architectural design" rather than what was actually
implemented though. Back when the 80386 was first designed, Intel
allowed for 32 bit physical addresses in the architectural design,
but (to save everyone money) they didn't implement full 32 bit
physical addressing. They knew no-one would need it (at the time)
so they only implemented 24 bit physical addressing, so that these
CPUs could only access 16 MB of RAM (and ROM, memory mapped devices,
etc).

For a 64 bit CPU, the physical address size is 52 bits and the virtual
address size is 64 bits, but these are the architectural limits which
aren't actually implemented in any 64 bit 80x86 CPU (and probably
won't be implemented for a very long time).

For example, Intel's first 64 bit CPU only implemented 40 bit virtual
addressing and 36 bit physical addressing. AMD's 'Hammer' supports
48 bit virtual addresses and 40 bit physical addresses.

Ironically, the main benefit of "64 bit 80x86" has nothing to do with
being (partially) 64 bit - the additional registers are far more
important and are responsible for any and all performance gains.


Cheers,

Brendan

Brendan Trotter wrote:
Ironically, the main benefit of "64 bit 80x86" has nothing to do with
being (partially) 64 bit - the additional registers are far more
important and are responsible for any and all performance gains.
It's true that the additional registers are the reason for a (small)
performance gain in most common software. However, it is NOT true that
this is resposnisble for "any and all" performance gains. There
definitely is some software out there which indeed can get quite a
massive performance boost from a 64-bit architecture (more than a factor
of 2). Encryption/Decryption software often belongs to that category.

Roland

David Schwartz wrote:
"Bill Davidsen" wrote in message
. com...


Were you around as a participant at that time?


Yes. Do you remember how quickly the requirements on typical games went
up? How one year most games required only a 386 and a year later they
required a 486? How long was it after that before 133Mhz or better was
needed?

I just grabbed a random program off my shelf and the box requirements
say "800Mhz Pentium 3 or faster". This would mean your CPU can be no more
than 5 years old *if* it was the very fastest CPU purchased at the time of
purchase! More realistically (few typical users buy the very fastest CPU
available), it means your CPU can be no more than 3 years old.


The 16 bit systems were so limited by what you could use and address and
do that people jumped from 16 to 32 bits as soon as they could, and that
was "way too long" for DOS users, much less for the UNIX users of us.


Right, and that will be the case with 32-bit systems in about three
years as people want to put more than 2Gb (and then more than 4Gb) of memory
in their systems.


It takes some level of dissatisfaction with what you have to make you
spend, "you" being personal or corporate. Maybe when that killer 64 bit
app comes out. When people are ready to buy a new system anyway they may
go 64 bit if the price is similar. It is at the bottom end and definitely
isn't at the server end.


All new x86 CPUs except perhaps the very low end will be 64-bit capable
in less than two years. How long do you think it will be before new PCs ship
with a 64-bit operating system by default?


Save this post, if a killer app comes out which needs 64 bits in two
years, you can reply.


There were ways to handle larger amounts of memory on 16-bit systems
too. It is my bet that at least 15% of commodity software will require a
64-bit CPU and OS within five years.

Oh there were, but they were painful to use in most cases. To address
your main point, it depends on your definition of commodity software,
but by any definition I don't see that as a "killer app" justifying
moving from 32 to 64 bit hardware before the old system is due for
replacement. Actually I would consider that over half of the computers
in desktop use are going to be replaced in 6-7 years, with nothing more
than attrition driving it.

You are essentially predicting that software requirements will lag
behind hardware availability by an amount that they have never lagged
before. Ever.

I am. Based on two different justifications. The best is that there
hasn't been a 64 bit killer app for the Mac, and that's been 64 bit for
a decade. The other is that there *is* a point when people have enough
and are not willing to make an upgrade because they don't see the need.

If you can get access to sales info, most 32 bit systems aren't ordered
with max memory, largest disk, or fastest CPU. That certainly suggests
that people don't feel the need.

I'm still confident that 64 bit hardware will come driven by replacement
rather than upgrade. I agree that most computers will include 64 bit
capability, but only because it will be standard. Intel and AMD are
unlikely to spend any money in 32 bit only products, when they need more
performance and lower power foar more.

I predict the big 64 bit software push will be driven by greed, I mean
marketing. When enough people have gotten 64 bit hardware, Microsoft
will suddenly release new versions of all apps, with new features, and
in 64 bit only. I predict they will offer *very* cheap upgrade from 32
bit versions, because they know they will make the money on Windows-64
o/s upgrades. But until most people have the hardware they won't push 64
bit only, because it locks them out of a majority of the market.

As for gamers? I define a gamer as someone who spends at least $100
extra on a computer for memory, faster CPU, or detter display. Oddly,
that lets out a fair percentage of people who do little else with their
computer. If they didn't spend money on hardware at 32 bits, will they
jump to 64 intesad of spending the money on more games? For that matter,
are the games on the 64 bit Mac better? (real question, I have no idea)


--
bill davidsen
SBC/Prodigy Yorktown Heights NY data center
http://newsgroups.news.prodigy.com

"Brendan Trotter" wrote in message
...

"David Schwartz" wrote in message
...

Right, and that will be the case with 32-bit systems in about three
years as people want to put more than 2Gb (and then more than 4Gb) of
memory
in their systems.

I think you're missing the difference between "physical addresses"
and "virtual addresses", and the difference between architectural
design and CPU implementation.

No, I'm not missing anything. What you said has nothing whatsoever to do
with what I said. It's *possible* to address an unlimited amount of memory
with an 8-bit CPU, but nobody does that if they don't have to. As soon as
the majority of computers are 64-bit capable, they won't have to, and so
they won't. In any event, the shortage of virtual addresses is the more
serious problem.

PAE is an ugly workaround. A lot of people specifically purchase the
maximum amount of memory their OS can handle without PAE because they want
to avoid it.

DS