Intel takes back lead in U.S. retail

[krw]

I didn't say that NO ONE needs it. Just that the vast majority of people
buying from retail won't. I have a 64-bit capable machine that I just built
and I expect to go to EOL with Windows XP 32-bit on it. And I'm not exactly
the average consumer.

Nor am I, though I've been around the block a few times. What I
need today, everyone else will "need"[*] in a few short years.

Given MS marketting I don't doubt they will think that they need it.

What??? Given M$ (and Intel when, and if, they get their act
together) marketing they will *KNOW* they need it. ...and it
doesn't matter what "it" is!

[*], yes I know I can surf the net on a P5-133, or whatever. My
K6-400 was fine until it decided it no longer wanted to boot (still
gotta figure out what happened).

 

[James Jones]

What??? Given M$ (and Intel when, and if, they get their act
together) marketing they will *KNOW* they need it. ...and it
doesn't matter what "it" is!

I can't do much about what marketting fools people into believing.
The fact remains that your machine with a 32-bit OS isn't going to
be obsolete any time soon.
- Jim

 

[krw]

I can't do much about what marketting fools people into believing.

That's exactly the point.

The fact remains that your machine with a 32-bit OS isn't going to
be obsolete any time soon.

My machine, with its 2bit OS won't be obsolete because I just
bought it (ThinkPad). Neither will the two-year-old 64bit system
with the 64bit OS sitting next to it. ;-)

 

[Tony Hill]

And you need 64-bits for what? Once again, in servers/workstations
absolutely, but mobile and desktop?

We're really starting to push up against the limits of 32-bit desktop
systems these days. 1GB of memory is pretty much the norm and 2GB
isn't all that uncommon. Typically going beyond about 1.5GB of memory
means that you'll want a 64-bit system to prevent having to use Ugly
Kludges to get all your memory working. Remember that it's not just
your physical memory that is counted in the 4GB limit, but your
virtual memory. 1.5GB of physical plus a 2.5GB page file equals your
4GB limit.

Most retail sales don't and won't care for several more years at least.

They should care because it's just the right thing to do and it's
free.

No one, who has a clue and in retail, is going to upgrade just for 64-bit
support. And yes, I like my Yonah laptop a lot! ;-)

That's fine. It's not like a 32-bit system is broken or anything,
they aren't going to disappear for a while. However when buying a new
desktop system now or a new laptop system in 3-5 months, there is NO
good reason to opt for a 32-bit chip. 64-bit offers sufficient
enhancements to make it worthwhile, Linux supports 64-bit fully now
and Windows Vista will be available as 64-bit out of the box (WinXP
64-bit is available too, but not as widely supported). The long story
short is that the days of new 32-bit systems are numbered.

 

[krw]

We're really starting to push up against the limits of 32-bit desktop
systems these days. 1GB of memory is pretty much the norm and 2GB
isn't all that uncommon. Typically going beyond about 1.5GB of memory
means that you'll want a 64-bit system to prevent having to use Ugly
Kludges to get all your memory working. Remember that it's not just
your physical memory that is counted in the 4GB limit, but your
virtual memory. 1.5GB of physical plus a 2.5GB page file equals your
4GB limit.

Don't forget the virtual memory space the OS reserves for itself
and virtual and real I/O. In reality you only have 2-3GB with a
32bit CPU/OS.

They should care because it's just the right thing to do and it's
free.

Exactly. Maybe people should still be buying 486s because they
don't "need" Pentiums.

That's fine. It's not like a 32-bit system is broken or anything,
they aren't going to disappear for a while. However when buying a new
desktop system now or a new laptop system in 3-5 months, there is NO
good reason to opt for a 32-bit chip. 64-bit offers sufficient
enhancements to make it worthwhile, Linux supports 64-bit fully now
and Windows Vista will be available as 64-bit out of the box (WinXP
64-bit is available too, but not as widely supported). The long story
short is that the days of new 32-bit systems are numbered.

Yep.

 

[nobody]

Don't forget the virtual memory space the OS reserves for itself
and virtual and real I/O. In reality you only have 2-3GB with a
32bit CPU/OS.

Exactly. Maybe people should still be buying 486s because they
don't "need" Pentiums.


Yep.

Hardware is getting faster, smaller, and cheaper.
Software is getting slower, bigger, and pricier. Even free Linux, if
you account for TCO.
;-)
NNN

 

[Rod Speed]

Hardware is getting faster, smaller, and cheaper.

Not always cheaper.

Software is getting slower,

Not always, some of it does some ops faster than had been done before.

bigger,

Not always.

and pricier.

Not always, quite a bit is now free.

Even free Linux, if you account for TCO.

Not necessarily.

 

[chrisv]

We're really starting to push up against the limits of 32-bit desktop
systems these days. 1GB of memory is pretty much the norm and 2GB
isn't all that uncommon.

But 2G is plenty for the vast majority of PC's being sold today.

 

[Kai Harrekilde-Petersen]

We're really starting to push up against the limits of 32-bit desktop
systems these days. 1GB of memory is pretty much the norm and 2GB
isn't all that uncommon. Typically going beyond about 1.5GB of memory
means that you'll want a 64-bit system to prevent having to use Ugly
Kludges to get all your memory working. Remember that it's not just
your physical memory that is counted in the 4GB limit, but your
virtual memory. 1.5GB of physical plus a 2.5GB page file equals your
4GB limit.

This will probably not be the first hurdle for 32bit OS'es. What is
more likely to hurt before this is fragmentation of your virtual
memory, when you have a single app that uses >1GB of memory (think
'games') and does frequent allocation/deallocation of memory to the
OS.


Kai

 

[Tony Hill]

But 2G is plenty for the vast majority of PC's being sold today.

Yes, and with 2GB you should really be using a 64-bit processor and
OS, otherwise you're having to deal with ugly hacks. For most OSes
(Windows and *BSD at least) you should always have more swap space
than you have physical memory, so going beyond about 1.5GB physical
memory means that your running into issues with 32-bit systems. In
fact, even at 1.5GB of physical memory you should probably aim for 2GB
or more of swap space for Windows to run ideally, and you just can't
do that properly on a 32-bit Windows system given that you've only got
4GB of total address space and Windows reserves 700MB to 1GB of that
for various I/O functions.

 

[Rod Speed]

Yes, and with 2GB you should really be using a 64-bit processor
and OS, otherwise you're having to deal with ugly hacks.

Not if you dont use it fully.

For most OSes (Windows and *BSD at least) you should always
have more swap space than you have physical memory,

Mindlessly silly. As the amount of physical ram
increases, the need for swap space DROPS.

so going beyond about 1.5GB physical memory means
that your running into issues with 32-bit systems.
Bullshit.

In fact, even at 1.5GB of physical memory you should probably
aim for 2GB or more of swap space for Windows to run ideally,

Mindlessly silly.

and you just can't do that properly on a 32-bit Windows system
given that you've only got 4GB of total address space and Windows
reserves 700MB to 1GB of that for various I/O functions.

Gets sillier by the minute.

 

[The little lost angel]

Yes, and with 2GB you should really be using a 64-bit processor and
OS, otherwise you're having to deal with ugly hacks. For most OSes
(Windows and *BSD at least) you should always have more swap space
than you have physical memory, so going beyond about 1.5GB physical

Why is this so though? Personally it seems that the more RAM I added,
the less swap space I set aside in proportion to the actual physical
RAM I have. After all, with all the RAM, it's supposed to help avoid
hitting the comparatively slow hard disk no?

 

[chrisv]

Yes, and with 2GB you should really be using a 64-bit processor and
OS, otherwise you're having to deal with ugly hacks. For most OSes
(Windows and *BSD at least) you should always have more swap space
than you have physical memory, so going beyond about 1.5GB physical
memory means that your running into issues with 32-bit systems. In
fact, even at 1.5GB of physical memory you should probably aim for 2GB
or more of swap space for Windows to run ideally, and you just can't
do that properly on a 32-bit Windows system given that you've only got
4GB of total address space and Windows reserves 700MB to 1GB of that
for various I/O functions.

Eh? Sorry, I don't buy that at all. "Oh, you installed more memory?
Now you need more HD swap space!" I don't think so...

 

[jack]

:
: Not if you dont use it fully.
:
: Mindlessly silly. As the amount of physical ram
: increases, the need for swap space DROPS.
:
: Bullshit.
:
: Mindlessly silly.
:
: Gets sillier by the minute.

Ostia! More words of wisdom from the resident troll and village idiot,
"Rod Speed" (aka Ron Reaugh). What a ****ing moron, what an idiot...more
dense than any black hole we can imagine...so dense that intellect can't
escape....LOL!!

j.

 

[Robert Redelmeier]

Eh? Sorry, I don't buy that at all. "Oh, you installed more
memory? Now you need more HD swap space!" I don't think so...

Well, you are not required to believe things even if they
are true. It just helps your credibility.

Surely you can see that a suspend-to-disk [hibernation]
file will have to get bigger to mirror installed RAM.

The size of the swap/pagefile is very OS/VM design dependant.
MS-WinNT & *BSD use an algorithm that requires every physical
page to have a spot in the pagefile. [May even require all of
VM]. Linux does not, and will happily run with a swapfile much
smaller than RAM.

These differences are very much like the difference between AMD &
Intel L2 caches. Intel L2 contains/mirrors all of L1. AMD uses
an exclusive[victim] L2 cache than has no L1 mirrored deliberately.

As Terje Metheson says "All programming is an exercise in caching".

RAM is just a big disk cache. The disk might be a network cache.

-- Robert

 

[Rod Speed]

Well, you are not required to believe things even
if they are true. It just helps your credibility.

We'll see...

Surely you can see that a suspend-to-disk [hibernation]
file will have to get bigger to mirror installed RAM.

Thats the hibernate file NOT THE SWAP FILE.

The size of the swap/pagefile is very OS/VM design
dependant. MS-WinNT & *BSD use an algorithm that
requires every physical page to have a spot in the pagefile.

Wrong. If that was so, and you have the swap file
set to automatic, its size would double if you double
the amount of physical ram, and it doesnt.

[May even require all of VM]. Linux does not, and will
happily run with a swapfile much smaller than RAM.

So does NT/2K/XP

These differences are very much like the
difference between AMD & Intel L2 caches.

Nope, nothing like.

Intel L2 contains/mirrors all of L1. AMD uses an exclusive
[victim] L2 cache than has no L1 mirrored deliberately.

As Terje Metheson says "All programming is an exercise in caching".

RAM is just a big disk cache.

Wrong again. What is in ram isnt necessarily on the drive.

The disk might be a network cache.

Mindlessly silly, most obviously with stuff specific to that PC.

 

[George Macdonald]

Why is this so though? Personally it seems that the more RAM I added,
the less swap space I set aside in proportion to the actual physical
RAM I have. After all, with all the RAM, it's supposed to help avoid
hitting the comparatively slow hard disk no?

If you're using swap space in any significant way you just don't have
enough memory for your workload, obviously. The thing is that swap space
is not a backup you allocate when needed - it's an integral part of the
memory management system; if you have a lot of real memory in anticipation
of running large memory jobs, you need the swap space. Growing it in an
"emergency", only when you actually have to page-out, is not the way to do
it - we saw that with Win98 and it's "elastic" swapfile.

In simple terms, if you have a process which has a page which due to its
"age" and demands from other processes has to be paged out, you don't want
the OS having to ask the file system to grow a 512MB page file out 2GB to
accomodate; even paging to HDD is supposed to be a fast, barely noticable
event.

Does anyone know whether the WinXP/2K3 pagefile.sys is a "sparse file"
under NTFS? This is one of the nutty things about Windows - other OSs,
which had a legacy of a real, full implemented file system, were using
"sparse files" for the paging file 20, maybe even 30, years ago. One gets
the feeling that M$ is just about to reinvent this... and probably try to
patent it.:-(

 

[Tony Hill]

Not if you dont use it fully.


Mindlessly silly. As the amount of physical ram
increases, the need for swap space DROPS.

In Linux yes, in Windows and *BSD, no. Read up on the differences
between how these OSes handle swap space. FWIW there isn't really a
"right"and a "wrong" way of doing things here, just two different
approaches.

Gets sillier by the minute.

Install 4GB of memory in a system sometime and boot into Windows. Now
tell me how much memory does Windows say you have? Any guesses on
where the remaining 700MB+ went?

 

[Tony Hill]

Why is this so though? Personally it seems that the more RAM I added,
the less swap space I set aside in proportion to the actual physical
RAM I have. After all, with all the RAM, it's supposed to help avoid
hitting the comparatively slow hard disk no?

If you are running Linux, then yes. If you're running Windows, then
no. Here's a quick explanation of why you should have at least as big
of a pagefile as you have physical memory:

http://blogs.msdn.com/larryosterman/archive/2004/03/18/92010.aspx

Note that pretty much every OS goes about handling swap space
differently. In Linux it works pretty much just like you expected,
more physical memory = less need for swap space.

Note that it's not really that one of these implementations is "right"
and the other "wrong", just slightly different ways of looking at
virtual memory.

 

[Rod Speed]

In Linux yes, in Windows and *BSD, no.

Fraid so.

Read up on the differences between how these OSes handle swap space.

Let XP handle the swap file size itself, double the amount of physical
ram in the system and watch what it does to the size of the swap file.

Then get a VERY large towel ready for the egg all over your face.

FWIW there isn't really a "right"and a "wrong" way
of doing things here, just two different approaches.

Its completely silly to require a swap file to double
in size when the amount of physical ram is doubled.

Install 4GB of memory in a system sometime and boot into Windows.
Now tell me how much memory does Windows say you have?
Any guesses on where the remaining 700MB+ went?

Irrelevant to what happens to the size of the swap
file when XP gets to manage it itself when you double
the amount of physical ram, say from 512M to 1G.

What you get with 4G is an artifact of how XP does the physical address space.