Price difference between Intel & AMD systems

On Sat, 18 Sep 2004 16:13:32 +0100, Franklin
wrote:

Is there a rough rule of thumb which indicates the price difference between
an AMD system and an Intel system of the same power?

In a word, no.

I am thinking of just the processor and mobo.
(I don't think memory depends on processor type)

To a certain extent memory does depend on the motherboard and/or
processor. For example, some setups (for both AMD and Intel) use
single channel memory while others use dual channel memory (ie memory
must be added in pairs). Some AMD systems, most notably the older
Socket 940 Athlon64 FX chips, require the use of registered memory,
while pretty much all others use unregistered memory.

Is it something like ... "Intel systems cost 25 to 30 percent more than an
equivalent AMD system"?

Well, first off, defining "equivalent" is not a very easy thing to do.
In some applications Intel's P4 design tends to do pretty well, while
in others AMD's AthlonXP line does well and in others still it's AMD's
Athlon64 line that really pulls ahead. So equivalency here depends
largely on what application is most important to you.

What's more, prices are rather fluid and tend to change a lot
depending on where in the price/performance scale you are looking.
For example, Intel's top-end P4 Extreme Edition chips are VERY
expensive ($900+), and generally perform about the same as an Athlon64
3500+ ($365) or 3700+ ($500). On the other hand, if you were to
compare a P4 3.0GHz, it would usually perform more or less on par with
AMD's Athlon64 3000+ (again, depending on the applications you use),
where here AMD's processor is only about $20 cheaper.

-------------
Tony Hill
hilla underscore 20 at yahoo dot ca

Tony Hill wrote:

On Sat, 18 Sep 2004 16:13:32 +0100, Franklin
wrote:

Is there a rough rule of thumb which indicates the price difference between
an AMD system and an Intel system of the same power?

In a word, no.

I am thinking of just the processor and mobo.
(I don't think memory depends on processor type)

To a certain extent memory does depend on the motherboard and/or
processor. For example, some setups (for both AMD and Intel) use
single channel memory while others use dual channel memory (ie memory
must be added in pairs). Some AMD systems, most notably the older
Socket 940 Athlon64 FX chips, require the use of registered memory,
while pretty much all others use unregistered memory.

Is it something like ... "Intel systems cost 25 to 30 percent more than an
equivalent AMD system"?

Well, first off, defining "equivalent" is not a very easy thing to do.
In some applications Intel's P4 design tends to do pretty well, while
in others AMD's AthlonXP line does well and in others still it's AMD's
Athlon64 line that really pulls ahead. So equivalency here depends
largely on what application is most important to you.

What's more, prices are rather fluid and tend to change a lot
depending on where in the price/performance scale you are looking.
For example, Intel's top-end P4 Extreme Edition chips are VERY
expensive ($900+), and generally perform about the same as an Athlon64
3500+ ($365) or 3700+ ($500). On the other hand, if you were to
compare a P4 3.0GHz, it would usually perform more or less on par with
AMD's Athlon64 3000+ (again, depending on the applications you use),
where here AMD's processor is only about $20 cheaper.

Which wouldn't be so bad if the Pentium 4 being discussed was a 64 bit one.
Unfortunately it is a 32 bit one. Assigning no extra value to the Athlon 64's
64 bit mode doesn't seem to make much sense. In 2005 many of those
who bought a high priced 32 bit processor in '04 might become upset
that they didn't use foresight and buy a 64 bit processor. I wonder what great
64 bit applications we will see in 2005. I wonder what 32 bit applications will
be ported to 64 bits and show tremendous improvements in performance
when the 64 bit is run compared to the 32 bit version on an Athlon 64
or Opteron. Here is a link to one application already out in 64 bits whose
64 bit version runs 25% faster than the 32 bit version on an Athlon 64.

http://www.short-media.com/review.php?r=257&p=1


Other applications might show a much greater performance increase.



-------------
Tony Hill
hilla underscore 20 at yahoo dot ca

JK wrote:

Which wouldn't be so bad if the Pentium 4 being discussed was a 64 bit one.
Unfortunately it is a 32 bit one. Assigning no extra value to the Athlon 64's
64 bit mode doesn't seem to make much sense. In 2005 many of those
who bought a high priced 32 bit processor in '04 might become upset
that they didn't use foresight and buy a 64 bit processor. I wonder what great
64 bit applications we will see in 2005. I wonder what 32 bit applications will
be ported to 64 bits and show tremendous improvements in performance
when the 64 bit is run compared to the 32 bit version on an Athlon 64
or Opteron. Here is a link to one application already out in 64 bits whose
64 bit version runs 25% faster than the 32 bit version on an Athlon 64.

Perhaps similar to the reaction of the "let's buy 64-bit because it's
the latest and greatest thing" brigade when technologies such as BTX,
PCI-Express, faster FSBs, new sockets etc hit the market en-mass in
2005 and render their "latest and greatest" machines obsolete. Then
they might well wonder why they didn't buy 32-bit machines, have
saving themselves some money and wait until they actually needed it?

--
iv Paul iv

Paul Hopwood wrote:

JK wrote:

Which wouldn't be so bad if the Pentium 4 being discussed was a 64 bit one.
Unfortunately it is a 32 bit one. Assigning no extra value to the Athlon 64's
64 bit mode doesn't seem to make much sense. In 2005 many of those
who bought a high priced 32 bit processor in '04 might become upset
that they didn't use foresight and buy a 64 bit processor. I wonder what great
64 bit applications we will see in 2005. I wonder what 32 bit applications will
be ported to 64 bits and show tremendous improvements in performance
when the 64 bit is run compared to the 32 bit version on an Athlon 64
or Opteron. Here is a link to one application already out in 64 bits whose
64 bit version runs 25% faster than the 32 bit version on an Athlon 64.

Perhaps similar to the reaction of the "let's buy 64-bit because it's
the latest and greatest thing" brigade when technologies such as BTX,
PCI-Express, faster FSBs, new sockets etc hit the market en-mass in
2005 and render their "latest and greatest" machines obsolete. Then
they might well wonder why they didn't buy 32-bit machines, have
saving themselves some money and wait until they actually needed it?

Buying a low priced 32 bit Athlon XP or Sempron might make sense,
especially for someone who runs only business software. Buying a 32
bit Pentium 4 at around the price of an Athlon 64 doesn't make much
sense for most people(notice I said most people, as there will be a few
who will say that more than 50% of their pc usage is video editing, and
they have no plans to ever want to upgrade to 64 bit editing software).



--
iv Paul iv

On Sun, 19 Sep 2004 19:44:50 -0400, JK wrote:

Tony Hill wrote:

Well, first off, defining "equivalent" is not a very easy thing to do.
In some applications Intel's P4 design tends to do pretty well, while
in others AMD's AthlonXP line does well and in others still it's AMD's
Athlon64 line that really pulls ahead. So equivalency here depends
largely on what application is most important to you.

What's more, prices are rather fluid and tend to change a lot
depending on where in the price/performance scale you are looking.
For example, Intel's top-end P4 Extreme Edition chips are VERY
expensive ($900+), and generally perform about the same as an Athlon64
3500+ ($365) or 3700+ ($500). On the other hand, if you were to
compare a P4 3.0GHz, it would usually perform more or less on par with
AMD's Athlon64 3000+ (again, depending on the applications you use),
where here AMD's processor is only about $20 cheaper.

Which wouldn't be so bad if the Pentium 4 being discussed was a 64 bit one.
Unfortunately it is a 32 bit one. Assigning no extra value to the Athlon 64's
64 bit mode doesn't seem to make much sense. In 2005 many of those
who bought a high priced 32 bit processor in '04 might become upset
that they didn't use foresight and buy a 64 bit processor. I wonder what great
64 bit applications we will see in 2005.

Don't hold your breath for too much. 64-bit is nothing new, it's been
around for 10+ years in other processors. The benefits and drawbacks
are well known. Usually those drawbacks (pointers twice as large and
therefore twice as much memory use/cache use/bandwidth use) outweigh
the benefits and applications tend to be slower unless you really need
64-bit integers (very rare for most apps) or you need more than ~2GB
of addressable memory (the real reason for 64-bit).

Of course, all is not equal in x86-64, as AMD also did a bit of
tidying and doubled the number of integer registers. This will tend
to make applications about 5% to 10% faster. For example, for SPEC
CINT2000 base, AMD showed an 8.9% improvement overall. However in
that 8.9% improvement there were three tests (181.mcf, 197.parser and
300.twolf) that ran slower, two that ran MUCH faster (186.crafty was
41% faster while 252.eon was 49% faster), and all the rest that were a
little bit faster.

Of course, all this will be for naught for 95%+ of all users if
Microsoft doesn't get their act together and get WinXP for x64
released sometime this decade.

I wonder what 32 bit applications will
be ported to 64 bits and show tremendous improvements in performance
when the 64 bit is run compared to the 32 bit version on an Athlon 64
or Opteron. Here is a link to one application already out in 64 bits whose
64 bit version runs 25% faster than the 32 bit version on an Athlon 64.

http://www.short-media.com/review.php?r=257&p=1

More important than the 25% improvement is the issue they ran into on
the first page, some things were just not possible on a 32-bit machine
due to lack of memory address space. Kind of flies in the face of
those who say 64-bit is not necessary on the desktop for the next 5+
years.

Other applications might show a much greater performance increase.

Some will. Some applications will show a 100% improvement in
performance. Others could easily show a 10% loss in performance.
Most will be about 5-10% faster. Not much, but it's free, so hey, why
not?

I do tend to agree with you, AMD's processors are often a better buy
these days even if the price is the same. However, that wasn't what
the original poster asked.

-------------
Tony Hill
hilla underscore 20 at yahoo dot ca

Tony Hill wrote:

On Sun, 19 Sep 2004 19:44:50 -0400, JK wrote:

Tony Hill wrote:

Well, first off, defining "equivalent" is not a very easy thing to do.
In some applications Intel's P4 design tends to do pretty well, while
in others AMD's AthlonXP line does well and in others still it's AMD's
Athlon64 line that really pulls ahead. So equivalency here depends
largely on what application is most important to you.

What's more, prices are rather fluid and tend to change a lot
depending on where in the price/performance scale you are looking.
For example, Intel's top-end P4 Extreme Edition chips are VERY
expensive ($900+), and generally perform about the same as an Athlon64
3500+ ($365) or 3700+ ($500). On the other hand, if you were to
compare a P4 3.0GHz, it would usually perform more or less on par with
AMD's Athlon64 3000+ (again, depending on the applications you use),
where here AMD's processor is only about $20 cheaper.

Which wouldn't be so bad if the Pentium 4 being discussed was a 64 bit one.
Unfortunately it is a 32 bit one. Assigning no extra value to the Athlon 64's
64 bit mode doesn't seem to make much sense. In 2005 many of those
who bought a high priced 32 bit processor in '04 might become upset
that they didn't use foresight and buy a 64 bit processor. I wonder what great
64 bit applications we will see in 2005.

Don't hold your breath for too much. 64-bit is nothing new, it's been
around for 10+ years in other processors.

Do any of the other 64 bit chips have integrated memory controllers?

The benefits and drawbacks
are well known.

Are they?

Usually those drawbacks (pointers twice as large and
therefore twice as much memory use/cache use/bandwidth use) outweigh
the benefits and applications tend to be slower unless you really need
64-bit integers (very rare for most apps) or you need more than ~2GB
of addressable memory (the real reason for 64-bit).

Of course, all is not equal in x86-64, as AMD also did a bit of
tidying and doubled the number of integer registers. This will tend
to make applications about 5% to 10% faster. For example, for SPEC
CINT2000 base, AMD showed an 8.9% improvement overall. However in
that 8.9% improvement there were three tests (181.mcf, 197.parser and
300.twolf) that ran slower, two that ran MUCH faster (186.crafty was
41% faster while 252.eon was 49% faster), and all the rest that were a
little bit faster.

Of course, all this will be for naught for 95%+ of all users if
Microsoft doesn't get their act together and get WinXP for x64
released sometime this decade.

I wonder what 32 bit applications will
be ported to 64 bits and show tremendous improvements in performance
when the 64 bit is run compared to the 32 bit version on an Athlon 64
or Opteron. Here is a link to one application already out in 64 bits whose
64 bit version runs 25% faster than the 32 bit version on an Athlon 64.

http://www.short-media.com/review.php?r=257&p=1

More important than the 25% improvement is the issue they ran into on
the first page, some things were just not possible on a 32-bit machine
due to lack of memory address space. Kind of flies in the face of
those who say 64-bit is not necessary on the desktop for the next 5+
years.

Other applications might show a much greater performance increase.

Some will. Some applications will show a 100% improvement in
performance. Others could easily show a 10% loss in performance.
Most will be about 5-10% faster. Not much, but it's free, so hey, why
not?

I do tend to agree with you, AMD's processors are often a better buy
these days even if the price is the same. However, that wasn't what
the original poster asked.

-------------
Tony Hill
hilla underscore 20 at yahoo dot ca

JK wrote:
Don't hold your breath for too much. 64-bit is nothing new, it's
been around for 10+ years in other processors.

Do any of the other 64 bit chips have integrated memory controllers?

The Sun UltraSparc 3 & 4 do. I think one of the Alphas had an integrated
controller too, but it was a RAMBUS controller, so it wasn't quite the same
thing as what's currently needed.

Yousuf Khan

On Mon, 20 Sep 2004 07:55:30 -0400, JK wrote:

Tony Hill wrote:

Don't hold your breath for too much. 64-bit is nothing new, it's been
around for 10+ years in other processors.

Do any of the other 64 bit chips have integrated memory controllers?

Yes, most of them do these days. Sun UltraSparc 3 and US4, the HPaq
Alpha EV7 and IBM Power5 all have integrated memory controllers. I'm
fairly certain that there is at least one 64-bit MIPS core out there
with an integrated memory controller.

Nothing particularly unique about integrated memory controllers or
64-bit, the only thing that Athlon64/Opteron do is bring these
technologies alongside the x86 instruction set and a much lower price
than most others.

The benefits and drawbacks
are well known.

Are they?

Yes, in fact they are. There's really nothing new about 64-bit, same
idea has been used in virtually every processor architecture on the
planet other than x86 for the past 10 years.

-------------
Tony Hill
hilla underscore 20 at yahoo dot ca

Ed wrote:
Does Intel have a CPU in the works that use an integrated memory
controller?
thanks,
Ed

There is rumours that it is working on it, yup. Some might even take the
fact that it's working on the FB-DIMM specification as a roundabout proof of
it. FB-DIMM has the potential to mask all technological differences between
different generations of DRAM. So a single memory controller that controls
DDR-RAM might be enough to control DDR2, or whatever else comes around in
the future.

Yousuf Khan

On Tue, 21 Sep 2004 13:41:53 -0500, Ed wrote:

On Mon, 20 Sep 2004 21:11:32 -0400, Tony Hill
wrote:
snip
Do any of the other 64 bit chips have integrated memory controllers?

Yes, most of them do these days. Sun UltraSparc 3 and US4, the HPaq
Alpha EV7 and IBM Power5 all have integrated memory controllers. I'm
fairly certain that there is at least one 64-bit MIPS core out there
with an integrated memory controller.

Does Intel have a CPU in the works that use an integrated memory
controller?

To the best of my knowledge, no. Err, well I guess their ARM chips
(XScale) have integrated memory controllers, but I'm guessing that
wasn't what you were thinking of.

Of course, I don't have any sort of secret insight into just what
Intel is doing. I would be rather surprised if they had not at least
looking into integrating memory controllers on their chips. There is
a certain trade-off between the added performance and reduced
flexibility that doing so entails, but all evidence seems to be
pointing the former is plenty reason enough to go down this path.

-------------
Tony Hill
hilla underscore 20 at yahoo dot ca