2800XP V's 2500XP

On Sun, 24 Aug 2003 13:49:47 +0000, J.Clarke wrote:

On Sun, 24 Aug 2003 15:20:49 GMT
"Wes Newell" wrote:

On Sun, 24 Aug 2003 12:24:14 +0000, SM wrote:

Is it worth spending the extra £60 for the higher speed processor(
both Barton Cores)

Of course not. The answer is in your question. They are both Barton
cores. IOW's, they are the same cpu, just with different defaults.
Would you pay 50% more for a tire that was inflated to the max
compared to the same tire that was inflated to the minimum tire
pressure, when inflating it to the max would be the same as the first?
Basically, that is your question.

Not quite. Would you pay extra for a tire that was guaranteed to take
full inflation pressure vs one that is guaranteed only to take 10 psi
less than full inflation pressure? Or would you save a buck and pray
that the cheap tire was merely mismarked rather than being actually
unable to take the pressure?

The tires were the same. That means same brand and model. Only one was
completely aired up. Note "same tire" above. Your response makes no
sense.

While all Bartons may be the same design and made in the same facility,
that does not mean that all will run at the same clock speed. And the

That's true, some rated at slower clock speeds will actually run faster
than some rated at higher clock speeds. Proven fact.

fact that you haven't personally encountered one that doesn't is not
relevant to the general statement. Personally I've never met a
sub-Saharan African with AIDS but that doesn't mean that they don't
exist in large numbers.

I think you're getting a little carried away John.:-)

--
Abit KT7-Raid (KT133) Tbred B core CPU @2400MHz (24x100FSB)
http://mysite.verizon.net/res0exft/cpu.html

On Sun, 24 Aug 2003 15:29:27 +0000, SM wrote:


"Wes Newell" wrote in message
news
On Sun, 24 Aug 2003 12:24:14 +0000, SM wrote:

Is it worth spending the extra £60 for the higher speed processor( both
Barton Cores)

Of course not. The answer is in your question. They are both Barton cores.
IOW's, they are the same cpu, just with different defaults.
Would you pay 50% more for a tire that was inflated to the max compared to
the same tire that was inflated to the minimum tire pressure, when
inflating it to the max would be the same as the first? Basically, that is
your question.

--
Abit KT7-Raid (KT133) Tbred B core CPU @2400MHz (24x100FSB)
http://mysite.verizon.net/res0exft/cpu.html

OT
go to tire tread NG????

So tell me if I'm wrong
1.You would buy the 2500

That's correct.

2.your tire pressure is low

Probably. I haven't driven my car in a few weeks.:-)

3.why pay over-inflated??? prices for basically the same core

Exactly.

--
Abit KT7-Raid (KT133) Tbred B core CPU @2400MHz (24x100FSB)
http://mysite.verizon.net/res0exft/cpu.html

His response makes a lot of sense. No two tires are the same. The same
applies to CPUs. There is no guarantee that you'll get the 2500 to run @
2800 speeds. I'm not sure if AMD does the same testing which Intel used to
do. Intel would take processors to see if they passed a set number of
tests. If the processor fails, they tested the same processor with lower
settings ... and so on. Now if you know that AMD ran only one set of tests
and marked the processor speeds randomly, you would be correct. Does anyone
know how the testing is done?

-------------------------
Politicians are used car salesmen who have the public speaking gene
"Wes Newell" wrote in message
news
On Sun, 24 Aug 2003 13:49:47 +0000, J.Clarke wrote:

On Sun, 24 Aug 2003 15:20:49 GMT
"Wes Newell" wrote:

On Sun, 24 Aug 2003 12:24:14 +0000, SM wrote:

Is it worth spending the extra £60 for the higher speed processor(
both Barton Cores)

Of course not. The answer is in your question. They are both Barton
cores. IOW's, they are the same cpu, just with different defaults.
Would you pay 50% more for a tire that was inflated to the max
compared to the same tire that was inflated to the minimum tire
pressure, when inflating it to the max would be the same as the first?
Basically, that is your question.

Not quite. Would you pay extra for a tire that was guaranteed to take
full inflation pressure vs one that is guaranteed only to take 10 psi
less than full inflation pressure? Or would you save a buck and pray
that the cheap tire was merely mismarked rather than being actually
unable to take the pressure?

The tires were the same. That means same brand and model. Only one was
completely aired up. Note "same tire" above. Your response makes no
sense.

While all Bartons may be the same design and made in the same facility,
that does not mean that all will run at the same clock speed. And the

That's true, some rated at slower clock speeds will actually run faster
than some rated at higher clock speeds. Proven fact.

fact that you haven't personally encountered one that doesn't is not
relevant to the general statement. Personally I've never met a
sub-Saharan African with AIDS but that doesn't mean that they don't
exist in large numbers.

I think you're getting a little carried away John.:-)

--
Abit KT7-Raid (KT133) Tbred B core CPU @2400MHz (24x100FSB)
http://mysite.verizon.net/res0exft/cpu.html

In article ers.com,
"The TweakOholic" wrote:

His response makes a lot of sense. No two tires are the same.

While the tire analogy might have not been absolutely flawless, the
point was clear and I think iy's a tad anal to critique it under a micrscope.

Perhaps 2 motors, virtually with the same potential, one with a governor
on, the other not ?

Jim

On Mon, 25 Aug 2003 00:15:51 +0000, J.Clarke wrote:

On Mon, 25 Aug 2003 01:43:40 GMT
"Wes Newell" wrote:

I think you're getting a little carried away John.:-)

No, Wes, I'm not getting carried away. Anyone who has had any real
engineering experience knows that designing a system so that a component
is running out of spec is eventually going to bite you in the ass. And
that is exactly what you are doing any time you overclock.

While this may hold true in bridge building, it has no impact when it
comes to these cpu's. There are all built to the same specs. They may not
all be tested the same, and one may clock higher than another, regardless
of the model number. A 3200+ is just as likely to fail as a 2500+ clocked
to the same speed, assuming it was thoroughly tested at that speed by
whiever clocks it. The IC industry has been doding the same thing for at
least 20 years. They build a part and then test to speed the order is for.
There's nothing thrown in to make one part slower than the other, They all
come from the same design. Now if you get a 2500+ that won't clock to a
3200+ speed, then that's another story.

You have some good advice for people, but your continued insistence that
every processor in the world is going to work the same way that the few
that you have personally encountered work is doing no service to
anybody.

I'm not sure what you think a few is, but I've probably worked with at
least a couple of hundred. I can't recall one that wouldn't clock up to or
close to the top speed of that core.

When you have a thousand of those all running the same way reliably,
with no rejects, then maybe I'll be impressed.

Should have read this first. Nope, not that many.

--
Abit KT7-Raid (KT133) Tbred B core CPU @2400MHz (24x100FSB)
http://mysite.verizon.net/res0exft/cpu.html

On Mon, 25 Aug 2003 00:18:39 +0000, J.Clarke wrote:

On Mon, 25 Aug 2003 01:57:19 GMT
"Wes Newell" wrote:

On Sun, 24 Aug 2003 13:54:18 +0000, J.Clarke wrote:

Unlike Wes, I don't believe in overclocking everything in sight.
There are times when it is appropriate and others when it is not.

I don't even consider it overclocking until I go over the speed that
the manufacturer sells the core at.

I see. So you just willy-nilly assume that every processor of a given
design is going to perform exactly the same way as every other processor
of that design.

Come on now, I never said that. In fact, I said just the opposite. Some
will clock higher than others.

I have mixed emotions about my wishes for you in this
regard--on the one hand I hope you continue to get away with this, on
the other hand it will serve you right if the time you get audited is
the time that the arithmetic functions on your overclocked processor are
off just enough to get you nailed with a half a million dollar fine.

I'm retired.:-)
But as I've stated before. I don't overclock critical operations machines.

There is something called a "manufacturing tolerance" with which you
might wish to familiarize yourself.

I designed and built memory controllers from ttl. I'm a litlle familiar
with tolerances.:-)

--
Abit KT7-Raid (KT133) Tbred B core CPU @2400MHz (24x100FSB)
http://mysite.verizon.net/res0exft/cpu.html

On Mon, 25 Aug 2003 07:30:50 GMT
"Wes Newell" wrote:

On Mon, 25 Aug 2003 00:15:51 +0000, J.Clarke wrote:

On Mon, 25 Aug 2003 01:43:40 GMT
"Wes Newell" wrote:

I think you're getting a little carried away John.:-)

No, Wes, I'm not getting carried away. Anyone who has had any real
engineering experience knows that designing a system so that a
component is running out of spec is eventually going to bite you in
the ass. And that is exactly what you are doing any time you
overclock.

While this may hold true in bridge building, it has no impact when it
comes to these cpu's.

It holds true in _every_ kind of engineering.

There are all built to the same specs.

They are built to the same drawings, with a tolerance range. Some come
closer to the middle of the range, some are on the edges. The ones on
the edges of the range will not perform as well as the ones that are in
the middle of the range.

They may
not all be tested the same, and one may clock higher than another,
regardless of the model number.

If they are all identical then why are they sold with different speed
ratings?

A 3200+ is just as likely to fail as a
2500+ clocked to the same speed, assuming it was thoroughly tested at
that speed by whiever clocks it.

Do you have a definitive test cycle that will guarantee that any chip
will perform identically to one with a higher rating from the
manufacturer under _all_ circumstances? If so, please be kind enough to
publish it, and publish the evidence on which you base the contention
that your test procedure is adequate.

The IC industry
has been doding the same thing for at least 20 years. They build a
part and then test to speed the order is for.

No, they build a part and then test to determine which speed bin to put
it in. In some cases, when they have more orders for a low-speed part
than they have parts binned at that speed from a given run they'll
downgrade part of a lot. But it doesn't always happen and all parts are
not downgraded.

There's nothing thrown
in to make one part slower than the other, They all come from the same
design.

They all come from the same design, but they don't all come from the
same identical dopant composition--there are tolerances there. And they
don't all come from the same identical dopant penetration, there are
tolerances there too. And they don't all have the identical crystal
structure--there are tolerances there. And they don't all have the same
identical feature dimensions--there are tolerances there as well.

Now it may be that AMD has their process tuned to the point that all
parts produced will pass the tests for the highest speed, or it may not.
But unless you work for AMD you have no way of knowing with certainty
that that is the case.

Now if you get a 2500+ that won't clock to a 3200+ speed, then
that's another story.

How do you determine that a given processor will clock to 3200+ speed?
Modern processors have an immense number of states--take the
transistor count and raise 2 to that power and you'll have an
approximation--if you aren't privy to the design you're not going to be
able to figure out which ones represent"worst case" and you aren't going
to live long enough to test all states of a single processor, so how do
you know that your tests are adequate?

You have some good advice for people, but your continued insistence
that every processor in the world is going to work the same way that
the few that you have personally encountered work is doing no
service to anybody.

I'm not sure what you think a few is, but I've probably worked with at
least a couple of hundred. I can't recall one that wouldn't clock up
to or close to the top speed of that core.

And on the basis of "a couple of hundred" you'll generalize to all the
millions that are produced, under all operating conditions and you'll
build mission-critical hardware with those out of spec processors?
You'll trust lives to those processors? You'll trust billions of
dollars to those processors?

When you have a thousand of those all running the same way reliably,
with no rejects, then maybe I'll be impressed.

Should have read this first. Nope, not that many.

Another question. Are those processors all running reliably at "the top
speed of that core" at the maximum rated temperature for the particular
chip? The ones that are labelled with a particular speed are guaranteed
to run at that temperature.

--
Abit KT7-Raid (KT133) Tbred B core CPU @2400MHz (24x100FSB)
http://mysite.verizon.net/res0exft/cpu.html


--
--
--John
Reply to jclarke at ae tee tee global dot net
(was jclarke at eye bee em dot net)

Thanks everyone for their help, 2500 barton on its way to me as we speak.

SM

On Mon, 25 Aug 2003 08:04:48 +0000, J.Clarke wrote:

If they are all identical then why are they sold with different speed
ratings?

Now that's one's real simple. Supply and demand. When their customers are
odering 1700+'s, they default them to 1700+'s, and test them for that
speed. When they order 2600+'s, ditto. Otherwise, you wouldn't see any
more cpu's from amd under 2600+'s in the XP line. Amd's actual cost is the
same for the 1700+ as it is for the 2600+. It's all about supply and
demand. Always has been, and always will be. As far the Barton, same thing
holds true. Their cost for a 2500+ is the same as for a 3200+. Our cost
however is considerably more.

A 3200+ is just as likely to fail as a
2500+ clocked to the same speed, assuming it was thoroughly tested at
that speed by whiever clocks it.

Do you have a definitive test cycle that will guarantee that any chip
will perform identically to one with a higher rating from the
manufacturer under _all_ circumstances? If so, please be kind enough to
publish it, and publish the evidence on which you base the contention
that your test procedure is adequate.

There are many test programs that are available. Me, I just get it to post
at it's highest speed, run memtest, then clock it down a little to be on
the safe side. I don't thnik you'll find what you are looking for
anywhere, including at amd. If it were available, I don't think you'd have
seen the recalls that have been put out by Intel for their cpu's.

The IC industry
has been doding the same thing for at least 20 years. They build a
part and then test to speed the order is for.

No, they build a part and then test to determine which speed bin to put
it in. In some cases, when they have more orders for a low-speed part
than they have parts binned at that speed from a given run they'll
downgrade part of a lot. But it doesn't always happen and all parts are
not downgraded.

Then there are a lot of downgraded 1700+'s out there.:-)

There's nothing thrown
in to make one part slower than the other, They all come from the same
design.

They all come from the same design, but they don't all come from the
same identical dopant composition--there are tolerances there. And they
don't all come from the same identical dopant penetration, there are
tolerances there too. And they don't all have the identical crystal
structure--there are tolerances there. And they don't all have the same
identical feature dimensions--there are tolerances there as well.

Now it may be that AMD has their process tuned to the point that all
parts produced will pass the tests for the highest speed, or it may not.
But unless you work for AMD you have no way of knowing with certainty
that that is the case.

Just look at the numbers. How many Tbred B cores have you seen reporting
to run well over 2200MHz. Now how many have you heard people complain that
they can't get over 2000MHz out of one. I haven't heard of a one. The
proof is in the numbers.

Now if you get a 2500+ that won't clock to a 3200+ speed, then
that's another story.

How do you determine that a given processor will clock to 3200+ speed?

This is the way I do it. I get it stable as high as possible, then back it
down at least 5%.

Modern processors have an immense number of states--take the
transistor count and raise 2 to that power and you'll have an
approximation--if you aren't privy to the design you're not going to be
able to figure out which ones represent"worst case" and you aren't going
to live long enough to test all states of a single processor, so how do
you know that your tests are adequate?

There are no garauntees that running at the rated speeds won't cause
unforeseen problems either. It's all the same.

And on the basis of "a couple of hundred" you'll generalize to all the
millions that are produced, under all operating conditions and you'll
build mission-critical hardware with those out of spec processors?

No John, You're trying to put words in my mouth that I didn't say. When in
fact, I've stated just the opposite for critical applications.

You'll trust lives to those processors? You'll trust billions of
dollars to those processors?

Now you know I never stated this.

Another question. Are those processors all running reliably at "the top
speed of that core" at the maximum rated temperature for the particular
chip? The ones that are labelled with a particular speed are guaranteed
to run at that temperature.

Anyone that would run a cpu at it's top rated temp is an idiot, just
asking for trouble. You know that as well as I. As for the garauntee. Just
try it. You'll be sending a lot of cpu's back. They will fail prematurely.
You'll have to take my personal garauntee on that. And of course, the
widely known fact that heat kills in the electronic component.

--
Abit KT7-Raid (KT133) Tbred B core CPU @2400MHz (24x100FSB)
http://mysite.verizon.net/res0exft/cpu.html

J.Clarke wrote:
A 3200+ is just as likely to fail as a
2500+ clocked to the same speed, assuming it was thoroughly tested at
that speed by whiever clocks it.


Well... not quite, but I'd expect it to be pretty much the situation in this
case. Depends what proportion of Barton cores fail the 3200+ test, I doubt
it's that many after they've been producing 3200+ for a little while.

Do you have a definitive test cycle that will guarantee that any chip
will perform identically to one with a higher rating from the
manufacturer under _all_ circumstances? If so, please be kind enough
to publish it, and publish the evidence on which you base the
contention that your test procedure is adequate.

Does AMD have a definitive test cycle? Didn't think so.

However, when running a prime95 test my 2500+ running at 3200+ speeds it
does seem to consistently fail on one of the calculations. I haven't tried
it at 2500+ speeds yet.

Ben
--
I'm not just a number. To many, I'm known as a String...