For the last time, AMD DOES NOT base its PRs on P4s

[Wingznut]

In the initial article that Ilmater linked, Richard Baker said: "I was wondering why you thought that AMD moves the goal posts on model numbers? The basis on which they are calculated is the same as it has always been, neither the applications nor their relative weightings have been changed since the original 1.4 Ghz Athlon."


Ok, let's look at some numbers. These are the "model number" with the respective actual clockspeed.

• 1500+ = 1333mhz
• 1600+ = 1400mhz
• 1700+ = 1466mhz
• 1800+ = 1533mhz
• 1900+ = 1600mhz
• 2000+ = 1666mhz
• 2100+ = 1733mhz
• 2200+ = 1800mhz

So far, there's a distinct pattern. 66 actual mhz = +100 PR. Even though it isn't logical because they are saying that this cpu scales 50% faster than whatever they are baselining it with. What I mean is: At 1333mhz, they say the "1500+" is ~12.5% faster than whatever they are comparing it to. But at 1800mhz, the "2200+" is now 22% faster, even though there are ABSOLUTELY ZERO architectural improvements. Nonetheless, we'll move on.

Here's where things change, for no apparent reason. Going with the pattern, you'd think that a "2400+" would be 1933mhz, right? Well, AMD disagrees and decided to give their 2000mhz cpu a "2400+" rating. If nothing has changed, then why the discrepancy??? Did the cpu all of a sudden get slower at 2000mhz? Of course not. Would the T-bird have gotten relatively faster at 2400mhz? Of course not.

Let's just say if their engineers used this kind of methodology, they'd be out of work in a big hurry.

 

[ElFenix]

Originally posted by: Wingznut
In the initial article that Ilmater linked, Richard Baker said: "I was wondering why you thought that AMD moves the goal posts on model numbers? The basis on which they are calculated is the same as it has always been, neither the applications nor their relative weightings have been changed since the original 1.4 Ghz Athlon."


Ok, let's look at some numbers. These are the "model number" with the respective actual clockspeed.

• 1500+ = 1333mhz
• 1600+ = 1400mhz
• 1700+ = 1466mhz
• 1800+ = 1533mhz
• 1900+ = 1600mhz
• 2000+ = 1666mhz
• 2100+ = 1733mhz
• 2200+ = 1800mhz

So far, there's a distinct pattern. 66 actual mhz = +100 PR. Even though it isn't logical because they are saying that this cpu scales 50% faster than whatever they are baselining it with. What I mean is: At 1333mhz, they say the "1500+" is ~12.5% faster than whatever they are comparing it to. But at 1800mhz, the "2200+" is now 22% faster, even though there are ABSOLUTELY ZERO architectural improvements. Nonetheless, we'll move on.

Here's where things change, for no apparent reason. Going with the pattern, you'd think that a "2400+" would be 1933mhz, right? Well, AMD disagrees and decided to give their 2000mhz cpu a "2400+" rating. If nothing has changed, then why the discrepancy??? Did the cpu all of a sudden get slower at 2000mhz? Of course not. Would the T-bird have gotten relatively faster at 2400mhz? Of course not.

Let's just say if their engineers used this kind of methodology, they'd be out of work in a big hurry.

i think the first part of your post explains the second part.

 

[dullard]

Here's where things change, for no apparent reason. Going with the pattern, you'd think that a "2400+" would be 1933mhz, right? Well, AMD disagrees and decided to give their 2000mhz cpu a "2400+" rating. If nothing has changed, then why the discrepancy??? Did the cpu all of a sudden get slower at 2000mhz? Of course not. Would the T-bird have gotten relatively faster at 2400mhz? Of course not.

Let's just say if their engineers used this kind of methodology, they'd be out of work in a big hurry.

AMD has used several formulas (in order of release date):
PR=frequency*1.5 - 500 (original XPs)
PR=frequency*1.5 - 600 (T-breds)
PR=frequency*1.5 - 250 (2500+ and 3000+ Barton)
PR=frequency*1.5 - 325 (2800+ Barton)
and I left school so I don't have the numbers in front of me, but I think for one chip they used:
PR=frequency*1.5 - 300.

In each case they are multiplying by 1.5. So lets go backwards (and thus use their first formula). An underclocked (unlocked multiplier) 1800+ XP at 1.0 GHz would have a PR rating of 1000+. That is the same performance as a 1.0 GHz T-bird. But that is with all the architectural improvement the XP had, this cannot be right. Ok lets underclock it more to 900 MHz. It would then get a PR rating of 850 - slower than the T-bird it is supposedly based on. This is just silly, but can be done, and shows that (1) either the original XP basis isn't the T-bird or (2) that their PR formula isn't too accurate. Note: I realize extrapolating can be erroneous, but that is why we can underclock a chip and actually test the speeds.

Why do the 3 Barton chips use two different formulas? Aren't they the same architecture? That seems odd too.

 

[Wingznut]

Originally posted by: dullard
...shows that (1) either the original XP basis isn't the T-bird or (2) that their PR formula isn't too accurate.

or (3) both.

 

[dullard]

Originally posted by: Wingznut
In the initial article that Ilmater linked, Richard Baker said: "I was wondering why you thought that AMD moves the goal posts on model numbers? The basis on which they are calculated is the same as it has always been, neither the applications nor their relative weightings have been changed since the original 1.4 Ghz Athlon."

But if the applications didn't change, why are there more benchmarks in the latest validation? Maybe they didn't use it in the formula, but then why even run the extra benchmark?

What are these relative weightings? If only we knew, then we could figure out EXACTLY what the basis is. We could even give the P4 and the T-bird a PR rating. That would settle things once and for all. But alas we don't know the relative weightings. Since some benchmarks were run 3 or 4 times (different versions of each benchmark program and different numbers of benchmarks depending on the CPU that AMD tested) do we count them all, or do they each get a 1/3 or 1/4 weight? We just don't know.

 

[Wingznut]

Originally posted by: Ilmater
This is, in fact, a true statement. If you remember, clock for clock, the T-Birds were fairly evenly matched to the P4s. That's why the PR system seems to follow the P4 architecture fairly closely.

Heh, not even THIS Intel employee would make that claim. But if you don't believe me, check out Anand's comparison of the P4-1.4 and T-bird 1.2ghz.

 

[Bovinicus]

It do beleive they base it off the TBird. The 1500+ is a 1333MHz CPU, correct? A 1333MHz XP is way faster than a 1.5GHz P4, but closer to the performance of a 1.5GHz TBird.

Anand's review of the Athlon XP shows the 1.33GHz XP (1500+) performing around the level of a 1.7-2GHz P4. However, it performs only a little better than the 1.4GHz TBird.

However, the REASON for making the rating system is because of the P4. The MEANS of doing so is by benching it against a P4.

 

[Ilmater]

Originally posted by: paralazarguer
Argh. It's pointless to even try to participate in a thread with Ilmater. he thinks that he's the god of computers and can make broad sweeping judgements and no one else can have a differing opinion. He starts entire threads the purpose is to set EVERYONE straight ONCE AND FOR ALL on the ammount of ram that everyone needs or AMD PR system or whatever. There's a word for that...
Anyway, it's not even worth the effort.

Why is it that our opinions differ so much? I've started very few threads that didn't just involve questions I had. In that RAM thread, I admitted that I was wrong because it turned out that there were many programs that DID need more than 512MB of RAM. However, I'm not going to tell you that you're right just because you want me to. I'm willing to admit when I'm proven wrong, are you?

 

[Ilmater]

Originally posted by: chizow

Originally posted by: LikeLinus
Sounds like their doing a great job

Sounds like you're confusing fraud and aggressive accounting methods with a firm's ability to audit.

Chiz

ZING!

 

[human2k]

Originally posted by: Ilmater
I've said it in forums before, and now an AMD marketing guy has confirmed it. Even though it seems that they base their comparisons on current P4 speeds, they still base their PRs on the same thing they did back when the first XPs came out. Read on:

The Inquirer article

You deserve a cookie!

/me gives cookie

 

[Wingznut]

But to answer the initial premise of this thread... Yes, AMD does use language to state that the PR is based on the T-bird.

HOWEVER, if that were actual fact, the methodology they use is completely illogical. And graphs such as these speak very different.

 

[Ilmater]

Originally posted by: paralazarguer
Most naive man in the world. It doesn't matter what the say. Reasoning people can look at a company's statement and try to understand the motivation for a potentially misleading PR statement. AMD obviously introduced their PR system to combat the diverging design philosophies between AMD's high IPC and Intel's high clocks and the effects that it has on consumers.

I'm sure I could come up with something witty to start this post with, but I'll simply say that this statement is 100% correct. I never said that the P4 isn't "the competition." I never said that AMD didn't come up with the rating system to compete with Intel. All I said was that the rating system itself is not based on a T-Bird 1.4. Nothing more.

 

[Wingznut]

If you say it is correct, Ilmater... Then explain the discrepancy of the numbers I illustrated above.

The PR rating is illogical and arbitrary marketing.

Do I blame them for doing it? Not really... What other choice did they have? But that doesn't make it a fact.

 

[Ilmater]

Originally posted by: Wingznut
In the initial article that Ilmater linked, Richard Baker said: "I was wondering why you thought that AMD moves the goal posts on model numbers? The basis on which they are calculated is the same as it has always been, neither the applications nor their relative weightings have been changed since the original 1.4 Ghz Athlon."


Ok, let's look at some numbers. These are the "model number" with the respective actual clockspeed.

• 1500+ = 1333mhz
• 1600+ = 1400mhz
• 1700+ = 1466mhz
• 1800+ = 1533mhz
• 1900+ = 1600mhz
• 2000+ = 1666mhz
• 2100+ = 1733mhz
• 2200+ = 1800mhz

So far, there's a distinct pattern. 66 actual mhz = +100 PR. Even though it isn't logical because they are saying that this cpu scales 50% faster than whatever they are baselining it with. What I mean is: At 1333mhz, they say the "1500+" is ~12.5% faster than whatever they are comparing it to. But at 1800mhz, the "2200+" is now 22% faster, even though there are ABSOLUTELY ZERO architectural improvements. Nonetheless, we'll move on.

Here's where things change, for no apparent reason. Going with the pattern, you'd think that a "2400+" would be 1933mhz, right? Well, AMD disagrees and decided to give their 2000mhz cpu a "2400+" rating. If nothing has changed, then why the discrepancy??? Did the cpu all of a sudden get slower at 2000mhz? Of course not. Would the T-bird have gotten relatively faster at 2400mhz? Of course not.

Let's just say if their engineers used this kind of methodology, they'd be out of work in a big hurry.

Sorry, I had to drive home from work, so I'm just now getting to these later posts. As I stated previously, I found that AMD is basing their PR system on adding another 100+ to their rating system for every ~3% performance increase. Therefore, I would assume that the 2200+ was at the lowest threshold of that ~3% increase. So - admirably, I'd say - when they released the 2400+, they increased the clock speed more than the previous 66MHz. This probably put them back to the higher limit of the ~3% increase over the 2200+ that they wanted to acheive.

Let's say that, for AMD, their instructions were at x instructions per second for every 66MHz on the T-Bred core and y instructions per second for every 100MHz on the TBird core, then y could be only slightly larger than x. And, assuming that was the case, if the first increase (we'll say a 1500+ vs. a 1.4GHz TBird) gave a 3.4% increase in IPS (instructions per second), then that % increase would decrease over the life of the product.

To put in numbers, lets say the 1.4GHz TBird spits out 100IPS, and the 1500+ AXP puts out, on average, 103.4IPS, which translates to a 3.4% increase. Now, if I increase the AXP by the same MHz increment (for comparison's sake, this would be the 1600+), theoretically, I'll get 106.8IPS, which is only a 3.29% increase. The 1700+ in our theoretical world would turn out a 3.18% increase. Now, assuming that was the rule, by the time they got to a 2400+, they would have only gotten a 2.6% increase over the 2300+. So, since this was too far below the ~3% threshold they originally put out, they increased the 2400+ by another 33MHz, giving it another 1.7IPS, or a total increase of 5.1IPS over its 2300+ counterpart. This is a 3.9% increase.

This was just an example, and I know that AMD normalizes all benchmarks to their 1.4 TBird, but it still illustrates the point.

 

[Wingznut]

First of all, when you have a real life example (like I pointed out above), using "assumptions" and "theories" are irrational.

By using the "1500+" PR, AMD stated that the "1500+" is 12.5% faster than a same speed Tbird. But then they say that the "2200+" is 22% faster than a same speed Tbird. Then they reverse their thinking and claim that a "2200+" is only 20% faster than a same speed Tbird.

If their PR were accurate, and based on the Tbird, then you wouldn't have the almost 2x difference. Nor would you have the flip-flop at "2400+".

Originally posted by: Ilmater
This was just an example, and I know that AMD normalizes all benchmarks to their 1.4 TBird...

Well, obviously they don't "normalize" anything. Otherwise the numbers would make sense.

 

[paralazarguer]

You know what? I have a 1ghz T-bird running right now in my basement. It's not actually a 1ghz T-bird though. It's an athlon 1,000,000,000,000. How's that you ask? Well, I'm comparing it to the speed a 8086 would have to run at to get its performance.

The 8086 was an 8 bit processor with few registers and could address almost no memory.

Obviously this rating would be meaningless.

So why is AMD's rating NOT meaningless?

Oh that's right, because each PR rating is meant to correspond to a Pentium 4.

 

[oldfart]

Did anyone look at the AMD Web page Wingz linked to? This is a text cut-n-paste, so the format is messed up, but you get the idea.

--------------------------------------------------------------------------------------------------------------------------
AMD has adopted a new nomenclature using model numbers. Competitive products are shown in the table below.

AMD Core Frequency AMD Processor & Model Number Competitive Processor & Speed
1.60 GHz AMD Athlon? XP 1900+ Intel Pentium® 4 1.9 GHz
1.53 GHz AMD Athlon XP 1800+ Intel Pentium 4 1.8 GHz
1.47 GHz AMD Athlon XP 1700+ Intel Pentium 4 1.7 GHz
1.40 GHz AMD Athlon XP 1600+ Intel Pentium 4 1.6 GHz
1.33 GHz AMD Athlon XP 1500+ Intel Pentium 4 1.5 GHz
AMD recommends that specifications for microprocessors be written to provide for open and competitive offerings from manufacturers. The table below shows several examples of how processor specifications can be written.

Processor Specification
Example 1 AMD Athlon XP processor 1500+ (min) or Intel Pentium 4 processor 1.5Ghz (min)
Example 2 AMD processor model 1500+ (min) or Intel processor 1.5GHz (min)
Example 3 PC processor model 1500+ (min) or 1.5Ghz (min)
Example 4 AMD or Intel processor having equivalent performance

-------------------------------------------------------------------------------------------------------------------------

Anyone find it an incredible coincidence that AMD's PR numbers line up EXACTLY with Intel's P4 CPU MHz speed? But of course, the PR rating has nothing to do with a P4 speed.

 

[RaynorWolfcastle]

IMHO, the whole ratings are fundamentally flawed because, like it or not, any benchmark you make is platform dependent.

Allow me to explain, imagine (hypothetically) that all the original benchmarks are initially done on an AMD SDR SDRAM chipset (AMD 751?). In these benchmarks the T-Bird and the Palomino's performance in benchmark XYZ is rather memory bound, as a result they post rather similar results AMD weighs this benchmark as x% of their PR scheme. As a result let's say 5% of the PR rating is based on the fact that the Palomino at 1.4 GHz is only 2.5% faster than the T-Bird 1.4 GHz on benchmark XYZ.

Fast forward a while to today, AMD releases the T-Bred 2800+. One of two things happen:

--> A) (unlikely) AMD bases their new PR rating by putting the T-Bred at 2.25 GHz in an SDR SDRAM MOBO. The benchmark XYZ is obviously STILL MEMORY BOUND.
result, the T-Bred @ 2.25 GHz only performs 3.4% faster than the T-Bird 1.4GHz. PR rating of the 2.25 GHz processor gets skewed (downwards) because of benchmark XYZ.

--> B) (much more likely) AMD bases their new PR rating by putting the T-Bred at 2.25 GHz and the T-Bird nForce2 boards with dual channel PC2700 RAM. The benchmark XYZ NO LONGER (or much less) MEMORY BOUND.
Result, the T-Bred @ 2.25 GHz only performs a full 100% faster than the T-Bird 1.4GHz and gets a 2800+ rating. PR rating of the of every other freq. Palomino processor gets skewed because their PR rating was computed based on the SDR platform.

This example is rather extreme, but remember that EVERY SINGLE TEST is at least somewhat dependant on platform (video card, RAM, hard disk, and so on). As you can probably tell, after changing platforms a couple of times this becomes a mess. Keep in mind that this doesn't even include for the fact that the mobos used to compute the original PR ratings (1400+ - 1600+ IIRC) are probably not compatible with the T-Bred and Barton core. Throw in an increase in FSB and the possibility that these results could never be verified by AMD or otherwise as no TBird is available at anything near 2.8 GHz and you get even more of a mess.

After reading this, YOU CANNOT POSSIBLY DEFEND THE POSITION THAT THE PR RATING IS IMPARTIAL. Whether or not it is based on the P4 or not is irrelevant. The conclusion is that the PR Rating is marketing, that is at best, loosely related to an objective method of comparing CPUs. At worst it is a system in which marketing types are free to assign more or less arbitrary values to a processor's speed since it is essentially impossible to either prove or counterprover their claims (again no T-Bird at 2.8 GHz).



*Warning* If you're thinking of calling me an Intel fanboy, please check the rigs page in my sig

 

[Bovinicus]

Did no one acknowledge my post showing how the performance of Athlon XPs in reference to their PR rating is much closer to the TBird than the P4. The 1500+ performs like a 1.7GHz-2GHz+ in Anand's initial review of the Athlon XP. Here is the link again. Take a good look.

Now, as I have said, AMD definitely created the PR rating system to compete with Intel. As well, it was meant to show the XPs performance relative to the P4. However, I do think they get their actual number from comparing it to a TBird.

By using the "1500+" PR, AMD stated that the "1500+" is 12.5% faster than a same speed Tbird. But then they say that the "2200+" is 22% faster than a same speed Tbird. Then they reverse their thinking and claim that a "2200+" is only 20% faster than a same speed Tbird.

The reason behind that is very simple. The inital speed of the XP (1500+) was set conservatively. The incriments of increase were set to make up for this gap. At a certain point, they turned the rating into one that was no longer conservative, so they had to normalize the new value. It like having a handicap at the beginning of a race, but being faster. You will eventually pass the other guy, even though you started behind.

 

[paralazarguer]

Did anyone look at the AMD Web page Wingz linked to? This is a text cut-n-paste, so the format is messed up, but you get the idea.

--------------------------------------------------------------------------------------------------------------------------
AMD has adopted a new nomenclature using model numbers. Competitive products are shown in the table below.

AMD Core Frequency AMD Processor & Model Number Competitive Processor & Speed
1.60 GHz AMD Athlon? XP 1900+ Intel Pentium® 4 1.9 GHz
1.53 GHz AMD Athlon XP 1800+ Intel Pentium 4 1.8 GHz
1.47 GHz AMD Athlon XP 1700+ Intel Pentium 4 1.7 GHz
1.40 GHz AMD Athlon XP 1600+ Intel Pentium 4 1.6 GHz
1.33 GHz AMD Athlon XP 1500+ Intel Pentium 4 1.5 GHz
AMD recommends that specifications for microprocessors be written to provide for open and competitive offerings from manufacturers. The table below shows several examples of how processor specifications can be written.

Processor Specification
Example 1 AMD Athlon XP processor 1500+ (min) or Intel Pentium 4 processor 1.5Ghz (min)
Example 2 AMD processor model 1500+ (min) or Intel processor 1.5GHz (min)
Example 3 PC processor model 1500+ (min) or 1.5Ghz (min)
Example 4 AMD or Intel processor having equivalent performance

-------------------------------------------------------------------------------------------------------------------------

Anyone find it an incredible coincidence that AMD's PR numbers line up EXACTLY with Intel's P4 CPU MHz speed? But of course, the PR rating has nothing to do with a P4 speed.

LOL! HA HA HA HA! My god.

 

[LED]

, AMD DOES NOT base its PRs on P4s

The Real deal is that the AMD PR System is based and compared to MHz's no matter who uses them

 

[LikeLinus]

Originally posted by: chizow

Originally posted by: LikeLinus

Originally posted by: chizow

Originally posted by: LikeLinus

Originally posted by: chizow

Originally posted by: paralazarguer
88% of Anandtechers agree: AMD's PR rating is based on the P4

Again, its b/c your poll is flawed. Its a loaded question. Any pollster would rip it to shred.

Chiz

The question might be "flawed" but you seem to not realize the results are true. I'm not quite sure why you keep defending this like you work for AMD, but the PR system is seriously flawed and not many people like it.

It's a direct result of the mhz performance lead of Intel. You still wont answer anything IVE asked, but damn well you sure can whine about that poll a lot.

I'm done pointing out the flaws in your arguments, its no longer worth the time and the effort. I've acknowledged many times that there would be no need for a PR system if AMD could keep up in the mhz race, but that doesn't invalidate the claim that AMD bases their PR rating on their T-bird. I don't purchase a CPU based on a PR rating, I purchase it based on real-world performance (and stepping ).

I don't feel like I'm defending anything, I just don't like seeing misconception becoming reality.

Chiz

No actually thats the first time youve said that. You've avoided my questions every time. Atleast we know the truth and even YOU know it's marketing BS.

<<I don't purchase a CPU based on a PR rating, I purchase it based on real-world performance>>

Yeah but they know joe blow doesnt know the difference, so it's a fabrication on what it is to generate sales based on name/numbers alone.

You also have many good excuses, but the test are fatally flawed. The fact that the chips don't have a standard testbed and the components around them change everytime is undisputible evidence that the PR system is a joke and has no merit.

It's really no different than Intel saying the P3 speeds up the Internet. I'm sure they have benchmarks and proof, but you wouldn't dare defend them would you? Cause you're not a AMD Fanboy are you? Just curious.

It all boils down to lies...and the fact that the ratings system wasn't created by a 3rd party without interest in either company is whats wrong here. It's not scientific and can be changed by what components are ran with you.

You must be good with your hands, as it seems you lack the mental capacity to grasp simple relationships. If you want to isolate 1 variable, you run it on an identical platform with the same testing measure, say a benchmark. If you changed that 1 variable, you would get a different result. If you ran the same test with 1 variable and changed all of the global variables, your end result would be different, but the RATIO or % delta between the results would remain the SAME. Don't talk about lies and flaws when you can't even understand such a simple concept.

On second thought, the PR rating might not be such a bad idea after all, it would help quantify real world performance so people like you won't have to think about it.

Chiz

Sorry I just saw this thread again.

But resulting to name calling and trying to belittle someone to make your point. You're a real winner pal, i bet you feel all man now.

I'm sorry but YOU apparently didn't understand the problem when basicing a products performance on another product, but the testbed around it is totally different, allowing the potential for different components to affect the outcome of the test at hand. You are introducing too many variables into the test that "could" alter the outcome of the products rating. It's not worth getting into again.

Sorry but i'm not going to resort to such childish antics to help you understand.

 

[Bovinicus]

Who really cares what they base their performance rating off of? It works relative to their other CPUs, so that is acceptable to me.

 

[Jeff7]

Originally posted by: sharkeeper
I wish they would do away with PR Ratings. I never liked them when Cyrix had them either. We know what the processor is capable of. Apple doesn't use a PR Rating either.

Cheers!

I think Mastertech01 said something to this effect, but I'll say it again in my own way:
WE know what the processor is capable of. Your average consumer, who doesn't know a processor from a modem, will see Intel 3.06Ghz Pentium IV. Then they'll see AMD 2.167GHz Athlon XP Barton. AMD who? Only 2.167GHz? Well no wonder that system's cheaper, gimme the Intel. And throw in a few megahertz of RAM too.

 

[Ilmater]

I know I'll regret this later, but instead of going for mostly courtesy or witty remarks, I'll just say what I feel. Some of you people are just idiots on this subject! I know that all of us are at least fairly intelligent individuals, but c'mon people. READ WHAT I'M F***ING TYPING!! paralazarguer, WTF man?! I don't know if you just don't read, or just want to push my buttons. If you do, that's fine. I can tell you, it's working. You have said over half-a-dozen times that AMD's PR system is there for the sole reason of competing with Intel's P4, which operates at higher frequencies. You're absolutely f***ing right. Congrats. Now please stop repeating yourself. You don't sound any smarter now than the first time you posted with that comment. Many others in here are stating the same basic thing. oldfart and Wingznut, I do want to say that I mean no disrespect to the two of you. I've seen MANY of your posts and they're very intelligent. But I really think you're missing the point on this one.

I'll say again, that of course AMD is going to put up benches of its processors versus it's biggest (and really, only) competitor. It may even use the same benchmarks to make those comparisons that it uses to come up with its model numbers. Heck, those benchmarks might even be benchmarks that highly favor AMD's architecture over Intel's architecture. I'm not arguing any of those facts. All I'm saying is that when it determines its model numbers, it's using the 1.4 T-Bird as the basis for comparison. That's it. As the marketing guy said:

The basis on which they are calculated is the same as it has always been, neither the applications nor their relative weightings have been changed since the original 1.4 Ghz Athlon.

Now, I work at a large company, and nobody hates telemarketers (edit: whoops, I meant marketers) more than I do. They'll make claims that sound ridiculous, but they DO NOT flat out lie. They can't. That could cost the company a lot of money. All I've been trying to say is that AMD bases its XP model numbers on a baseline determined by a 1.4GHz Athlon. This guy is saying that I'm exactly right. Do you honestly think he's lying? Because of the wording of his statement, there is absolutely no room for reading between the lines. You're either saying he's a liar or he's not. I'm saying he's not, and I'm giving you reasons to back up my arguments.

Wingznut, I'm just going to refer you to my previous post. I know you say you don't want my "theoretical" examples, but your examples aren't logical. Look at the simple math. If you go to the basest form, IPS should determine speed. That should tell you how much the processor can calculate in a given amount of time. Increasing the frequency with which data is thrown through the processor, should theoretically increase on a constant, incremental basis, not a percentage basis. So, since AMD says that every 100+ increase in model number translates to roughly a 3% increase in performance, they had to increase the frequency by more than 66MHz to get that 3% increase. I understand that the 2400+ is more of an increase over the 2200+ than the 2200+ was over the 1800+, but the system, as a whole, wasn't deviating from their initial plan, which was to increase the model number by 100 for every ~3% increase in performance.

As for the thread in general, its purpose was to simply establish the fact that AMD bases the actual numbers it puts on its XP line by normalizing benchmarks to the performance of a 1.4GHz T-Bird. Nothing more. I don't want to claim that they're fair, or that they aren't used to compete with Intel. However, if anyone argues the validity of the PR system, then you have got to know what that system is based on. I've read many times where people (I believe SexyK was the one I last argued this with) talk about the PR ratings, and then somebody chimes in with, "Yeah, but they're not based on the T-Bird like they used to be. In fact, even AMD doesn't claim they are any more." That's clearly not the case. AMD does use the T-Bird 1.4 and it DOES claim that that's the case. So if this thread accomplishes what I want it to, I can later say to someone, "This has already been argued about, and we've already established that the T-Bird is the basis."