AMD ? Performance to Rating ratio

[Noj]

OK, so we all know about AMD's Performance Rating on the Athlon XP, and we all know that:
1.33 Ghz = 1500+
1.40 Ghz = 1600+
1.47 Ghz = 1700+
1.53 Ghz = 1800+

However, while the 1.33 Ghz chip is rated at 12.5% higher, the other chips are not. In fact:
1.33 Ghz = 12.5% higher than it?s rating
1.40 Ghz = 14.3% higher than it?s rating
1.47 Ghz = 15.9% higher than it?s rating
1.53 Ghz = 17.4% higher than it?s rating

In order to give an accurate measure of performance with the rating system, shouldn?t these processors all be rated at the same percentage of their clock speed?

 

[Duvie]

INteresting!!! Good question....

The only thing that needs to be said is that the 1.53ghz or athlon 1800+...whatever the hell you want to call it beats the p4 2ghz....

 

[Noj]

<< the 1.53ghz or athlon 1800+...whatever the hell you want to call it beats the p4 2ghz.... >>



I agree, but that's not the point. If AMD wants to tell us that the Althon XP 1.33 Ghz CPU is 12.5% faster than a P4 1.33 Ghz CPU (if there was such a thing), then fine I have no problems with that.
However, if they then tell us that their Althon XP 1.4 Ghz CPU is 14.3% faster than a P4 1.40 Ghz, they are clearly lying. The 1.4 Ghz chip has no architectural improvements to warrant that extra performance boost.

Perhaps it doesn't matter, since we can just look at the benchmark scores to see what the real performance is, but I don?t see how they expect us to accept their performance rating when it?s clearly illogical.

 

[Jiggz]

The ratings were supposedly made with comparison to P4's speed but then just like everybody says, the XP 1.53 Ghz can whip a 2Ghz P4 anytime, anyplace. Yet it's only rated at 1800+. How AMD came out with their rating numbers is anybody's guess. To make the rating a constant factor of the speed maybe logical but neither is it accurate. Which is basically true with P4 speed and performance. A 2.0 Ghz P4 is 43% faster than a comparable 1.4Ghz P4 speed wise but I don't think its the same ratio in performance.

 

[Noj]

<< A 2.0 Ghz P4 is 43% faster than a comparable 1.4Ghz P4 speed wise but I don't think its the same ratio in performance. >>



True, but that has more to do with the systems other bottlenecks than the CPU itself, I would expect a 2.0 Ghz P4 to perform 43% better the a 1.4 Ghz P4 if the CPU was the only bottleneck. AMD aren't rating your computer, they're rating their CPU.

 

[twong82]

If I haven't read wrong from articles regarding the performance rating for Athlon XP processors, PR is based on the equivalent to the Thunderbird processors not the P4's although a Thunderbird will also beat a P4. i.e. a 1.53Ghz Athlon XP is equivalent to a 1800+ Athlon Thunderbird.

 

[Duvie]

misread earlier post...

NOj appears to understand xp rating is to tbirds, but is just questioning the growing percentage of performance against similar tbirds...

A 1.33 ghz tbird vs a 1.4ghz tbird = 5.26 percent increase in performance

a 1.33ghz athlon 1500+ vs a 1.4ghz athlon 1600+ should also equal 5.26 percent increase

like architecture should equate to similar architecture, when void of other system bottlenecks.

HOwever noj points out the 1.33xp vs 1.33tb is a 12.5 percent increase and the 1.4zp vs 1.4tb is 14.3percent which would equate to a 6.9 percent increase in 1.33xp to 1.4xp which is greater then the ratio of clock speed with clock instructions aside since they should be similar...

 

[Noj]

Duvie, either you completely misread my post, or I have completely misread yours. Please clarify.



<< The p4 has a longer pipeline and though the clock speed is 2ghz the amount of instruction per clock cycle is less >>



In this case I was comparing a Pentium 4 to another Pentium 4, not the Athlon.



<< It is being rated to a tbird not a p4 >>



Fair enough, but if the 1.33 Ghz XP is 12.5% faster than the 1.33 Ghz T-Bird, then the 1.4 Ghz XP must be 12.5% faster than the 1.4 Ghz T-Bird. However, their performance rating indicates that it is 14.3% faster.

 

[Duvie]

< Fair enough, but if the 1.33 Ghz XP is 12.5% faster than the 1.33 Ghz T-Bird, then the 1.4 Ghz XP must be 12.5% faster than the 1.4 Ghz T-Bird. However, their performance rating indicates that it is 14.3% faster. >

I totally agree...that is why I said good question earlier...

< would expect a 2.0 Ghz P4 to perform 43% better the a 1.4 Ghz P4 >

oops!! I guess I read over that as saying 1.4ghz amd... Sorry I will correct earlier post...

 

[Noriaki]

<<

<< the 1.53ghz or athlon 1800+...whatever the hell you want to call it beats the p4 2ghz.... >>



I agree, but that's not the point. If AMD wants to tell us that the Althon XP 1.33 Ghz CPU is 12.5% faster than a P4 1.33 Ghz CPU (if there was such a thing), then fine I have no problems with that.
However, if they then tell us that their Althon XP 1.4 Ghz CPU is 14.3% faster than a P4 1.40 Ghz, they are clearly lying. The 1.4 Ghz chip has no architectural improvements to warrant that extra performance boost.

Perhaps it doesn't matter, since we can just look at the benchmark scores to see what the real performance is, but I don?t see how they expect us to accept their performance rating when it?s clearly illogical. >>

You are reading way to much into this.

If your average person was intelligent enough to sit down and work all that out, AMD wouldn't need the Model number ratings.

The model number ratings are for joe blow six pack that doesn't know a CPU from a hole in his ass....so that the guy doesn't just look and see oh wow...AMD 1.53Ghz, Intel 2Ghz...Intel must be faster..bigger number and all...

I would guess the Reasons AMD use 1500, 1600, 1700 and 1800 is because the Thunderbird went up to 1400Mhz.
Therefor they have consistancy in their product lines. Their newer CPUs have bigger numbers, therefore are faster.

If you are smart enough to know that the ratios are wrong, then you are probably smart enough to realzie there is more to the performance story than the numbers.

The Model ratings aren't for people like us, they are for the majority of the users out there.


It's all about marketting, it's not suppose to be an accurate measure of performance, nor is it meant to be logical. Marketting rarely is.

 

[Noj]

<< ...that is why I said good question earlier... ... Sorry I will correct earlier post... >>



Thanks.



<< It is being rated to a tbird not a p4. >>



Do you have links to any information where AMD states how they determined their performance rating (i.e. what benchmarks they used, what they compaired the Athlon XP to, etc...)?

 

[Zebo]

Simple. round mumbers.

If they went srictly by % the model numbers would be something like

1500+
1593+
1647+
1753+

 

[Noj]

Noriaki, I agree with you, I?m just disappointed that AMD?s ?True Performance Initiative? is just another way to dumb down the masses.
Perhaps you?re right and I should just shut up and be happy that I know how these CPU perform in a real world environment, I just feel that people deserve a little better.

 

[Remnant2]

Actually that's not necessarily the case.. the XP has several things, like hardware prefetch, that have increasing impact at higher clock speeds. It's entirely possible that the gap between the Tbird and Palomino cores widens as the frequency scales.

Put another way, the constant terms in the performance equation are the memory subsystem and the other system components. A Tbird might begin to scale slower and slower as the frequncy increases since it is waiting longer and longer for data from RAM. With HW prefetch, the wait actually decreases in terms of ns time because the frequency of the CPU (and thus the cache/prefetcher) is increasing. This is assuming that the memory bus isn't saturated, which with DDR is probably true.

 

[Noriaki]

<< Noriaki, I agree with you, I?m just disappointed that AMD?s ?True Performance Initiative? is just another way to dumb down the masses.
Perhaps you?re right and I should just shut up and be happy that I know how these CPU perform in a real world environment, I just feel that people deserve a little better. >>

Yes, you are quite right. I'm dissappointed in the whole model number thing altogether.

But, the fact is, joe consumer won't know that performance doesn't depend on clock speed....
How do you suppose Dell sells so many SUPER FAST systems with P4 1.7Ghz with 128MB of PC133 on i845 with a GF2MX?

That system would get spanked by a Duron 1Ghz with a GeForce2GTS....but jo consumer won't know that...

AMD has to play the stupid marketting games...if they don't, they will sell no chips...and they will die

I don't like it either, but I accept it as necessary for AMD to stay afloat...I wish that the majority of consumers were smart enough to see beyond clock speed...but they aren't :disgust:

 

[Sugadaddy]

<< Noriaki, I agree with you, I?m just disappointed that AMD?s ?True Performance Initiative? is just another way to dumb down the masses.
Perhaps you?re right and I should just shut up and be happy that I know how these CPU perform in a real world environment, I just feel that people deserve a little better. >>




the masses are already dumb beyond belief, so it actually helps them...

 

[Zebo]

<<
the masses are already dumb beyond belief, so it actually helps them... >>



Ignorance is bliss?

 

[cvlegion]

The problem is that it isn't rated high enough. My Athlon T-bird 1.4GHz would lick a Pentium 4 2GHz any day of the week. Considering that the Athlon XP is faster than the T-bird for the corresponding speed, the 1.33 XP should be at least an Athlon XP 2000+. Now when I get my 1800+....

 

[Mookow]

<< The problem is that it isn't rated high enough. My Athlon T-bird 1.4GHz would lick a Pentium 4 2GHz any day of the week. Considering that the Athlon XP is faster than the T-bird for the corresponding speed, the 1.33 XP should be at least an Athlon XP 2000+. Now when I get my 1800+.... >>


The T-Bird 1.4 and the P4 2GHz are as close to equals as two different microprocessors can get today. Check Anand's review of the P4 2GHz. I'm not disputing that the Athlon XP 1800+ is, on average, faster than the P4, but.....