a little confused about AMD Ghz rating

[Maximilian]

Originally posted by: Brunnis

Originally posted by: Viditor
I think you misunderstand what they did...
They had a 3rd party (one of the big 5 accounting firms I believe) compile a series of benchmarks, and then test the 1GHz Tbird. This was the benchmark for a 1000+ numbering...then they ran the same benchmarks on each chip as it was produced to get the model number (based on how it compared to the Tbird on those benchmarks).
This same numbering system has been carried on for quite some time and continued through the XPs, though I really think they have abondoned it for the X2s.

That's one idiotic system. Why? Because no one is interested in how an Athlon64 compares to a CPU that's pretty much forgotten by now. An interesting aspect of the numbering system is that it matched the P4 much better than it did with a Thunderbird. How strange...

Another interesting thing is why AMD would use a system that maybe 1% of the computer buyers understand. Do you think anyone outside the enthusiast scene would get anything out of a comparison with an old discontinued CPU. Of course not. They don't even know what a Thunderbird is and the relative performance to one is quite uninteresting.

Whatever the official line was/is, it's not really up for questioning that the system exists solely because of the P4 and that it's meant to compare to this CPU.

Last week I actually met the guy that designed the system (Patrick Moorhead at AMD). It's a shame that this discussion didn't pop up earlier so that I could have asked him face to face. It's likely that he'd stick to the official line, though.

Well thats the system, like it or lump it. It was abandoned after the end of the XP line, the numbers are now just to distinguish between models.

Yes originally a lot of people figured it was against the P4 they were comparing the AXP's with but an AXP 3200+ isnt the equal of a p4 3.2 ghz, the p4 is faster. Its about the same as a P4 2.8/3.0 with HT.

Bottom line, they are not performance indicators against anything anymore, they are just to differentiate between products within their own line.

 

[Brunnis]

Originally posted by: Soviet
Well thats the system, like it or lump it. It was abandoned after the end of the XP line, the numbers are now just to distinguish between models.

So, the model numbers for the Athlon64 just happened to match up very well with the Pentium 4. That's a little naive, don't you think?

Originally posted by: Soviet
Bottom line, they are not performance indicators against anything anymore, they are just to differentiate between products within their own line.

I believe that for the X2 line, but not the regular Athlon64 line.

 

[hans007]

i think they will get rid of it again.

they originally had the system back during the k5/k6 days, then got rid of it during the original athlon, and brought it back again during the athlon xp.

 

[Brunnis]

Originally posted by: hans007
i think they will get rid of it again.

they originally had the system back during the k5/k6 days, then got rid of it during the original athlon, and brought it back again during the athlon xp.

I agree. I think they should adopt a model number scheme similar to that of the Opteron instead.

 

[coolingoff]

The P4 does run at higher clock speeds than the athlon, implying the processor completes 3.4 or whatever billion cycles a second.
The AMD processor in theory run slower, ie fewer clock cycles per second but for each cycle they are able to do more than the P4, hence the equivalent system or whatever.

The efficiency in each cycle is obviously better, since even intel reverted to slower clock speeds but efficient ones for the core duo.

 

[Gdepp519]

On a Bit of a Side Note is there any websites that compares AMD to Pentiums?? or gives a calculator showing the difference??

So what i guess I am asking is there some website that offeres a simple drop down menu where you can choose an AMD and Intel Processor
and then be able to have the two compared against each other??

 

[Maximilian]

Originally posted by: Brunnis
So, the model numbers for the Athlon64 just happened to match up very well with the Pentium 4. That's a little naive, don't you think?

They didnt match up very well at all. If anything AMD held back with their model numbers. A 3200+ A64 is better than a 3.2 ghz P4, in some cases its better than a 3.6ghz p4, and all of them are better than a 3200+ AXP, matches up wonderfully that :roll:

This is why we on anandtech use benchmarks to indicate performance

Originally posted by: Gdepp519
On a Bit of a Side Note is there any websites that compares AMD to Pentiums?? or gives a calculator showing the difference??

So what i guess I am asking is there some website that offeres a simple drop down menu where you can choose an AMD and Intel Processor
and then be able to have the two compared against each other??

You can do that here You can find even older cpu's on the same site under a different chart in case you wish to compare somthing like a 1ghz thunderbird to one of the AXP's and see how it matches up.

 

[dennilfloss]

Originally posted by: Brunnis

That's one idiotic system.

Not for those who upgraded from a Tbird to subsequent AMD offerings. It allowed us to assess how much performance increase we could expect from each subsequent AMD offering compared to our current CPU.

 

[justly]

I can't give a difinitive answer to the rating system AMD uses, but the way I understand it there are some details missing in what people have described.

First, the rating system was introduced when SSE was incorporated into the XP line. Since the "added performance" from SSE was not accompanied by a clock speed increase the best way to market this feature (and its additional performance) was to give it a performance rating. I also agree that the XP line used a 1000Mhz Athlon as a base of comparison.

When AMD intoduced the Athlon 64 there was a lot of confusion about the rating system because a A64 3200+ performed better than the XP 3200+. Some people overlook the fact that SSE2 was incorporated into the AMD line with the introduction of the A64. Because of this "added feature" a new test suite (benchmark) had to be developed to show the increased performance (just like what had to be done when SSE was added to the Athlon). The results from this testing suite also had to have a base model processor to compare it to, most people still think the 1000Mhz Athlon is the base for comparison. I don't believe this to be true for a couple of reasons, first why compare to a processor that is no longer in production and has niether SSE or SSE2, second I happen to remember reading in an Anandtech article stating that the A64 uses the XP 1800+ as a base for comparison ( I can't remember the specific article so please don't ask for a link).

I expect some people might wonder why the top of the line XP wasn't used, I suspect the same reasons they didn't use the top model Athlon, they probably used one of the more popular models and it also was in the range where it scaled well (not limited by factors outside the scope of the processor, like memory bandwidth).

I do think the rating system was somewhat a marketing tactic, but it also was (and still is) a valid way to judge performance for AMD processors (with the possible exception of the X2 line).

The rating system definaty has its use as a marketing tool against the P4, but using it to compare performance to a P4... sorry I can't agree with that. The performance of the P4 had as much to do with the type of memory, FSB speed and the amount of cache as the actual Mhz, in many respects you couldn't judge the performance of a P4 at all based only on Mhz. Lets also not forget that the rating system was established when the P4 was still using a 400Mhz FSB so if the rating system was actually designed to compare performance against the P4 the rating numbers where to low at the time they where introduced.

 

[Brunnis]

Originally posted by: Soviet
They didnt match up very well at all. If anything AMD held back with their model numbers. A 3200+ A64 is better than a 3.2 ghz P4, in some cases its better than a 3.6ghz p4, and all of them are better than a 3200+ AXP, matches up wonderfully that :roll:

This is why we on anandtech use benchmarks to indicate performance

Ahh, yes, the famous one sided gaming comparison. The fact is that a P4 3.2GHz is very competitive with an Athlon64 3200+ if a wide range of benchmarks are used.

Originally posted by: denniflossNot for those who upgraded from a Tbird to subsequent AMD offerings. It allowed us to assess how much performance increase we could expect from each subsequent AMD offering compared to our current CPU.

I did just that (upgraded from a T-bird to an AXP). Just didn't find the model number to correspond well with T-bird frequency. It did however match up quite well with the P4, but as it turns out, it's all a big coincidence.

Moving on now...

 

[Gikaseixas]

Btw, most games now come with a Pentium and a AMD recommended model numbers. It's a no brainer that a A64 cpu will do more calculations per cycle than a P4 cpu and that's why you see the difference in Mhz. The Conroe is more efficient than a X2 clock per clock, that's why a E6600 will be faster than a FX62 in same cases...

 

[Arkaign]

Yeah, the number systems on both sides have gotten horrible. Mhz/Ghz are sort of useless now as well, unless you're comparing speeds of the same cores, and even that can be skewed by varied performance scaling and efficiency.

It takes patience to explain :

Socket A Sempron 2800+ > Socket A Athlon XP 2800+ > Socket 754 Sempron 2800+, Socket 754 Athlon 64 2800+ (for the most part)

 

[Maximilian]

Originally posted by: Arkaign
Yeah, the number systems on both sides have gotten horrible. Mhz/Ghz are sort of useless now as well, unless you're comparing speeds of the same cores, and even that can be skewed by varied performance scaling and efficiency.

It takes patience to explain :

Socket A Sempron 2800+ > Socket A Athlon XP 2800+ > Socket 754 Sempron 2800+, Socket 754 Athlon 64 2800+ (for the most part)

No, the socket 754 varients are better than the socket A cpu's.

Socket 754 Athlon 64 2800+ > Socket 754 Sempron 2800+ > Socket A Athlon XP 2800+ > Socket A Sempron 2800+

EDIT: Oh wait maybe you were putting them in lowst to highest order, rather than using the > (greater than) sign to its actual meaning.

 

[BigDH01]

I never understood this obsession with Athlon 64s. I too had this obsession once, thinking that the A64 was clearly superior than the P4. Then, I went back and took a look at the benchmarks. In reality, at stock speeds, the 3800+ A64 and P4 at 3.8 GHz were roughly comparable in performance except in games.
If games were excluded, I'd say the 3.8 P4 barely gets the nod.

Anandtech Review

Of course, the FX was the fastest chip but this wasn't really a mainstream chip. The EE was a huge ripoff at this point.

With the introduction of dual core, AMD clearly took the lead until the Conroe. I guess it's kind of off-topic but I was just amused looking back at the benchmarks and realizing just how close in performance the mainstream A64s and P4s were. I suppose that because the A64 dominated gaming, it became the chip that everyone remembers.

 

[Brunnis]

Originally posted by: BigDH01
I never understood this obsession with Athlon 64s. I too had this obsession once, thinking that the A64 was clearly superior than the P4. Then, I went back and took a look at the benchmarks. In reality, at stock speeds, the 3800+ A64 and P4 at 3.8 GHz were roughly comparable in performance except in games.
If games were excluded, I'd say the 3.8 P4 barely gets the nod.

Anandtech Review

Of course, the FX was the fastest chip but this wasn't really a mainstream chip. The EE was a huge ripoff at this point.

With the introduction of dual core, AMD clearly took the lead until the Conroe. I guess it's kind of off-topic but I was just amused looking back at the benchmarks and realizing just how close in performance the mainstream A64s and P4s were. I suppose that because the A64 dominated gaming, it became the chip that everyone remembers.

Exactly! The Athlon 64 was never that much better than the P4, especially the Northwood. Sure, gaming was faster, but I'm pretty sure most people were graphics card limited anyway. It wouldn't surprise me if a HyperThreading enabled 90nm Northwood with some extra cache and x86-64 would have have been the better chip overall, compared to the A64.

The Prescott was a horrible product, though, mainly because of excessive power consumption and no performance gain.

 

[mhahnheuser]

Except that the reference target/product was both the 1st. GHz CPU and 2ndly. the fastest, so measuring in multiples of it would keep the numbers out in front of the opposition, especially as some other posters here have elluded to already that the early releases of netburst CPU's ran slower than the latest of the P3's. Good ploy by AMD in making Intel's MHz strategy work against them. C2D is also fighting some legacy battles in this regard in that for sometime now AMD CPU's have been as quick and in some cases considerably quicker than Intel's, now to try and steal market share back with 1.8 and 2.13 GHz CPU's will take some effort. Average Jo won't pay a premium for flash harry CPU's, and cheaper AMD systems are hitting the shelf with heavier GPU power, so Intel will have its work cut out ahead of it convicing the masses its back in front, and that's assuming that there is no quick response from AMD.