AMD's PR system is now totally useless!

[RedShirt]

With the release of the new, 133 FSB Pentium IV chips, the PR system used by AMD does not work! Just look at Tom's benchmarks at Tom's Hardware (see the Athlon XP 2400+ compared to the PIV 2400 with 133mhz FSB).

Sure, AMD may be comparing the XP speeds with the older T'birds, but these speeds just add to the confusion now... They used to be conservative when comparing to a PIV, but now they are not. AMD should just go back to stating their REAL speed.

And no, I don't have anything against AMD, in fact, I use AMD chips in both of my rigs and like them, it just seems a little dishonest of them now to maintain this rating system.

 

[ChrisIsBored]

And ya know what, Dell sells this laptop... they call it the 8200.. but it's only a 1.4Ghz

 

[viper007]

Where the heck it the Athlon XP 2400+? It hasn't been offically anouced so quiet

 

[Rand]

I tend to disagree, for one the highest rated AthlonXP is 2100+.... there are no 533MHz FSB P4's at such clockspeeds. The 2100+ would be comparable most closely to the P4 2GHz/2.2GHz on the 400MHz FSB. In such a comparison the 2100+ is still rated slightly on the conservative side as it tends to outperform the 2.2GHz P4 in most cases.

Putting a hypothetical AthlonXP 2400+ up against the P4 2.4/533 is a more appropriate comparison. Given that a 2100+ typically performs roughly comparable to a theoretical P4 2.3GHz/400, I find it reasonably likely that a 2400+ would perform extremely closely to P4 2.4/533... if anything I might expect the 2400+ to have an extremely slight lead.

Naturally their are going to be exceptions... SSE2 optimized apps will favor the P4, while strongly X87 FPU limited apps will favor the AthlonXP. But on the whole I feel the two should perform very similarly.

Also, as it stands the Barton core is evidently going to see a 512K L2 cache, if this hypothetical 2400+ was of the Barton core it is quite likely it would be superior to the P4 2.4/533 in most cases.

In short, I don't consider the Model rating scheme useless at all.
At the very least it is certainly still a viable means of comparison for the 2100+ and all current AthlonXP processors as the closest P4's are all Willamette and Northwood/400.

 

[RedShirt]

<< And ya know what, Dell sells this laptop... they call it the 8200.. but it's only a 1.4Ghz >>



Yeah, but that is a totally different matter. If Dell sells computers, not just the processors. Dell can call the complete system what they want because it is a whole computer. Usually, the higher the number, the better the system. Dell simply could not call all their computers with a 2 GHz processor a DELL 2000, then it would be very confusing.

I have had no problems with AMD doing a PR rating in the past because the ratings were conservative, but now they are not.

 

[RedShirt]

<< I tend to disagree, for one the highest rated AthlonXP is 2100+.... there are no 533MHz FSB P4's at such clockspeeds. The 2100+ would be comparable most closely to the P4 2GHz/2.2GHz on the 400MHz FSB. In such a comparison the 2100+ is still rated slightly on the conservative side as it tends to outperform the 2.2GHz P4 in most cases.

Putting a hypothetical AthlonXP 2400+ up against the P4 2.4/533 is a more appropriate comparison. Given that a 2100+ typically performs roughly comparable to a theoretical P4 2.3GHz/400, I find it reasonably likely that a 2400+ would perform extremely closely to P4 2.4/533... if anything I might expect the 2400+ to have an extremely slight lead.

Naturally their are going to be exceptions... SSE2 optimized apps will favor the P4, while strongly X87 FPU limited apps will favor the AthlonXP. But on the whole I feel the two should perform very similarly.

Also, as it stands the Barton core is evidently going to see a 512K L2 cache, if this hypothetical 2400+ was of the Barton core it is quite likely it would be superior to the P4 2.4/533 in most cases.

In short, I don't consider the Model rating scheme useless at all.
At the very least it is certainly still a viable means of comparison for the 2100+ and all current AthlonXP processors as the closest P4's are all Willamette and Northwood/400. >>



Well, let's put it this way, if AMD continues the scaling like they do now, the scheme will not be correct. 512k of Cache may help the Athlon, but these PIV's are just getting better and better the faster they get (especially when the memory bus gets faster).

What would be REALLY confusing is if AMD decides to change their PR system, altering it later on, to adjust it to these faster PIV chips. This is why I say they should ditch the whole thing alltogether

 

[Adul]

the coul just step up the speed one notch to make up to the slant in the ratings.

 

[RedShirt]

<< the coul just step up the speed one notch to make up to the slant in the ratings. >>



This is true, but they would have to do this every time Intel does something to make the PIV faster. This is probably what they will end up doing in the end.

Looking at the benchmarks now (at Tom's site) his simulated 2400+ doesn't do as well as the PIV 2400 in the majority of the benchmarks.

 

[FishTankX]

Redshirt, has anyone yet mentioned that nearly everyone uses DDR instead of PC1066 and there's a very very very extremley large performance disparity between the performance of a PC266 Pentium4 machine and a PC1066 pentium4 machine? Next time try comparing the AthlonXP 2100+ with PC266 memory running at CAS 2.5 to the Pentium4 running the same (This is standard config for alot of OEM's) and better yet test the 2.2 or the 2.0 against the AthlonXP 2100+ and I guarantee you you'll be eating your words.

 

[FishTankX]

Example
http://www6.tomshardware.com/mainboard/02q2/020501/ddr400vsrambus-06.html
QuakeIII arena

PC266 CAS2.0 SiS645DX 294FPS

i850E PC1066 381FPS

Now, let's assume that in the rest of the benchmarks the performance disparity is only 7% or so. Okay, that's fine.

Now, adjust all of the Pentium4 scores 7% and *then* compare them to the AthlonXP. Fair? I think so. Same memory technology...

Well, Redshirt, what is your response?

 

[ChrisIsBored]

<<

<< And ya know what, Dell sells this laptop... they call it the 8200.. but it's only a 1.4Ghz >>



Yeah, but that is a totally different matter. If Dell sells computers, not just the processors. Dell can call the complete system what they want because it is a whole computer. Usually, the higher the number, the better the system. Dell simply could not call all their computers with a 2 GHz processor a DELL 2000, then it would be very confusing.

I have had no problems with AMD doing a PR rating in the past because the ratings were conservative, but now they are not. >>




You're not a quick one to pick up on sarcasm are you RedShirt?

 

[sMashPiranha]

Dude, AMD's PR system wasn't supposed to compare the Athlon XP to the P4. It was supposed to compare the Athlon XP to the Thunderbird. (I'm pretty sure thats what it was anyway)

 

[FishTankX]

Smash Pirana:That doesn't help the consumer pick a CPU, does it. And that's what PR ratings are suposed to do. That's what he was talking about.

 

[Gstanfor]

<< AMD's PR system is now totally useless! >>


Whatever leads you to that conclusion?

AMD's PR system was never intended to compare AthlonXP to intel CPU's in the first place, it was a comparison between an Athlon XP and a (theoretical) equally clocked Athlon Thunderbird.

This comparison is still perfectly valid since it in no way involves intel specifications. Those who believe the PR rating is in comparison to P4 are wrong. Check amd's website out and read up on the rating system.

Greg

 

[imgod2u]

In no way is the AthlonXP "1600+" comparable to a T-bird at 1.6 GHz. AMD has to claim this officially of course, however, I think it is pretty much implied that it is used to be comparable to the P4's. I never liked the PR system to begin with. As I see it, it is just a way one company tries to make itself the standard for performance (AMD says it is equal to another processor, therefore it must be?). Companies should make the chip and let third parties do the speed comparisons. I know there's a lot of other marketing stuff involved but AMD is always viewed as a more honest company (albeit not by me). It shocks me how many people still think this. Credit where credit is not deserved is much worse than the PR system.

 

[Athlon4all]

I agree with Rand. I was pouring over the benchys from yesterday and it was clear that the XP 2100+ was neck and neck, sometimes pulling ahead of the 2.2 P4. So it is not useless, does it need to change, yes

 

[TSDible]

Hmmm...

Each time you "change" the rating system, it will become more confusing.

The fact of the matter is... as long as AMD is behind in the MHz race, they will use a Rating system. I would put down money right now that if AMD were ahead of Intel in the MHz game there would be no PR-Rating! Therefore, it appears to me that the PR-Rating system is directed to comparing AMD CPUs to Intel CPUs... even if it is not on paper.

Is the PR-Rating good or bad? I honestly don't know, but it is a direct attempt to remain competitive with Intel. There is no other way to see it. You can't remain competitive with a company and only compare your products to your own products. Think about that.

Just my $0.02

 

[HardWareXpert]

<<

<< AMD's PR system is now totally useless! >>


Whatever leads you to that conclusion?

AMD's PR system was never intended to compare AthlonXP to intel CPU's in the first place, it was a comparison between an Athlon XP and a (theoretical) equally clocked Athlon Thunderbird.

This comparison is still perfectly valid since it in no way involves intel specifications. Those who believe the PR rating is in comparison to P4 are wrong. Check amd's website out and read up on the rating system.

Greg >>



AMD's own website say PR is comparable Athlon speeds, however AMD's own white paper on PR shows AthlonXP 1500+/1600+/1700+/1800+ againest P4 1.5/1.6/1.7/1.8.

This white paper AMD admits that clock frequencys are not the hold picture, because Intel are so far ahead in the Mhz, AMD us the PR system to offset this.

AMD's PR White paper

 

[Rectalfier]

Face it, AMD will not drop the PR rating. Intel purposely engineered a processor that would scale highly in megahertz, so that they could push AMD into the low end, since prices are based on MHZ. Although AMD's Tbird 1.4 performed on par with the Willy 2.0, AMD had to price their part depending on Intel's 1.4. AMD lost tons of money, and they would go bankrupt if they did not impose such a rating. So far, the ratings have been very fair, even conservative, so I do not have a problem with it.

 

[randomlinh]

and because the general public is always "bigger/more is better," if AMD didn't use the PR rating, where would they be? They'd look like they were WAY behind intel. I admit, i really don't like the PR rating, but what else could they do really?

 

[HardWareXpert]

The PR system has done well after all, the AthlonXP1600+ -2100+ do piss all over there Intel counter parts, so people are really getting a 2Ghz CPU(2000+) for a third of the price.

 

[imgod2u]

<< Face it, AMD will not drop the PR rating. Intel purposely engineered a processor that would scale highly in megahertz, so that they could push AMD into the low end, since prices are based on MHZ. Although AMD's Tbird 1.4 performed on par with the Willy 2.0, AMD had to price their part depending on Intel's 1.4. AMD lost tons of money, and they would go bankrupt if they did not impose such a rating. So far, the ratings have been very fair, even conservative, so I do not have a problem with it. >>



Ok, that's BS and you know it. The P7 design started in 1996. You honestly think Intel thought "gee, well AMD will release the K7 and beat us in the MHz in 1999, so we've gotta start on a MHz monster now?". It constantly amazes me how people think Intel can see the future. The P7 core was created because the engineers knew that eventually, the limitations on how fast signals can trasmit between logic gates will be limited no matter what you shrink the transisters to (for an example, take a look at the T-bred and how it overclocked in that overclockers.com.au article, 2.0 GHz stable with a Vapchill cooling, that's a .13 micron chip with copper interconnects and it's maximum yield was 11.11% higher than its .18 micron transister size predecessor). In contrast look at the Northwood and its scalability compared to the Willamette even with double the cache. 4.1 GHz (demonstrated at IDF) compared to 2.2ish GHz (I think that was the relatively high Willamette overclock with extreme cooling). There is a limit to how fast logic gates can switch on and off when in complicated stages.
And with the proper prediction algorithms, enough memory bandwidth and coding that provides enough instructions to process every second (or multiple threads of instructions), it could have just as high IPC as its predecessor (the P6 core). It's all about maximizing the gains while minimizing the losses. Marketing had nothing to do with the design (well, prolly something to do with it but it wasn't just for the MHz increase). The P7 core's early implementation and release in the form of the P4 on the other hand, I suspect was due to severe competition from AMD (I doubt the P7 core was even meant for sub-2GHz chips).

 

[HardWareXpert]

Dude, Intel made the P4 with a 20 stage pipeline, no other Pentium has ever been able to scale that far, you no why?
Because Intel realized that AMD could out perform them easy clock for clock so they made the P4 scalable to aleast 3Ghz.

This is partly the reason why the P4 is so slow, the 20 stage pipeline crippled it, by introducting enhancements to the core which softed the blow of such a big pipeline. Intel's FPU is also inferiour to Athlon's FPU and SSE2 still cannot make the difference in performance.

Intel are also fooling people with the 400MhzFSB which is 4x100Mhz quadpumped, adding more pipes to the bus dont make things better, making the water flow faster down the pipes does. Very few apps and games take advantage of heavy bandwith thats why in anything using FPU the Athlon wiped the floor with P4.

 

[LikeLinus]

<<Because Intel realized that AMD could out perform them easy clock for clock so they made the P4 scalable to aleast 3Ghz.>>

Actually thats not a very accurate statement. The P3 and Original Athlon ran about neck and neck in most benchmarks, they were not "easily outperforming them". The Tbird started to pull away, but by that time they created the P4 for higher mhz cause AMD is the one who started this whole Mhz war. AMD started something they obviously couldn't finish and are now getting toasted mhz to mhz. AMD chose to create this mythical "PR" system to make their chips give the appearance they were faster than what their Mhz rating is. See people really only care about Mhz. AMD was pushing for that because at they were spitting out chips with Higher mhz quicker than intel at the time...always trying to stay ahead. Intel created the P4 and destroyed that little game and you know the rest.

Intel knew they would take a performance loss, but it was worth it to them because of the ability to scale so high and sell chips.

AMD's PR system is a crock of shiit and pathetic. They really need to re-think that whole plan. They have some killer chips already (and more coming in the future we hope). Just sad they can't call their chips what they are.

 

[Sid03]

<< AMD's PR system was never intended to compare AthlonXP to intel CPU's in the first place, it was a comparison between an Athlon XP and a (theoretical) equally clocked Athlon Thunderbird. >>

this is exactly what i meant in another thread when i stated that "the zealots just eat it up."

so, if the above is true, then you are telling me that a p4 is equal to a t-bird, clock for clock. (i have yet to get a reply on this from any of the zealots in the other threads, so i'll try a third time)

as per amd (and gstanfor) the performance of an athlonxp 2000+ is on par with a t-bird 2000mhz, if one existed. also, an athlonxp is about equal to a 2000mhz p4. so that means that a p4 is equal to a t-bird.

are you going to tell me that you believe that a tbird clocked equally to a p4 will perform about the same?