Does PR rating still have meaning?

Raystonn

Distinguished
Apr 12, 2001
2,273
0
19,780
Please see <A HREF="http://www.rx-7.net/uploads0/1040338389.jpg" target="_new">this chart</A> displaying the clockspeed and PR ratings of the Athlon processors. Would someone explain the odd variations I am seeing? Has there been some official reason given for any of this?

If PR is supposed to be a stable measurement of the processor's performance, why are we not seeing a stable slope of clockspeed/PR? Surely a 333MHz FSB Athlon should require less clockspeed per PR point than the 266MHz FSB processor. Yet we see the reverse?

-Raystonn


= The views stated herein are my personal views, and not necessarily the views of my employer. =
 

eden

Champion
AMD lowered the clockspeed on the new 333MHZ FSBs, and of course altered the PR again. We all knew it could not be linear all the time, because in the end a 2.53GHZ P4 can be better than an XP3000.


--
<A HREF="http://forumz.tomshardware.com/modules.php?name=Forums&file=faq&notfound=1&code=1" target="_new">The THGC Photo Album, send your pics and see others'!</A>
 

Raystonn

Distinguished
Apr 12, 2001
2,273
0
19,780
Please try the link again. Thanks.

-Raystonn


= The views stated herein are my personal views, and not necessarily the views of my employer. =
 

paulj

Distinguished
Feb 15, 2001
523
0
18,980
I am not sure you can measure and develop a graph like that and expect a static slope. I think AMD is trying to give you a comparison to an Intel chip. On a lot of the real life benchmarks the rating fairly closely matches the frequency of an equivalently performing Intel chip. So as the Intel performance varies so does the XP rating. And as the Athlon performance varies so does the XP rating. At the end of the day if you buy an XP 2000+ you want it to perform as well as a P4 2000 MHz processor. That's just my opinion but it seems to hold true.

<font color=red>The solution may be obvious, but I can't see it for the smoke coming off my processor.</font color=red>
 

blah

Distinguished
Dec 31, 2007
2,694
0
20,780
yer brouzer in <A HREF="http://www.rx-7.net/uploads0/1040338389.jpg" target="_new">coo-coo</A> then.

..this is very useful and helpful place for information...
 

Hoolio

Distinguished
Jun 26, 2002
291
0
18,780
I am much more interested that with some benchmark programs an intel 1Ghz chip for example could have a PR rating or 1200 or more, I find it interesting :p


The last time I tested a P4 williamette 1.5Ghz it had a PR rating of 1700 or thereabouts :p
 

Schmide

Distinguished
Aug 2, 2001
1,442
0
19,280
I think its almost common knowledge that the PR rating, although it is suppose to be compared to the Tbird, is actually relative to the P4. For the first 8 ticks of the PR(1500-2200) they were up against the Willamette. Since the Northwood had a bigger cache and faster FSB, it without a doubt scaled better. Thus the rating from 2400-2600 was relatively less aggressive. The final scaling revision for the 333mhz bus XP’s, is more aggressive due to the improvement gained by the improved memory transfers.

Just my spin on it.

Dichromatic for your viewing plesure...
 

paulj

Distinguished
Feb 15, 2001
523
0
18,980
The rating is based on the tbird, but that implicitly creates a relationship to the P4 as well. Back then the performance of the Athlon and the P4 were closer. <A HREF="http://www.amd.com/us-en/assets/content_type/white_papers_and_tech_docs/Benchmarking_Methodology2_v2.5.pdf" target="_new">"AMD Athlon™ XP Processor Benchmarking and Model Numbering Methodology"</A> explains their basis for the rating.

"AMD believes that the idea of solely using a processor's frequency to compare performance between AMD and Intel processors needs to be replaced by a new approach to measure processor performance." This is true. What if I bought a P4 2.0 GHz processor and expected it to perform as well as a AMD Athlon XP processor running at 2.0 GHz (2400+). I would be sorely disappointed.

There should be some sort of standard evaluation for the general public. Of course Intel doesn't mind the higher frequency on their processors because it makes for better marketing. Without a rating most consumers would not know how to compare the two processor brands. That's why Tom's Hardware should be required reading for prospective computer purchasers. :wink:

<font color=red>The solution may be obvious, but I can't see it for the smoke coming off my processor.</font color=red>
 

phsstpok

Splendid
Dec 31, 2007
5,600
1
25,780
The graph seems to prove you point pretty clearly. It sure looks like the curve is adjusted for

1. Palomino vs Willamette.
2. Thoroughbred vs Northwood.
3. Thoroughbred B (333 Mhz) vs Northwood

I wonder if the PR number vs clockspeed will change when AMD switches to Tbred B core for all Athlon XPs.

<b>99% is great, unless you are talking about system stability</b><P ID="edit"><FONT SIZE=-1><EM>Edited by phsstpok on 12/20/02 06:06 PM.</EM></FONT></P>
 

FUGGER

Distinguished
Dec 31, 2007
2,490
0
19,780
Juin just posted links to <A HREF="http://forumz.tomshardware.com/modules.php?name=Forums&file=faq&notfound=1&code=1" target="_new">barton benchmark</A> screen shots, I was quite amused that PR rating on the CPU being so close to Sandra PR rating.

You are limited to what your mind can perceive.
 

phsstpok

Splendid
Dec 31, 2007
5,600
1
25,780
I wish AMD would use Sandra's PR numbers. At least it's something that we, the consumers, could reproduce and verify.

It seems clear that AMD's own numbers are adjusted for marketing purposes, to make their product look equivalent or superior to Intel's. I've always ignored the PR number and accept it as simply a model number.

On a side note (but related), MaximumPC magazine (sorry paper magazine so no links) speculates in the Fast Forward editorial on how Intel might go about marketing Banias. MaximumPC questions how will Intel market that it packs more punch per clock cycle than P4, implying that Intel will come up with PR numbers of it's own.


<b>99% is great, unless you are talking about system stability</b>
 

Col_Kiwi

Distinguished
Aug 8, 2002
429
0
18,780
I've always ignored the PR number and accept it as simply a model number.

As is the best way to do it. Benchmarks show well enough how performance/cycle varies on P4 vs AXP.

The thing that other guy (forgot your name, sorry) said about being adjusted to the three P4 types seems right to me. That kinda shows AMD is being hypocritical when they say their PRs are relevant to a tbird core!

Ah well, nothing's to stop me from ignoring them.

-Col.Kiwi
 

eden

Champion
Ah well, nothing's to stop me from ignoring them.
Stop having a closed mind will ya?

AMD themselves at least admitted the PR had some modifications to do when the XP2400+ and the XP2600+ were "launched", so I think that's gutsy enough for me to still respect them. Honesty isn't too common in the industry.

--
<A HREF="http://forumz.tomshardware.com/modules.php?name=Forums&file=faq&notfound=1&code=1" target="_new">The THGC Photo Album, send your pics and see others'!</A><P ID="edit"><FONT SIZE=-1><EM>Edited by Eden on 12/22/02 10:28 PM.</EM></FONT></P>
 

slvr_phoenix

Splendid
Dec 31, 2007
6,223
1
25,780
Anyone who ever thought to compare an Athlon 1400MHz to an Athlon XP 1500+ knows that the PR system was <b>never</b> intended to compare the XP to the T-Bird.

Honesty isn't too common in the industry.
In the last few years honesty is a burden that AMD has not even remotely attempted to hold on to. I'm not saying that Intel, or even anyone else for that matter has ever been any better. I'm just saying that AMD is certainly no pinacle of truth.

<font color=red>--The turkey may burn and the family may fight,
but video games are forever, so to all a good byte.--</font color=red>
<font color=green>Winter Solstice: <i>The</i> reason for the season.</font color=green>
 

eden

Champion
Which I agree of course, however it's always a fresh air to hear a company admit something.
I recall hearing that SiS admitted their old chipsets had some problems or whatever and that they addressed the issue.
Point is, if some company admits something, well I think it deserves a bit of respect. But no, AMD is not any different from Intel really since they are moving towards an Intel-esque marketting attitude.

Anyone who ever thought to compare an Athlon 1400MHz to an Athlon XP 1500+ knows that the PR system was never intended to compare the XP to the T-Bird.
While that is about 90% true, the remaining 10% does show the XP1500+ sometimes overtaking the 1.4GHZ Tbird, in multimedia applications with SSE. I think you and I can agree on that one, as SSE brought a major advantage for Palominos. The PR IMO is still standing well against similar clocked P4s. In fact if P4s used DDR, the PR rating rapes them in gaming. Did you see the VGA charts? The XP2700+ was beating the 3.06GHZ P4 with HT and DDR333 CAS 2. However I'd like to see a P4 2.8GHZ against an XP2800+ under DDR with i845PE. It should be interesting to see.
Otherwise, with RDRAM, the AthlonXP PR is barely able to compete, though IMO it ties. The close battle with the XP2800 and the 2.8GHZ shows that there was still a bit of life left in the K7, with no 512KB L2 and 200MHZ FSB, like Barton is expected to use. And the multiplier unlocked nForce 2 mainboards option is another thing that creates some value again in purchasing an AMD. I recall over 3 months ago it was barely justified to buy one, except if you're extremly low on cash. Now IMO with the XP2400+ coming unlocked on the nForce 2s as Anandtech confirms it, you can have an extreme FSB and memory speed in sync under Dual Channel aggressiveness, and keep it clocked the same.

--
<A HREF="http://forumz.tomshardware.com/modules.php?name=Forums&file=faq&notfound=1&code=1" target="_new">The THGC Photo Album, send your pics and see others'!</A>
 
Attention EVERYONE!!! You must now cease and desist any conversation regarding the AMD PR rating system!!!

If Intel EVER found out that AMD is actually comparing their chips to Intel chips, and adjusting, then there would be a flurry of lawsuits to follow! So any conversations about this need to be kept hush hush!!!!



Quick!!! Whats the number for 911?
 

slvr_phoenix

Splendid
Dec 31, 2007
6,223
1
25,780
Point is, if some company admits something, well I think it deserves a bit of respect. But no, AMD is not any different from Intel really since they are moving towards an Intel-esque marketting attitude.
Moving towards? Sometimes I think they've surpassed Intel in every Intel-ism except for the marketting budget itself.

While that is about 90% true, the remaining 10% does show the XP1500+ sometimes overtaking the 1.4GHZ Tbird, in multimedia applications with SSE.
In very rare applications (far less than 10% of real-world applications), that <i>don't</i> support 3D Now, and even then it's only just barely. And for <b>everything else</b>, it loses to the 1.4GHz T-Bird.

Come on, you can't really say that just barely winning in an extremely narrow range of software justifies a model number that portrays the CPU as considerably more powerful than it really is, can you? At absolute best, it would have been worthy of a 1350+ rating. (If such a thing existed.) It didn't even come remotely close to a 1.5GHz T-Bird in performance.

I think you and I can agree on that one, as SSE brought a major advantage for Palominos
Actually, I don't think that I do agree. Surely it was a minor advantage, surely it made a lot more sense than pushing 3D Now just because of how well accepted SSE was in comparison, but ultimately, I don't think it really did much at all for AMD. Having AMD market 3D Now optimized compilers that plugged seamlessly into MS Visual Studio would have done more for them than adopting SSE had. At least, that's my observation as a software engineer.

The PR IMO is still standing well against similar clocked P4s.
That depends on your point of view. It certainly won't stand up through all of 2003 though. It'll have to be adjusted once again. And in all of these adjustments, do they ever adjust the already-rated CPUs? That's one thing that seriously bugs me.

In fact if P4s used DDR, the PR rating rapes them in gaming.
And that's an awfully biased point of view to take. You could run Athlons with CAS 3 PC100 SDRAM and compare <i>that</i> to a P4. Why is it wrong to look at it that way? Because it's simply using memory that you <i>know</i> bottlenecks the CPU so that it can't perform to it's fullest. Besides, high quality DDR that <i>isn't</i> running on overclocked timings is only about 5%-10% less in price than an RDRAM solution.

Now IMO with the XP2400+ coming unlocked on the nForce 2s as Anandtech confirms it, you can have an extreme FSB and memory speed in sync under Dual Channel aggressiveness, and keep it clocked the same.
You make it sound as though people <i>aren't</i> OCing high end P4s with PC1066 RDRAM and totally slaughtering the AXPs.

Yes, AMD is coming back into the OCers minds. However, even then as far as absolute performance goes, AMD can't compare to Intel. AMD only wins against Intel in price.

Be this as it may, it still has nothing to do with the very simple fact that AMD's PR rating <b>never</b> made sense when compared to T-Birds. Even from the <b>very</b> beginning, it was wildly skewed to compare to P4s, not to the noticably more powerful T-Birds. AMD still won't admit <i>that</i>.

And who cares if they admitted that they adjusted the PR calculations? This is blatantly obvious information that would have been instantly noticed whether AMD had admitted it or not. So they <i>had</i> to admit it, becuase there was absolutely no way that they could have hidden it.

<font color=red>--The turkey may burn and the family may fight,
but video games are forever, so to all a good byte.--</font color=red>
<font color=green>Winter Solstice: <i>The</i> reason for the season.</font color=green>
 
<i> And in all of these adjustments, do they ever adjust the already-rated CPUs? That's one thing that seriously bugs me. </i>

Now that really is a scary thought. Wouldn't that be such a mess if they actually did change the PR on chips that are already rated! And since it is considered only a "Model" number, there is nothing legally stopping AMD from changing the PR rating.

Heh, so in the future, you could purchase an Athlon XP 2700+, and in reality get what is currently known as a 1500.....and you couldn't do anything about it, except grumble...that would be a hoot!!!

Although, obviously, you would think that AMD wouldn't do this, as it would tear their sales apart, and destroy customer loyalty. However, if they did it on a small enough scale.....




Quick!!! Whats the number for 911?
 

eden

Champion
In very rare applications (far less than 10% of real-world applications), that don't support 3D Now, and even then it's only just barely. And for everything else, it loses to the 1.4GHz T-Bird.
I realize my argument is weak but I'll have to contend with the following for now. Notice that in MP3 encoding the XP1500 does majorly better than the 1.4GHZ Tbird. And also scroll some pages, notice MPEG 2 performance. It is clear MM is affected by the advent of SSE.
<A HREF="http://www17.tomshardware.com/cpu/20020107/p42200-13.html" target="_new">http://www17.tomshardware.com/cpu/20020107/p42200-13.html</A>

Surely it was a minor advantage
It is significant in multimedia for many applications, and Anandtech supposedly got a 90% boost once. I don't know how true it is, but I know that SSE was something missing.

It'll have to be adjusted once again.
They only adjusted the XP2600 and XP2400 for some reason I forgot. I'd check the article where it was paper launched to see why. But they did not modify any other previous cores' model.

You could run Athlons with CAS 3 PC100 SDRAM and compare that to a P4. Why is it wrong to look at it that way? Because it's simply using memory that you know bottlenecks the CPU so that it can't perform to it's fullest.
How many people do you think are paying for the premium of PC1066 systems?
I can be sure when I say that the majority of P4 owners are buying DDR memory. So it's safe to say for gamers, that DDR will be their buy if they want value for performance. Yes PC1066 is still the absolute winner, there is no doubt, and it easily overtakes AthlonXPs. However the least AthlonXP system you can get from the worst OEMs is one with DDR 266. Least from P4 OEMs is lol, SDRAM of course.
However you were talking about the CAS 3 100MHZ RAM equipping, so that's all I have to say about common systems on both sides of the platforms.

Besides, high quality DDR that isn't running on overclocked timings is only about 5%-10% less in price than an RDRAM solution.
I disagree. Around town, 512MB PC1066 that is actually OCZ, a barely reputable brand name for quality RDRAM, is an insane 559$ CDN. DDR 333 from Crucial, 512MB is roughly half that, and Crucial is known to be a very high quality RAM maker. All for 249$. Here is the proof: <A HREF="http://www.shoprbc.com/ca/shop/?cid=162" target="_new">http://www.shoprbc.com/ca/shop/?cid=162</A>
So while currently DDR333 is the top and certified DDR and PC1066 is the official top performer for Pentium 4 systems, it is clear that your point is not as valid anymore. DDR400 is also not too expensive. Actually CAS 2 DDR400 from OCZ, rated as PC3500 for 433MHZ usage if needed, is 365$, still far from RDRAM.

You make it sound as though people aren't OCing high end P4s with PC1066 RDRAM and totally slaughtering the AXPs.
I agree, I was a bit disregarding, my bad.
Except that those who would buy say, a 2.4GHZ P4 and OC it to 3GHZ on air, are often trying to get value, and again those users will buy DDR systems. (I know it because I got two friends who have bought a 2.4GHZ with DDR and wish to overclock later on, and of course were short on money) If you can get an XP2400 unlocked, overclock it to say 2.2GHZ, then up the FSB to 200MHZ, keep the same clock, use the nForce 2 Dual Channel aggressive timings, you will get similar value, if not better. I would imagine the P4 still would kick ass in multimedia with SSE2, however you are still getting somewhere with the AMD system, which is not a bad thing at all.

However, even then as far as absolute performance goes
Again, as I said above, it is slowly changing in my perception, because I still think that most P4 overclockers will opt for a DDR setup, and as seen in THG's VGA Chart benchmarks, it will not perform as good as an AXP that is not even using the same PR to clock rating. In other words the XP2700+ is able to beat the 3.06GHZ, but then again HT might be hurting it more now than helping.
Yes though Intel with Rambus, are still the absolute performance winners. But again, with the RDRAM price, what's the worth really? I wish it had been like 200$ CDN less, then it'd be widely considered. You yourself seem short on cash when you had announced the system for Xmas that you wish to build. Can you opt for the RDRAM system?

Be this as it may, it still has nothing to do with the very simple fact that AMD's PR rating never made sense when compared to T-Birds. Even from the very beginning, it was wildly skewed to compare to P4s, not to the noticably more powerful T-Birds. AMD still won't admit that.
And I am not against you on that one. It sure isn't worth comparing to a Tbird, especially when it takes 66MHZ for 100 PR points. No Palomino rising 66MHZ will compare to an unproportionally rising 100MHZ of Tbird performance. But then again, would they have ever sold this many Athlons if they had kept real clock speeds to sell?

--
<A HREF="http://forumz.tomshardware.com/modules.php?name=Forums&file=faq&notfound=1&code=1" target="_new">The THGC Photo Album, send your pics and see others'!</A><P ID="edit"><FONT SIZE=-1><EM>Edited by Eden on 12/23/02 04:28 PM.</EM></FONT></P>
 

ejsmith2

Distinguished
Feb 9, 2001
3,228
0
20,780
RAYSTONN!!!!!

DUDE!!

You're still alive!!!!

I've been wondering about you for a while. God I was missing these propoganda threads. And the forums have been seeming a little dull lately. I totally hope this is like, a Renaissance or something. With all the company people in here, maligning and debasing each other, and entire flaming threads and week long pouts.

And Wusy and AMDinside trolling.

What we really need is some good thermodynamic debates, too.

WELCOME HOME!

AND MERRY CHRISTMAS!

"I personally think filesystems should be rewritten from scratch every 5 years..." --- Hans Reiser
 

lhgpoobaa

Illustrious
Dec 31, 2007
14,462
1
40,780
bah. just put the chip in and run it faaast :smile:

<b><font color=purple>[Rik_]</font color=purple> I wonder how many people have made their own phasechange system?
<font color=blue>[LHGPooBaa]</font color=blue> I get phasechange whenever i eat a hot chillie :lol: </b>
 

eden

Champion
Yup, that's ol' Ray... :wink:

--
<A HREF="http://forumz.tomshardware.com/modules.php?name=Forums&file=faq&notfound=1&code=1" target="_new">The THGC Photo Album, send your pics and see others'!</A>