AMD PR ratings, new chips, 333mhz bus, clawhammer

[drogue]

read this article from CNet

couple things to point out:
the PR ratings are being directly compared to P4's by a reasonably reliable news source.....in addition to being unwise, thats also illegal for AMD to do (i think its illegal at least, something about marketing laws: any lawyers/paralegals wanna back/correct me on that?). they need to clarify that it is rated aginst Tbirds or lose the PR system. i understand the marketing implications, and thats what they REALLY mean, but when news sources report like that...its simply gotten out of hand. i agree amd needs a good way to market their chips, but thats what marketing majors are for right? i have full confidence in AMD's ability to market the xps better than that.

clawhammer debuts at 3000+? perhaps amd has lowered their sites. it seems all the other previews/information points to a 3400+ debut.

333mhz bus for athlons. i cant wait! they might really be able to get back into the game against the top end p4's. prescott will really wallop even a 333mhz bus athlon, but prescott is still quite a ways off. a good mobo with kt400 implementation (true DDR 400 for athlons) and a 333 mhz bus for the proc will really put the top end athlons in striking distance, or even equal to the top of the line p4s (i hope).

 

[Duvie]

333mhz with 400ddr....Not really...Do the math...The athlon at 333mhz fsb wil produce 2.7gb/s of bandwidth and so the pc3200 (400) will show the kind of performance pc2700 on kt333 mobos show with a stock 266fsb athlon...NOT MUCH....

P4 is different...the quad pumped makes it so even at 533fsb p4bs the pc3200 ddr doesn't deliver all the p4 can take...


I agree on the pr rating!!! It obviously of late needs a downgrading with the northwoods 512kb cache and 533fsb...PLus the fact that amd scales at 100pr points per 66mhz..I Dont think so!!! I know they mean Tbird but they are playing the consumers just as much and lkely should be sued as well...

 

[Budmantom]

"the fact that amd scales at 100pr points per 66mhz..I Dont think so!!!"

At the same time you can't compare an amd xp2200+ (1800mhz) to a p4 1800, out of the box performance the amd is alot faster.

 

[drogue]

regardless of the math, a 333mhz-bus athlon can simply handle more incoming data, the 400mhz ram simply ensures that the athlon takes everything it can. at 333mhz ram, missed cycles and the like will keep the athlon from recieving enough data to be really efficient within a single clock cycle
i'm not sure what the real world performance gain is, but it seems like 5-6% would seem reasonable.an extra 5-6% on top of the faster clocks (extra 2 or 3%?) should increase the xp architecture enough to allow the 2800+ to compete with the 2.8 p4. although, right now its still speculation, and indeed, the 2800+ may not have a 333mhz bus, it may take the next speed to introduce that, if they ever produce anything past 2800+. either way, time will tell.

edit: oh yeah, as long as AMD states its rated against its own product they cant technically be sued

 

[Duvie]

One thing your percentages only will mean something in ceratin apps...I have studied this and not all appz will gain the same or have the same use of the extra bandwidth....so your 5-6 then 2-3 is extremely ambitious and best case scenario....

Ie there are apps that pc800 with a 400mhz fsb chip only does 3 percent lees in eprformance versus pc1066 with a 533fsb...They have a 800mb/s differing in bandwidth...


I agree on the stating it....Hwever they do not state it enough and should stae it with each chip sold....

At the same time you can't compare an amd xp2200+ (1800mhz) to a p4 1800, out of the box performance the amd is alot faster.

that is a product of differing architecture...you need not preach to the choir about the mhz myth....The bottomline is performance...

The question still goes and you did not answer why does 66mhz real clock equal 100 pr points...I mean the xp 1600 is not that muh faster then the 1.4ghz tbird c chip!!! Where did they arrive at this number???

 

[drogue]

i would assume they got that number from their benchmarking system. they do actually have sets of tests the run to determine what speed they list it at. i think i saw the list of apps they use, but to be honest i cant remember where i saw it or what was on it. its safe to assume its commercial programs/utilities that are hand selected to perform well with the xp procs, which is probably how they arrive at 1600. but then again, i dont work for amd so i cant say for sure about that last part.

 

[JMicaelK73]

I'm sorry to ask but in what way is it illegal by AMD to use the PR system. They newer say that their processors are runing @ some GHz number, they only rate the processor with a "+" rating, don't see the illegal thing there. But if there is something that I'm missing please tell me.

Micael

 

[CrazySaint]

No, there's nothing illegal about the PR rating. If there were, you can be sure that the PR rating would have been dead long ago.

 

[BowlingNut]

i think the legality question lied within the amd rating against P4 chips.....its perfectly legal (though i think, unwise) for them to rate them against the "tbirds"

 

[Boonesmi]

AMD always claimed the PR rating was comparing the athlonXP to the Tbird... NOT comparing it to the pent4

but many people still think it was supposed to be a rating to compare it to a P4

 

[BowlingNut]

Originally posted by: Boonesmi
AMD always claimed the PR rating was comparing the athlonXP to the Tbird... NOT comparing it to the pent4 but many people still think it was supposed to be a rating to compare it to a P4

well, most of us around here know that, but a quick looksee at the article reveals this:

Under the company's model numbering system--designed to reflect the actual performance of the chip rather than its clock speed--the new 2400+ and 2600+ chips will perform as well as or better than Pentium 4 chips running at 2.4GHz and 2.6GHz. The 2400+ and 2600+ could have clock speeds in the range of 2.06GHz and 2.2GHz, based on AMD's previous Athlon XP models. Its Athlon XP 2200+ chip, for example, runs at 1.8GHz.

seems to be an awfully common mistake. i think technically it could be illegal for AMD to rate their chips against the P4 b/c of performance differences...but i'm no lawyer, so i'm open to enlightenment.

 

[mechBgon]

I wish they'd hurry up and spit out Barton already.

Back to my previous post, what I might not have expressed very well is that a system benchmark would be useful to the everyday consumer for gauging how the system as a whole will perform. Content Creation 2002, for instance, would give a higher result on a system with a faster model of hard drive, all other factors being equal. Maybe I'll try to look up Consumer Reports' contact info and send them an email saying we need more than just a red circle with a dot in the middle, we need some CC2002 benchies to help educate the average consumers that actual system performance isn't just a matter of GHz, Mb and Gb. What do you guys think of that?

 

[DAPUNISHER]

"Educating the average consumer" is an oxymoron.

 

[BowlingNut]

Originally posted by: DAPUNISHER
"Educating the average consumer" is an oxymoron.

true, ranks right up there with "military intelligence" and "corporate efficiency"

edit: for all you military people: that is a joke for all you corporate people....well, no joke for you

 

[mechBgon]

Originally posted by: BowlingNut

Originally posted by: DAPUNISHER
"Educating the average consumer" is an oxymoron.

true, ranks right up there with "military intelligence" and "corporate efficiency"

edit: for all you military people: that is a joke for all you corporate people....well, no joke for you

LOL!

I was mulling over the developments on AMD's horizon and remembered that at some point, the Duron name is supposed to be stuck onto the present AthlonXP core. Maybe they'll simply re-label the present AthlonXP as the new Duron at some point, perhaps when the 2800+ appears on 333MHz bus.

 

[DivideBYZero]

It's got to be Barton on 333Mhz all the way for my next upgrade, then wait for Hammer to settle and go to that maybe 6 months after introduction.

Unless something that knocks my socks off comes out from Intel at a realistic price.

 

[BowlingNut]

so......any idea when barton is gonna hit stores? i dont really carea about the 333mhz bus since the performance wont be very much greater, but the 512kb cache seems like a pretty big deal.

 

[mechBgon]

Well, the AMD roadmap shows it in the second half of this year, but that's no help. Guess we'll see...

 

[dexvx]

XP simply doesnt scale as good as the P4. What you're saying about 1.6Ghz P4 vs 1.6Ghz XP, the XP simply beats the p4 by a large margin. At high Mhz, tis a different story.

Go look on HardOCP's Radeon9700 Pro review. They are running an XP on a "new unreleased core" at 15x165 = 2.48Ghz (not PR). At that speed, it would have a massive PR rating (over 3000+), but it barely edges the p4-2.53 in all the benchmarks (+5 % avg, I'd say). It may be 3000+ to a Tbird, but its not more than 2700+ to a P4 with a 533 bus.

The only noteworthy thing is that the p4 uses PC1066 RDRam.

 

[SupermanCK]

like i said before...pr rating means jack...
real life performances would be the key...

 

[BowlingNut]

Originally posted by: dexvx
XP simply doesnt scale as good as the P4. What you're saying about 1.6Ghz P4 vs 1.6Ghz XP, the XP simply beats the p4 by a large margin. At high Mhz, tis a different story. Go look on HardOCP's Radeon9700 Pro review. They are running an XP on a "new unreleased core" at 15x165 = 2.48Ghz (not PR). At that speed, it would have a massive PR rating (over 3000+), but it barely edges the p4-2.53 in all the benchmarks (+5 % avg, I'd say). It may be 3000+ to a Tbird, but its not more than 2700+ to a P4 with a 533 bus. The only noteworthy thing is that the p4 uses PC1066 RDRam.

well, i heard a story somewhere (not sure where) that for the new processors, the PR ratio was going to change. i'm not sure how its going to change, but i suppose we'll see. 2.48Ghz? the 2400 and 2600 are supposed to be running at 2.0 and 2.13ghz repsectively. one would assume that the 2800 (even with the 333mhz bus) will run in the 2.2-2.3range, perhaps even 2.35 - 2.48 would indeed be a huge jump...perhaps a new barton core? if it is a barton, its probably a pre-production model, and thus not likely to yield incredible performance numbers until things have been finalized and optimized. hmmm, food for thought

in any case, the athlon will always be rated against the tbird, so 3000+ could be correct - since we all know that amd would "never" rate their procs against intel's....

 

[dexvx]

My bad on that part.

The HardOCP benches were at 1024x768, so not entirely CPU limited.

I would expect a 2.5Ghz XP to be at least a 3000 p4. I doubt its pre-production, lower performing part seeing that Kyle will do a review of it sometimes within the following week, as the NDA cancels.

 

[MajinVegeta]

The issue isnt clock speed but chip design, at higher speeds the chips decoder units cant keep pace, 1/2 the P4 core(not counting caches) is decoding units.

Decoding units make up most of the core of the chips, the reason the P4 is so huge and scales so well is massive decoding units can feed the chip at higher Mhz.

This is why the P4 is so huge it will need theses massive decoding units when HT is enabled without them the pipelines wait for the decoding units to move onto the next stage.

The problem I fore see with hammer is with adding the northbrigde to the core. Your taking up very costly SOI wafer space that could be used for decoding units and cache to scale better at higher Mhz.

 

[SSXeon5]

Well I did see the P4 2.53GHz keep up with the 2.48GHz 333Mhz fsb XP (lol). And Your kidden me that you still think the PR was ment for the Thunderbird, ahahaha. There is no way a 1.53GHz XP equals a 1.8GHz TBird, then i guess the 1.3GHz P4 is clock to clock as fast as a 1.3GHz Thunderbird

Originally posted by: dexvx
My bad on that part.

The HardOCP benches were at 1024x768, so not entirely CPU limited.

I would expect a 2.5Ghz XP to be at least a 3000 p4. I doubt its pre-production, lower performing part seeing that Kyle will do a review of it sometimes within the following week, as the NDA cancels.

No No they were 1280x1024, and 1600x1200 so your theory is shot And the sad part is, it is also a 2.48GHz 333Mhz fsb XP (lol). But i also did a thread about the new PR and how i guess amd is madly losing IPC with the high clocks they are getting That or they are figuring there PR should be different with the release of the 533Mhz FSB Northwoods

SSXeon

 

[MajinVegeta]

Like I was saying SSXeon.

This is why Itanic is so powerful in its native state. But once it have to do a decode for an x86 instruction it curls up and dies, becuse it doesnt need massive decoding units like all current x86 chips.