NEW: Internal Intel Document Slams AMD.

[AGodspeed]

http://www.amdzone.com/articleview.cfm?articleid=1033

Thanks to NFS4 for the heads up in this thread.

A few comments...

First off, I find it very strange that Intel uses SPEC, Sysmark, WebMark, and Content Creation to prove to channel partners that the "Pentium 4 processor 2.2GHz establishes unrivaled performance leadership" and that the ?Athlon XP 1.67GHz does not live up to its PR rating of 2000.? These benchmarks are almost as useless as 3DMark and SiSoft. Real application performance is the only thing anyone is going to care about, right? Who in their right mind would base a purchase on Sysmark, I would be looking at video, audio, gaming, etc. applications to see how a system compared to the other, not Sysmark.

Secondly, Intel makes another very confusing conclusion, stating that "The Press, Analysts, and Review Sites are already picking up on the excitement surrounding Northwood", however this conclusion is based on a random statement from Hardware Central referring to the prospect of the performance of an extremely highly overclocked system of 3GHz. Review sites like Anandtech, Aceshardware, Tech Report, Tomshardware, and many other review sites have published very neutral or uninspiring conclusions about Northwood versus the Athlon XP. Unless Intel is referring to hardcore overclockers (snickers), it's clear that their conclusion is off.

In this example, Intel makes another very confusing conclusion, stating that "Higher Frequency + Bigger Cache = Better Performance", yet Intel states in this very same document that "Thoroughbred is just a die shrink, so it will only help with frequency, not performance." Anyone else see the flawed logic here?

The last thing that confuses me is Intel?s list of OEM support. For some reason, Intel thinks that Alienware, Emachines, Fujitsu, NEC, and Systemax don?t sell Athlon XP systems. Read here and then here. Quite confusing.

Please, discuss.

 

[Priit]

Well, it's pretty obivious that document was not meant for computer freaks who actually know how things really are (more or less). It's not hard to pick tests/benchmarks where P4 runs considerably better than Athlon and vice versa. Typical marketing FUD, not much to discuss about IMO...

 

[Adul]

heh intel and it marketing. I hope amd will do something to counter this. like release hammer early.

 

[christoph83]

It's not hard to pick tests/benchmarks where P4 runs considerably better than Athlon and vice versa. Typical marketing FUD, not much to discuss about IMO...

Ditto.


First off, I find it very strange that Intel uses SPEC, Sysmark, WebMark, and Content Creation to prove to channel partners that the "Pentium 4 processor 2.2GHz establishes unrivaled performance leadership" and that the ?Athlon XP 1.67GHz does not live up to its PR rating of 2000.? These benchmarks are almost as useless as 3DMark and SiSoft.

I've seen Tom's and Anandtech uses Sysmark and Content Creation in their reviews. If they are so useless then why dont they just take them out? Why do we crown a speed king based off these benchmarks? Anyway I dont see why we are trying to analyze Intel's Marketing Deparment

 

[Sid03]

<< First off, I find it very strange that Intel uses SPEC, Sysmark, WebMark, and Content Creation to prove to channel partners that the "Pentium 4 processor 2.2GHz establishes unrivaled performance leadership" and that the ?Athlon XP 1.67GHz does not live up to its PR rating of 2000.? These benchmarks are almost as useless as 3DMark and SiSoft. Real application performance is the only thing anyone is going to care about, right? Who in their right mind would base a purchase on Sysmark, I would be looking at video, audio, gaming, etc. applications to see how a system compared to the other, not Sysmark. >>

which benchmarks (specifically) would you consider useful? why do i get the feeling that you'll rattle off a bunch that favor amd?



<< In this example, Intel makes another very confusing conclusion, stating that "Higher Frequency + Bigger Cache = Better Performance", yet Intel states in this very same document that "Thoroughbred is just a die shrink, so it will only help with frequency, not performance." Anyone else see the flawed logic here? >>

i guess you missed the big orang bubble which says "expect a 1.67ghz t-bred will have the same performance as a 1.67ghz palomino"... obviously their point being that there are no enhancements to the cpu outside of the shrink to .13 micron. (unlike the northwood, which is faster than a willamette at the same clockspeed.)

it seems as though you are just someone who favors amd over intel. if you want to start bashing for marketing, amd has more than their share as well.

for the record, i own both a tbird 1.4ghz and a northwood 1.6@2.2ghz... (at least until the t-bred comes out, then i'll swap my bird for a bred. i think i'll keep my northwood for a while. it's an awesome cpu. hopefully my t-bred system can be as quiet as my northwood system with no worries.)

 

[KenAF]

<< First off, I find it very strange that Intel uses SPEC, Sysmark, WebMark, and Content Creation to prove to channel partners that the "Pentium 4 processor 2.2GHz establishes unrivaled performance leadership" and that the ?Athlon XP 1.67GHz does not live up to its PR rating of 2000.? These benchmarks are almost as useless as 3DMark and SiSoft. >>

I don't agree at all. 3DMark and Sisoft are simulated benchmarks that have nothing to do with real world applications or games. Sysmark attains its benchmarks by running through scripted tasks in Office XP (Word2002, Excel2002, Powerpoint2002, etc), while WebMark attains its results through loading of various complex web pages in the web browsers we use, among other things. Most people (although perhaps not the HS kids that frequent this forum) spend far more time working on their computers in these applications than they do gaming; I know I do.

I've heard other's say that at the speed of processors today, there is no real world difference in Office applications. This is baloney. The Office and productivity tests are at least as relevent as the 120 vs 150 fps gaming tests for those that do real world work. While you aren't going to see any difference between a 1GHz Athlon and 2.4GHz P4 when just typing some report in Word XP, you do see the difference if you are doing any sort of serious work in these applications. You see it in creating complex Powerpoint presentations with animation and video, and you definitely see it when doing financials work (where you may have sheet after sheet performing calculations based on other sheets...which base their results off calculations in still other sheets).

Also of note, in Hardocp's latest review, Kyle found the Athlon to thoroughly best the P4 in tests with Office2000, while the situation is reversed with Office XP, giving some credence to Intel's claim that the latest applications tend to do better on its processors.

While I don't disagree with the use of a performance rating system, I do agree with Intel on the inappropriateness of AMD's current system moving forward. According to AMD, their performance rating system is and will continue to be based on the original Athlon, long after its life has passed. With it's performance rating system, AMD can basically ignore any improvements Intel makes, be it with a faster bus or more execution units, because they are basing their rating on an older [AMD] processor. In a year, AMD's performance rating system could be highly misleading as to comparative system performance. In my view, performance ratings should reflect actual performance relative to shipping processors, not past products that no one is buying. Moreover, the ratings should be assessed based on some third-party firm, not the needs of AMD's marketing department. Of course, this is problematic, as ratings for processors will change depending on what the competition is doing (i.e. 2200+ might no longer be 2200+ with 533FSB, and 3400+ might no longer be 3400+ when Intel significantly reworks the P4 with 100+ million transistors next year).

 

[MadRat]

Internal documents are not marketing, they are dogma.

 

[tenoc]

Company I and company A both make sports cars.

Both cars have the same weight and horsepower, same acceleration, same top end.

Salesman for I says, "It doesn't matter what the performance is. Ignore the horsepower
and acceleration and top end. Our car revs to 10,000rpm, theirs only revs to 7000rpm.
Therefore, ours is better and only costs twice as much!"

 

[EdipisReks]

Salesman for A says "look, our car only revs at 7000rpm but performs the same as their 10,000rpm, so our car is obviously more efficient. we call our reving quantirev and give it a rev rating of 10,000+" Customer says "what about I's new car that equals your car when reving at 8500, but can go all the way to 12,000. shouldn't you modify your quantirev rating? their 10,000 no longer your 7000". salesman: "no, you see, our quantirev was measured against one of our older cars, not company I's car". customer: "but doesn't that mean that the quantirevs are now useless and misleading....?"

it cuts both ways, buddy.

 

[tenoc]

And the price differential for the same performance?

 

[EdipisReks]

hey, i paid less for my 1.6a PIV than i did for my 1.0 gHz athlon. if you are talking about overclocking (and i hope you are) intel provides just as good if not better bang for the buck than AMD.

--jacob

 

[BlvdKing]

This document is pure FUD.

 

[KenAF]

The document is no more inappropriate than the benchmarks AMD quotes on it's site (www.amd.com). Each vendor quotes the benchmarks where it's processor does best. Big surprise! Give me a break.

AMDZone's rebuttal amounts to "the P4 has RDRAM and we don't." Honestly, that argument just doesn't cut it.

 

[formulav8]

Intel must definitely be worried about something or they wouldn't be acting like cry babies. They may also be afraid of people getting educated on the mhz = performance myth.



Jason

 

[Sid03]

<< First off, I find it very strange that Intel uses SPEC, Sysmark, WebMark, and Content Creation... These benchmarks are almost as useless as 3DMark and SiSoft. Real application performance is the only thing anyone is going to care about, right? Who in their right mind would base a purchase on Sysmark, I would be looking at video, audio, gaming, etc. applications to see how a system compared to the other, not Sysmark. >>

check out the list of benchmarks amd uses...

apparently amd thinks that sysmark, content creation, and 3dmark are valid benchmarks to use. of course, not the latest version of each benchmark, though. notice that outside of last year's content creation, amd uses no video or audio benchmarks.

 

[Jman13]

<< hey, i paid less for my 1.6a PIV than i did for my 1.0 gHz athlon. if you are talking about overclocking (and i hope you are) intel provides just as good if not better bang for the buck than AMD.

--jacob >>



Ok, you can't compare prices from a long time ago to prices today. Otherwise, I can complain that 486s were SO overpriced, because they were $5000 for a good machine in 1992, and this year, I built a kick ass AthlonXP system for less than $1000!!!

As of right now, the Athlon 1.0GHz is $76.
However, The AthlonXP 1600+ (1.4GHz) is only $93. And the 1800+ is only $109.

The P4 1.6A is reasonably priced at only $114, but it's still more expensive than an AXP 1800+ which at stock speed KILLS it in performance. Either way, the P4 1.6a is 40 bucks more than a 1GHz Athlon...but the 1GHz are expensive since they are so rare right now.

 

[christoph83]

Internal documents are not marketing, they are dogma.

These dont seem to be internal documents but the slides used in the presentation. I dont see what the big deal is.

 

[Diable]

<< Intel must definitely be worried about something or they wouldn't be acting like cry babies. They may also be afraid of people getting educated on the mhz = performance myth.



Jason >>



I doubt their worried since every company on earth releases internal memos and documents that slam their competitors. Until AMD starts running adds on TV educating the greater populus i.e. non geeks on the "mhz myth" I don't think Intel has anything to be afraid of.

AGodspeed, SPEC, Sysmark, and Content Creation are used by almost every hardware site in their cpu and motherboard test so I don't see whats wrong with Intel using them. Your also overlooked that fact that most people that buy complete systems from OEM's don't encode or edit video much less audio, unless by audio you mean making mp3's? Games are a good barometer but which ones do you use? Games that favor Intel's processors like Quake3 or games that favor AMD's processors like Serious Sam?