Did anyone trust BAPCO anyway?

[Rugor]

Both Vanshardware and the Inquirer have been commenting on the recent revelations that the reason the P4 does so much better on Sysmark 2002 than Sysmark 2001 is because of the changes to the test.

As an example, all the tests in one category that the Athlon beat the P4 in have been removed. In the new version that category consists of repeats of the tests the P4 excelled in plus a few more similar tests.

If it was common knowledge that the tests removed were for things not in common use, and the ones the P4 excels in where commonly used, then this revision might make sense. However, all the available information indicates the removal was completely arbitrary. Arbitrary unless the goal was to create a benchmark to showcase the P4 against the Athlon.

The end result has been to turn the suite from a CPU test to a Memory bandwidth based system test. There's nothing wrong with that, or wouldn't be if it was advertised as such. But it's marketed as a CPU test.

I won't go into the fact BAPCO was once headquartered in the same building as Intel, and that Intel chairs it.

The truth of the matter is that there are some things in which the P4 excels, and I may lose my AMD fanboy status for saying that. But it's true, P4 is better for things that require a lot of bandwidth but not much computation.

I just don't like so much being placed on such a skewed benchmark.

[gibsinep]

Well, I think Intel is just floppying around some of its money. :( :rolleyes:

[absalom]

Originally posted by Rugor
Both Vanshardware and the Inquirer have been commenting on the recent revelations that the reason the P4 does so much better on Sysmark 2002 than Sysmark 2001 is because of the changes to the test.

As an example, all the tests in one category that the Athlon beat the P4 in have been removed. In the new version that category consists of repeats of the tests the P4 excelled in plus a few more similar tests.

If it was common knowledge that the tests removed were for things not in common use, and the ones the P4 excels in where commonly used, then this revision might make sense. However, all the available information indicates the removal was completely arbitrary. Arbitrary unless the goal was to create a benchmark to showcase the P4 against the Athlon.

The end result has been to turn the suite from a CPU test to a Memory bandwidth based system test. There's nothing wrong with that, or wouldn't be if it was advertised as such. But it's marketed as a CPU test.

I won't go into the fact BAPCO was once headquartered in the same building as Intel, and that Intel chairs it.

The truth of the matter is that there are some things in which the P4 excels, and I may lose my AMD fanboy status for saying that. But it's true, P4 is better for things that require a lot of bandwidth but not much computation.

I just don't like so much being placed on such a skewed benchmark.

SysMark 2002 is skewed towards making the Intel Pentium 4 results look better than the AMD CPU here goes the site from Vanshardware.com http://www.vanshardware.com/reviews/2002/08/020822_AthlonXP2600/020822_AthlonXP2600.htm
http://www.vanshardware.com/reviews/2002/08/020822_AthlonXP2600/SYSmark%202002%20Analysis%20Presentation%20FINAL.p df:t Thanks Rugor

[NDD]

Few good facts, BUT ...
P4 got SSE2, Athlon XP DON'T.
P4 has greater bandwidth. 3.2GB/s vs. 2.1GB/s IS a difference.

Besides, subjectivity of many benchmarking programs is a well known fact :rolleyes:

[$1500-P4 gamer]

"Besides, subjectivity of many benchmarking programs is a well known fact"

So true!;) :t

[gibsinep]

The only true benchmarking is the ones you for yourself. :p ;)

[Rugor]

I don't deny the higher bandwidth is a difference, I expect the P4 to win at bandwidth intensive tasks, that's what it's designed for. What I object to is someone taking a so called "CPU Benchmark" and without telling anyone, turning it into a "Bandwidth Benchmark." Meanwhile it's not marketed as such, but as a raw CPU benchmark.

That and removing all the test the Athlon beat the P4 on is just a little fishy, especially when half of what replaced them are tests the P4 was better at.

I know benches are usually skewed, but this is clumsy, heavy-handed and ugly.

[NDD]

Chill out, Rugor, have a "Kit-Kat" ;)

[wallie_x]

I think a lot of us are just getting plain tired of the corruption seen in big business. The heavy handed tactics of M$ squash all competition attitude while ole B. Gates stating,"Anti-competetive? Certainly not me?" is just plain disgusting. Intel shows itself as no different. My own opinion is that Intel thinks most people who buy their CPU/chipsets are dummies: brain washable by TV adds that is. (However, their prominence in America shows the latter premise as somewhat true.) Anything that makes them look cool or glittery, even if it’s false, is fair to use in advertising. e.g. the MHZ = performance tripe they continue to try and sell us. Stacking the deck in benchmarking just shows repetition of an old behavior, using deception to promote sales.

[gibsinep]

wallie_x- so true.:(

[wallie_x]

Truely, I didn't know how REALLY bad it was until I read 'Van's Hardware' page. He's giving up his web site. And it sounds like after some serious intimidation from ole Inhel and (or) BAPCO. If the P4 Northwood is so great, why can't it rest on its own laurels? Instead Infidel and DIPCO cook the benchmarks to make the P4 look even better. If its so much better, then why does it need a 700mhz handicap and a 1.1GB/s greater memory transfer rate? (P4 2.8ghz vs. XP 2600+ running @ 2.1ghz) If someone in golf or another sport beat a pro by using a very large handicap, who would think anything of it? In one sense the P4 is a dunce when compared to the Athlon XP by the latters vastly superior FPU and IPC ratio. I used to think a little better of Intel, now I feel like returning to my old mantra: The P4 SU*KS
(Not that it does, but because I believe Intel does.) :r

[gibsinep]

Wallie- once an Intel hater always an AMD lover. :D

[$1500-P4 gamer]

Where do I fit in that equation? I have...
1 P4 sys/rdram of course on Asus P4TE
1 Cell II 800mhz system
1 Duron 1gig
and now coming soon to the family I'm getting a T-bird 950 and mobo cheap so add that to the list. That one is for recording my tv shows like HDTV!;) Now before I guess you could say since I had 1 more Intel sys that I was prefering Intel but now its 50/50:p

[tking]

Originally posted by wallie_x
...after some serious intimidation from ole Inhel and (or) BAPCO.... Infidel and DIPCO cook the benchmarks to make the P4 look even better ... In one sense the P4 is a dunce when compared to the Athlon XP by the latters vastly superior FPU and IPC ratio.

If people are intimidating and cooking numbers to make something look better than it is then they need to be caught and shown as liars. It's hard to believe (with the strength of the media) that a hot story like "INTEL LIES!" wouldn't happen if they did.

I'm far from a big fan of modern business, but putting that aside I'd really like to try and understand the engineering differences between AMD and Intel.

P4$1500G has explained it before and (if I'm getting it), the P4 went the specialists path while the Athlon is a generalist. If the Athlon uses brute force to process, the Intel uses finesse. Of course if you program outside of what the Intel can do, you're SOL. Conversely, if most programming is done within the P4's specialties, then it will run away with the prize.

Intel's business practices are dictated by the sleezy business class - probably guys who graduated with the former managers of Enron. I can hardly advocate their actions.

Back in the recesses of Intel there are a lot of engineers though, and the work they do is what really powers the company. You might be able to criticize the business guys (they're really just middle men at the end of the day though). The real value at Intel are their engineers. Politics aside, is the P4 really a dunce, or simply the expression of a particular engineering philosophy?

T

ps: since I'm on engineering, no one ever seems to bring up the physics management of the chip. From heat to cooling systems, the Intel is much more user friendly. Isn't that all part of the 'best engineered' equation too?

[wallie_x]

tKing,
I was a little flustered when I wrote that. Aside from the psychological value of venting steam at perceived injustice, if you cut through the anger I believe I have a strong point. The P4, while having a lot of merits, could also be thought of as sheer ploy by Intel to extort the value of the MHZ = performance myth. It's designed with the ability to ramp-up its clock speed. The P4 Willie was somewhat of a dunce if one screens by synthetic benchmarks. The P4 Northwood is an example of engineering overcoming the shortcomings of a design. The .13 micron change was planned, but was the increase in L2 cache? Or the adding of an extra over layer for better interconnections? I think not. We’ll never know for sure, but If one were to speculate current trends indicate Intel continues with playing on the ignorance of Americans by promoting deception. The P4 Celeron is an absolute retard in some benchmarks yet runs at 1.4 mhz or greater. Moreover, the circumstantial evidence is not good for Intel. Intel and BAPco were deeply interconnected before AMD joined. The performance of the Athlon XP has been a hurtful thorn in Intel's side. I think they tried to remove it through devious means.

[tking]

Originally posted by wallie_x
...Aside from the psychological value of venting steam at perceived injustice, if you cut through the anger I believe I have a strong point. ...but If one were to speculate current trends indicate Intel continues with playing on the ignorance of Americans by promoting deception....

Fair enough on the venting steam part of things. Venting is all a part of the Sysopt experience.:)

I appreciate the thoughtful reply. I usually get something along the lines of AMD rocks, just shudup!

The current trends you mention are where I think this might play out and prove you right. Considering Intel's approach, have they limited themselves into a dead end technological vector? What I mean to say is that will there be a point where it will be patently obvious to anyone that they are reving the Pentiums higher and higher with less and less benefit?

The reason that I ask is that the mhz difference between AMDs and Intels seem to be widening, but the AMDs seem to stay in the same position (just a touch behind the fastest Intels). A year ago T-birds were running a few hundred mhz behind comparible Pentiums. Now the Pentiums seem to be almost 1000 mhz ahead (in terms of clock speed) but the AMDs are still keeping up. Will the gap keep widening? Has Intel painted itself into a corner?

T

[wallie_x]

tKing,
I was playing devils advocate a little in that last post. I don't perceive Intel or its motives wholly in those terms. Perception is a funny thing. We can deeply influence our conclusions by adding cognitive filters. Unfortunately, most of these are subconscious filters, which usually prejudice our ideas somewhat unbeknownst to us. (I love philosophy; if I tried hard enough I might be able to convince myself the moon is not really made of green cheese...nah! I like green cheese.)
At another web page I saw the author state that the P4, in its current state, wouldn't be able to go much beyond 3ghz. (Some had also said the Athlon was at the end of its MHz rope too though.) I am certainly not Joe computer. These reluctant beasts are my part time hobby, but I believe the 64 bit CPU is about to change the game radically.
On the other point, I think executives think more in terms of ploys to win adherents than do engineers. The real problem is that the morality of capitalist America has sunk to new lows; and for those of us who care, exposing falsehood should be done rather loudly.

Thank you too for the thoughtful input.

[Rugor]

Intel has some very good engineers, and a very large budget.

The P4 we see today is very much stripped down from the original design, which did feature a much more powerful pipelined FPU. Unfortunately the die size was prohibitive, being more than 50% greater than Willamette on the same process.

In many ways the P4 really is a triumph of engineering over design. It was designed to scale to high clock-speeds as the first priority. Nothing else was as important. The secondary goal was to make sure the processor was optimized for what Intel's marketing department considered important, primarily websurfing and streaming media.

AMD designed the Athlon with something else in mind. They wanted to make the most powerful x86 core they could, and like Intel, they succeeded. The P!!! was only able to compete with the Athlon because it had better L2 Caching structures, SSE support, and more TLB's. It could get data to the core faster, so the Athlon's greater power was eaten by having less time to work. Remember, ever since the coppermine, Intel's L2 cache bus has been four times as wide as AMD's.

Both companies have very good engineers, Intel simply has more of them. Intel's resources are much greater than AMD's, or than any other semiconductor company. AMD, like most companies, transitioned to .13 micron on a single fab. Intel used 4 megafabs, letting each one try different methods. No other company can afford that kind of parallelization.

Intel's resources also allow them to put more effort into packaging than AMD can afford. They can devote the time and effort to packaging, and AMD can't afford to match them.

Intel is more likely to run into pure speed based physics related limitations due to the double speed ALUs. The current 2.8GHz P4's ALUs are already running at 5.6GHz. AMD's fastest part runs at 2.13GHz. AMD will probably not run into physics limitations as quickly.

Personally, I like AMD's design philosophy with the Athlon core better than Intel's with the P4. I have a huge problem with what happened with BAPCO, but that doesn't invalidate Intel's engineering prowess. I just think they aimed it in the wrong direction.

[gibsinep]

Nice Rugor, I also like the design of AMD chip better, like you said I like the though of having more power per punch so to say instead of having to pucnh twice(intels higher mhz scale)

lol I know bad example. :t

[wallie_x]

I've read most of both the articles at Van's and the one from AMD. They both put forth very convincing arguments about how pro-Intel rigged Sysmark 2002 realy is. It's somewhat disgusting actually to think Goliath needs to rig the fight to beat David. Very low Intel. Do you think we're all dummies? Not everyone buys Celeron's you know. Which actually means some of us do have the ability to know something beyond the MHz=performance myth you still promulgate. The P4 Celeron proves beyond a doubt you have few morals, or you wouldn't attempt to rip people off so blatantly. Poor fellow who's eyes gleam as he see's "1.7ghz" Celeron written across his new machine thinking he's bought the world. Perhaps your new nickname should be Infidel (a non-believer). You don't believe in yourself or your products; if you did then you wouldn't of had to rig Sysmark 2002 to make yourself look better. :eek: :x

[tking]

Originally posted by Rugor
Intel has some very good engineers, and a very large budget.

Thanks Rugor, that's just the kind of thing I was looking for. I've read the articles but I'm kind of new to this and I often get lost in politically biased reporting which ends up turning into bashing rather than analysis.

That was the clearest it's been defined for me.

.. and Wallie? Fear not, that same user with stars in his eyes over his new Celeron 1.7 will be the first to shake his head in disbelief when it won't run his new copy of Doom III the way he thought it would. If the Celeries and P4's are hobbled it'll show. I just hope AMD can stay with them till they run out of scalability, then it'll get real interesting.

Of course, if Intel can parallel test manufacturing they've probably got an RnD department the size of a small country. Who knows if the P5 will be evolutionary or revolutionary.

T

[gibsinep]

You dont see AMD trying to "trick" a product like the Celeron on the public. The Duron was a very good CPU also.

Dont tell me the Pr rating is a trick as it isnt it IS true.