AMD's little PR exercise : from the inquirer

[Spitfire_x86]

<i> Original Article </i>

<b> AMD's little PR exercise

Speculative out of prediction order

By INQUIRER staff: Tuesday 10 September 2002, 01:57


MAKE OF THIS WHAT you will. It just popped into our mailbox without any warning.
XP2300+ = 1.87GHz T-bred B 333FSB
XP2400+ = 2.00GHz T-bred B 266FSB
XP2500+ = 2.00GHz T-bred B 333FSB
XP2600+ = 2.13GHz T-bred B 266FSB
XP2700+ = 2.13GHz T-bred B 333FSB
XP2800+ = 2.27GHz T-bred B 266FSB
XP2800+ = 2.13GHz Barton 333FSB
XP2900+ = 2.27GHz T-bred B 333FSB
XP3000+ = 2.27GHz Barton 333FSB

AthlonXP Thoroughbred B / Barton Models
Secrets: Barton Cache means add 100+ to PR ratings. T-bred B 333FSB means add 100+ to PR from the original T-bred B.

All new PRs are based on the T-bred B (XP2400 = 2GHz). </b>


What do you think about this article?

 

[baldurga]

Well, IF this is true, now we can guess the benefit of 333 FSB and 512 L2. Anyway, I won't believe anything until I can see some benchmarks. PR rating is good as long as benchmarks tells you the same.

Also, how they verify that a CPU that runs ok at 266FSB can't run at the same frecuency but at 333FSB? I would feel quite confident in buying the 266 model, unlock it, and raise the FSB to 333.

DIY: read, buy, test, learn, reward yourself!

 

[MeTaLrOcKeR]

If thats true....than the Barton has no chance against the P4 NW....especialy C1 NW's.........

thats a crock of sh|t.......and if it is true.......than WTF is AMD doing ?

<A HREF="http://www.anandtech.com/mysystemrig.html?id=13597" target="_new">-MeTaL RoCkEr</A>

 

[eden]

Actually the 200 more PR points for new FSB and cache make sense since the AthlonXPs won't benefit majorly.
But the exaggerated XP3000 at 2.27GHZ will truly not make sense IMO. You are about 733MHZ away competing a 3.06GHZ HT enabled P4, that to me sounds like AMD has got to change PR to linear scaling, by 100MHZ and 100PR points.

--
When buying an AthlonXP, please make sure the bus is at 133MHZ, or you will get a lower speed!

 

[MeTaLrOcKeR]

they cant do that....otherwise theyd never be able to compete with Intel...in that case...why ot just drop the whoel PR ratings.........and take Apple's approach....

<A HREF="http://www.anandtech.com/mysystemrig.html?id=13597" target="_new">-MeTaL RoCkEr</A>

 

[Dark_Archonis]

Unless AMD is trying to pull a miracle out of their hats, that's bogus. XP3000+ at 2.27ghz? Ya right, and christmas is in the summertime...

- - -
<font color=green>All good things must come to an end … so they can be replaced by better things! </font color=green>
<P ID="edit"><FONT SIZE=-1><EM>Edited by Dark_Archonis on 09/10/02 05:56 PM.</EM></FONT></P>

 

[chuck232]

Plus looking at the tests at THG, it showed the 166MHz FSB XP3400+ performing near a 2.66GHz P4. Now that says a lot for AMD's PR huh? At 3400+, that's 2.66GHz. Now I really don't think the extra cache is gonna make up for <i>that</i> much.

...And all the King's horses and all the King's men couldn't put my computer back together again...

 

[stealthbig]

Chuck I think the real 3400+ is going to have larger L2 cache than the overclocked one.

 

[chuck232]

LOL.... predictable. Someone over at SharkyForums brought up the same thing as you!!

If you say that the XP3400+ will have more cache, that'll be extremely confusing. The PR is supposed to give consumers an idea of what kind of performance they can expect. Now if a t-bred XP3400+ and a Barton XP3400+ aren't the same speed, then that'll be very bad. More than likely the Barton 3400+ will probably be a tad slower in clock speed terms but have more cache.

Here's their posts:

<i>macutty says:</i>
Look at the difference between a XP3400+ and a P4 3.6GHz, can you say "byebye AMD"? they better have one hell of a chip with barton, cause with the NW core going this high I would expect prescott to at least add 10-15% performance increase over the NW at launch.

<i>Mithrandir3017 says:</i>

Remember, the "3400+" has a 266 FSB and 256k L2 cache. A Barton 3400+ would be faster.

<i>Moridin says:</i>
If this is indeed the case, doesn’t it indicate how useless AMD’s model number system really is? The model number is supposed to help the “uninformed” user understand how the processor really performs. If the same model number results in two different levels of performance how can this possibly be happening?

...And all the King's horses and all the King's men couldn't put my computer back together again...<P ID="edit"><FONT SIZE=-1><EM>Edited by Chuck232 on 09/10/02 08:50 PM.</EM></FONT></P>

 

[eden]

Chuck, I thought you were there before, but yes we actually had quite a long discussion about why each higher speeds, the AXP loses IPC.

Some of us theorized, the cache design is becoming poor. 256KB L2isn't high on such speeds, but the most is inside the cache, how it works, the associativity, the bit width, all that contributes to weakened core performance.
There might be core design problems as well but I am sure AMD is aware if they ever are to release PR3400, it cannot be this way for sure. No way the AthlonXP will become a clock per clock comparison against P4s in non-SSE2 apps, unless P4 upgrades its IPC or AMD starts core stripping.

Also, the PR will continue to change, AMD is the owner of it, by their CPU's, and as they just did with XP2600, they can admit there are bogus errors at high speeds vs PR, so they will probably revise it again. I am always gonna say this, until they stop that illogical 66=100 technique, they cannot expect the AXPs to continue to compete when PR is 1000 points higher than actual MHZ, it would mean the PR is losing IPC as it goes up, which is both non-existant and not logical.

--
When buying an AthlonXP, please make sure the bus is at 133MHZ, or you will get a lower speed!

 

[chuck232]

Hey yeah I'm sure the real XP3400+ will be very much different. AMD'll probably have to change their ratings again, maybe Hammer will be able to fix that without even a rating change. I was merely looking at the tests that THG did. And you've got to admit, looking at those tests, the *XP3400* performs awefully close to a P4 2.66GHz or a 2.8GHz. I'm sure AMD won't get caught doing that though.

Just playing devil's advocate again, Eden .

...And all the King's horses and all the King's men couldn't put my computer back together again...

 

[a2112a]

Read somthing about this elsewhere

XP2800+ = 2.13GHz Barton 333FSB
So fsb = 166
Mult = 12.8 ????
or
XP2300+ = 1.87GHz T-bred B 333FSB
166 x 11.2 ????
or
XP2900+ = 2.27GHz T-bred B 333FSB
166 x 13.6 ????

<P ID="edit"><FONT SIZE=-1><EM>Edited by a2112a on 09/11/02 01:54 AM.</EM></FONT></P>

 

[chuck232]

Hmmm... you bring up actually quite a good point. I've not seen any multipliers other than xx.0/5. If they're that obscure, it probably means that the clock speeds aren't right, but they may be something close.

...And all the King's horses and all the King's men couldn't put my computer back together again...

 

[a2112a]

Just edited, forgot the .6666666
Still looks strange
1/5 multi steps?

 

[chuck232]

The XP2800 166MHz will probably 13 mult, so ~ 2.16GHz
The XP2300 will probably be 11, so ~1.83GHz
Then the XP2900 will probably be 13.5 mult and ~2.24GHz.

That's my prediction.

...And all the King's horses and all the King's men couldn't put my computer back together again...

 

[juin]

I think that is a good PR-rating it will be acurate

At the end i have speak with a horny lady

 

[nja469]

I want a real 3GHz AMD processor, when is that gonna happen?

 

[shallowbaby]

why ot just drop the whoel PR ratings.........and take Apple's approach....

yes!

<font color=green> there's more to life than increasing its speed -Ghandi</font color=green>

 

[eden]

Which is?

--
Where did your THGC username come from and why did you choose it? <A HREF="http://forumz.tomshardware.com/community/modules.php?name=Forums&file=viewtopic&p=19957#19957" target="_new">Tell here!</A>

 

[shallowbaby]

i just hope that the PR supporters will still be here when amd decides to abandon it completely.

i'm sure there will be those will still defend the PR afterwards even.

<font color=green> there's more to life than increasing its speed -Ghandi</font color=green>

 

[texas_techie]

This from anandtech:
"ExtremeTech has a story on Intel's Banias claiming that it might use a performance rating similar to what AMD does with its XP processors. Banias is going to run at slower clock rates than P4 so, the Intel marketing machine has to come up with some other way of hyping the higher end of the line. Of course, Intel won't be doing the same thing as AMD, but this might be fun to watch"

I think it would be ironic for Intel to adopt a similar PR scheme as AMD.

----------------------------------------------------
AFAIK - that PR model/rating scheme in the original post is dead on accurate (read: thats how the numbers/performance will be upon release).
Second - TSMC has announced it will reduce the amount of 300mm wafers its making. From an estimated 10,000 month to 5,000 month. I have always speculated TSMC is having problems at the .13 micron level. This cant be good for AMD.

Third - there is an unconfirmed report that the Hammer will be delayed even further - ie.. not until march.

Fourth - there are two rumors floating around. Take them with a grain of salt. First - is that AMD is considering selling off more of its assets in an attempt to stay afloat.
Second, that If Hammer doesnt do well - then AMD will very likely fold within a couple of years of Hammer's release.
My personal opinion is that both are true. THe incredible pace Intel is keeping coupled with Hammer's delays = big, big trouble.


Benchmarks are like sex, everybody loves doing it, everybody thinks they are good at it.

 

[a2112a]

Re: texas_techie

It's sad but I think your right.

 

[eden]

Coming from a guy that has contacts at AMD, this looks too true, and too grim.

Man these are not good times, however I blame it on AMD's slow pacing. They hyped too much, they made a paper launch in the middle of Intel's rampage, they continue not to do some improvements WE demand. Hammer HAS to be succesful and powerful or otherwise point 4 is the true point.

--
Where did your THGC username come from and why did you choose it? <A HREF="http://forumz.tomshardware.com/community/modules.php?name=Forums&file=viewtopic&p=19957#19957" target="_new">Tell here!</A><P ID="edit"><FONT SIZE=-1><EM>Edited by Eden on 09/11/02 10:09 PM.</EM></FONT></P>

 

[chuck232]

That will seriously bad for competition. What'll happen? Will the gov try to pump cash into the company or just let it die?

I kinda think that AMD brought this on themselves. As Eden said, they hyped their products so much then when we ask, when can we get it? They say... oh I dunno... a couple years at best. :frown:

...And all the King's horses and all the King's men couldn't put my computer back together again...

 

[why]

Tell me you don't really think that AMD would fold based only on the success or failure of one line of its products. AMD makes a lot more than CPU's and owns an amazing number of patents. Success or failure of a company is not so simplistic.