Sign in with
Sign up | Sign in
Your question

A64 3000+ 90nm Review

Last response: in CPUs
Share
September 28, 2004 2:48:18 PM

<A HREF="http://www.amdzone.com/modules.php?op=modload&name=Sect..." target="_new">Doesn't quite</A> look as if AMD truly managed the extra 5% performance boost per clock, but quite good processors nonetheless.

More extensive testing should come out from other sites, and it will then be easier to judge.

More about : a64 3000 90nm review

September 28, 2004 3:14:26 PM

Oh, and you might have some trouble getting a connection to the site. I had a few "[-peep-] happens" screens before I could read the whole article, but just keep trying and it will work eventually.

<i>Edit: I read the whole article more carefully. It seems a 3000 A64 @ 90nm is slightly slower than the 130nm counterpart indeed. Not only not faster, slightly slower. Not as "OMG" as prescott, but still not better than old tech.

Also, the 1.8Ghz 3000+ did not manage past 2Ghz or so, and showed poor overclockability. It might be another indication that only future generation processes (extra layers) will scale adequately.

In any case, it doesn't look as if 90nm is as rosy as 130nm was back in the northwood days. AMD seems to have done the 90nm without much commendability either.

<P ID="edit"><FONT SIZE=-1><EM>Edited by Mephistopheles on 09/28/04 02:35 PM.</EM></FONT></P>
September 28, 2004 4:23:42 PM

OTOH, clock for clock it is a little faster than 130. AMD needs to biatch slap their marketing department hardcore on this naming convention. This was an extremely stupid mistake. They should have released it with the same name as the 2800+, then the review would be read, the speed increase seen, and no negatives. Now they will have to deal with a negative public perception.

Leave it to marketing to mess things up.

I have to admit though, when you compare the 90nm 3000+ to it's clock equivalent 130(2800+), it's only marginally faster. That's kind of disapointing, but at least it is a little faster.

<i>Cigarettes - No cholesterol, high in fiber, low in fat, how could they not be good for you?</i>
Related resources
Can't find your answer ? Ask !
September 28, 2004 6:06:03 PM

AMD doesn't have the 5% boost per clock because revision E hasn't been released. With Revision E we get SSE3 and such. Besides being on .09u process and dual channel memory controller the A64 3000+ Winchester is the same as a 2800+ Newcastle. I can't see how AMD can justify 200 PR Points by going from a single to dual channel memory controller. This is not good news at all if a 939 3000+ (1.8GHz) is slower than 754 3000+ (2GHz).

Abit AN7 (Nvidia Nforce2 Ultra 400), AMD Athlon XP-M 2600+ @ 2.43GHz (211 X 11.5), Nvidia Geforce4 Ti4200 128MB OCZ DDR EB Platinum 2 X 256MB PC3700 @ 211MHz 1:1 2-3-2-7
3dMark2001SE: 12,997
September 28, 2004 6:46:50 PM

Rev Es also run cooler according to the likeness of OPPAINTER and Macci (who have had them for a surprisingly long time).

Wanna know why it's called 3000+? Cause 2900+ isn't a cool name and the performance certainly is better than 2800+! Also the way I see it...wasn't the S754 naming scheme initially for PRing Northwood P4s????? S939 is *NO DOUBT* comparing itself to Scotty...so the 3000+ is at least somewhat justified (especially if you factor in DDR2 :tongue: ). I really see no problem with their naming scheme if you keep in mind that they're trying to compare performance to Scotty and (somewhat) not against itself (or older versions of itself or even Northwood anymore).

Maxtor disgraces the six letters that make Matrox.
September 28, 2004 6:58:36 PM

My mistake then, I thought Rev E was the yet to be released core revision with SSE3 support...

I do see a problem with their naming scheme. AMD doesn't say A64 Northwood 3000+ OR A64 Prescott 3000+, they say 3000+! In my opinion if one 3000+ performs under another 3000+ then theres something wrong. Yes, 2900+ might not sound cool but it's more accurate than 3000+.

Kinda like AXP 3200+ may sound cool but its nowhere near P4 3.2C, more like 2.8C. Now if Intel released a P4 3.2B then the 3200+ rating might be more accurate. Bah, that's why I hate these model numbers...but hey we don't live in a perfect world so they have to be there I guess.

Abit AN7 (Nvidia Nforce2 Ultra 400), AMD Athlon XP-M 2600+ @ 2.43GHz (211 X 11.5), Nvidia Geforce4 Ti4200 128MB OCZ DDR EB Platinum 2 X 256MB PC3700 @ 211MHz 1:1 2-3-2-7
3dMark2001SE: 12,997
September 28, 2004 7:05:41 PM

Rev E will have the (few) elements of SSE3 that are non-HT related...all I said was that Rev E will <b>also</b> be cooler.

Here's a guide for you if you have such a problem with a naming scheme:

AXP PR ~= P4B
A64 S754 PR ~= P4C
A64 S939 PR ~= P4E

If, however, P4E's weren't slower than P4C's, no doubt they would have kept with the old naming scheme. However, by comparing to Scotty, their marketing group can get a precious 100-200 extra PR points, so they take full advantage of that.

Maxtor disgraces the six letters that make Matrox.
September 28, 2004 7:08:01 PM

It should have been named 2900+
This would tell all the geeks, all those who would bother specing out a system, that its an oddball advantage mhz for mhz. This would then make the other 3100, which also would be truthful.
I hate this whole xp3000 thing.
theres no way an xp3000 goes as fast as a p4 3.0 ghz, at least in the HT models (which 90%+ of the ones bought at this speed are currently).
If only amd would just say 2800 and 3000 when it makes sense to.
September 28, 2004 7:24:32 PM

Well the thing is if you read THG's "AMD Travels through Time: AMD Athlon XP 2800+" you can see that the 2800+ T-Bred (not available anymore and only released in limited quantities) was equal to if not better than a 2.8B. So they had everything correctly up till they got whacked by the barton core and P4C.

AMD Athlon XP-M 2600+ @ 2425MHz (220.5 X 11.0), OCZ Enhanced Bandwidth PC3700 2X 256MB 220.5MHz 1:1 @ 2.5-2-2-8 Dual Channel, ATi Radeon 9600 Pro @ 468MHz/346.5MHz

3dMark2001SE: 14,076
September 28, 2004 7:43:10 PM

Regarding the naming thing, that's also why people became confused with the Sempron naming convention. It's matched up to the Celeron D, not to the AXP or A64 naming convention.

I'm just your average habitual smiler =D
September 28, 2004 7:48:02 PM

Exactly...it's all about what product it is designed to comepete against (now these aren't always right...the 3400+ does have a struggle to keep it's name against the 3.4C). The 3.0E (and DDR2) is a pretty poor performer and a dual-channel, 512KB, 1.8GHz A64 DOES keep up with it pretty easily--there is nothing wrong with the naming convention.

Maxtor disgraces the six letters that make Matrox.
September 28, 2004 7:53:28 PM

AXP 3000+ competes quite nicely with my P4 3.06 w/ HT. Just because AMD came out with new processors after Northwood C came out using the same PR scheme does not mean that it is rated for those new MHz...the target competition remained the same while Intel released a superior chip that was generally unrivaled until the A64 S754 rolled along.

Maxtor disgraces the six letters that make Matrox.
September 28, 2004 7:57:18 PM

<i>Cause 2900+ isn't a cool name and the performance certainly is better than 2800+!</i>

Marketing gone wrong. 2900+ is a fine name, especially if it accurately reflects the performance placement within THEIR OWN lineup. AMD should focus on offering a line up of chips with ratings relative to each other. They no longer need to try to match Intel tit for tat. They have the performance advantage now, they should start acting like it.

Try to look at it from Joe users perspective. Joe user doesn't want to hear from his tech geek friends things like "Oh, the 3000+ is a good chip just make sure you don't get the 90nm." Joe user hears that and goes "WTF are you talking about?" Then his geek friend attempts to explain how it all makes sense "Well, the 90nm chip is in direct competition with Intels Scotty which is also 90nm, so the 3000+ rating is in reference to that. If you get the 130nm chip rated at 3000+, it's actually faster."

Now joe doesn't have any idea what "nm" really means,nor does he have any idea why there should be a difference between chips with the same rating. Joe also doesn't want to hear why 2 chips are rated the same but also have different clock speeds, yet other than die size, have no discernable technological difference. You know why he doesn't want to hear it, because joe has no fricking idea what the hell half of the words in that sentence mean. If you try to explain it to joe, he will start drinking heavily while wishing he didn't have you as a friend. Poor Joe already has a hard time understanding cache size, 64bits, TLA-RAM differences, etc...

I'm of course engaging in a bit of hyperbole, but really, AMDs ratings should be straightforward. A 3000+ should perform the same as a 3000+. You can't go wrong with keeping it simple.

<i>Cigarettes - No cholesterol, high in fiber, low in fat, how could they not be good for you?</i>
September 28, 2004 8:05:14 PM

Quote:
A 3000+ should perform the same as a 3000+. You can't go wrong with keeping it simple.

We agree to disagree...I stand by my ratings relative to Intel (their competition, anyway) and you stand by their consistent performance ratings throughout their entire line.

Maxtor disgraces the six letters that make Matrox.
September 28, 2004 8:18:18 PM

I know, read this though.

http://www.theinquirer.net/?article=18733

"There's worse to come. The current top of the Sempron range is the 3100+. But it is slower than an Athlon 64 2800+. Yes, 3100 is somehow slower than 2800. Where the marketing department spent years and millions of dollars convincing a sceptical press that the Performance Rating system was reasonable and logical, it has just blown all of the slowly built credibility completely out of the water."

That guy actually thinks 3100+ is up against P4 speed....

AMD Athlon XP-M 2600+ @ 2425MHz (220.5 X 11.0), OCZ Enhanced Bandwidth PC3700 2X 256MB 220.5MHz 1:1 @ 2.5-2-2-8 Dual Channel, ATi Radeon 9600 Pro @ 468MHz/346.5MHz

3dMark2001SE: 14,076
September 28, 2004 8:40:57 PM

That's exactly the problem. It doesn't matter how logical the rating system is if you happen to understand the logic behind it. If your system generates bad press and misunderstandings in the market, it's not good.

Logically speaking, within the context of having the technical knowledge to understand why the ratings are what they are, there is nothing wrong with the ratings, but human beings are not logical and it's human beings they are marketing these chips to. This is bad marketing. Very very very bad marketing.

<i>Cigarettes - No cholesterol, high in fiber, low in fat, how could they not be good for you?</i>
September 29, 2004 12:42:38 AM

Well, the S939 3000+ 90nm seems slower than the 3000+ on 130nm and costs more. So, while in technology land the 1.8Ghz 3000+ on 90nm might be slightly faster per clock than the 2.0Ghz, this advantage has been literally thrown away by marketing, pricing and higher clock of the 130nm version!!! So you'd still recommend old 130nm tech (2Ghz 130nm) rather than newer technologies.

Which is stupid. There is still <i>no reason</i> to buy a 90nm chip at all so far, except maybe for Dothan. And the pricing means that AMD either doesn't want to or can not flood the market with 90nm products. Because it's quite early for their 90nm process, I'm thinking it's the second alernative. Spreading A64s now in mass quantities and with the s939 platform would only do them good. Unless, of course, they want to keep S754 infrastructure around a little longer...
September 29, 2004 12:56:24 AM

<i> So, while in technology land the 1.8Ghz 3000+ on 90nm might be slightly faster per clock than the 2.0Ghz, this advantage has been literally thrown away by marketing, pricing and higher clock of the 130nm version!!! So you'd still recommend old 130nm tech (2Ghz 130nm) rather than newer technologies.</i>

100 percent in agreement.

<i>Cigarettes - No cholesterol, high in fiber, low in fat, how could they not be good for you?</i>
September 29, 2004 1:08:48 AM

Quote:
That guy actually thinks 3100+ is up against P4 speed....

:eek:  <i>What?</i> From were did you get that? I think you're mistaken. This guy simply thinks that the performance rating should be coherent amongst AMD products, which it isn't! Obviously! He didn't even mention Intel! Your own quote shows that!

Oh, and BTW, that inq article was quite an interesting read.
a b à CPUs
September 29, 2004 1:57:43 AM

Hey, if it's 2% faster, and the Prescott is 3% slower than the Northwood, AMD claims it's fair. Also, AMD has their own set of benchmarks to proove the name is fair, and is willing to use crippled platforms for lesser processors in order to PROVE their point.

<font color=blue>Only a place as big as the internet could be home to a hero as big as Crashman!</font color=blue>
<font color=red>Only a place as big as the internet could be home to an ego as large as Crashman's!</font color=red>
September 29, 2004 2:15:49 AM

My opinion on this isn't about whether it's technically fair or not. I see this as adding confusion to an already confusing(for the uninitiated) marketplace.

Can't you see the clarity and advantage of AMD offering processor lines whose rating numbers clearly identified their performance level with regards to other chips in their lineup? Just my opinion.

<i>Cigarettes - No cholesterol, high in fiber, low in fat, how could they not be good for you?</i>
September 29, 2004 2:26:31 AM

"The current top of the Sempron range is the 3100+. But it is slower than an Athlon 64 2800+. Yes, 3100 is somehow slower than 2800."

Somehow? Maybe because it's not competing against a Pentium 4! That's JUST like saying this:

"The current top of the Celeron range is the 3.1GHz. But it is slower than a Pentium 4 2.8GHz. Yes, 3.1GHz is somehow slower than 2.8GHz."

The way he words that makes me come to the conclusion that either

A) He's referring to what the average consumer would perceive
B) He's a freakin moron.

This make sense now?

AMD Athlon XP-M 2600+ @ 2425MHz (220.5 X 11.0), OCZ Enhanced Bandwidth PC3700 2X 256MB 220.5MHz 1:1 @ 2.5-2-2-8 Dual Channel, ATi Radeon 9600 Pro @ 468MHz/346.5MHz
3dMark2001SE: 14,076
September 29, 2004 2:27:02 AM

Personnaly, before I spend a grand on toys, I want to know what I'm getting. If people are befuddled by names, they need help. Could be part of what some places offer, when they sell systems.
September 29, 2004 2:38:13 AM

Well put yourself in this situation, I used to sell PCs at Circuit City. Do you know that I basically lie to a customer when I tell them, "Yes this Athlon XP 3000+ is equivalent to this Pentium 4 3.0GHz right here." I have to bite my tongue man. And it's especially hard to tell the consumer that this A64 3000+ is a better performer than this AXP 3000+! If there the same number the consumer should believe that the only advantage is 64 bitness right? When in fact it is so far from the truth...now that's a lie to consumers, no matter how much I love AMD, their technology, and their processors. I do not like their marketers. They should just put the clock speed of their processors and market how much better clock for clock they are than Pentium 4s. That's why I hate the clockspeed of the P4, I really believe it was made just for marketing sometimes...

AMD Athlon XP-M 2600+ @ 2425MHz (220.5 X 11.0), OCZ Enhanced Bandwidth PC3700 2X 256MB 220.5MHz 1:1 @ 2.5-2-2-8 Dual Channel, ATi Radeon 9600 Pro @ 468MHz/346.5MHz
3dMark2001SE: 14,076
a b à CPUs
September 29, 2004 3:50:38 AM

It doesn't matter if they screw up the ratings as far as the average consumer is concerned, the number is bigger so they'll buy it, and that's it. Intel sells millions of Celerons that way. So it would be confusing to the uninitiated if they cared, but appearently they don't care. Let them spend their money.

<font color=blue>Only a place as big as the internet could be home to a hero as big as Crashman!</font color=blue>
<font color=red>Only a place as big as the internet could be home to an ego as large as Crashman's!</font color=red>
September 29, 2004 4:29:23 AM

""Doesn't quite look as if AMD truly managed the extra 5% performance boost per clock, but quite good processors nonetheless."""

It does not support sse3 which was part of that 5% gain claim. I told you a couple of weeks ago that these early 90nm did not apear to support sse3 so no suprize if its not quite 5% on a clock per clock = cache comparison.

That being said their is no excuse for a 90nm 3000+ preforming much slower then a 130nm 3000+. AMD's ratings should not change for same generation cpu. the + is there for a little wiggle room when running the amd benchies so they should be very close. I belive the prescott did not exist when the a64 was launched so the rating formula was most likley designed to compare to the p4C which is why it beats the prescott so badly.

One thing that was cool about amd's ratings is it let you know that a higher rating always meant better performance (OVERALL) when dealing with same generation cpu's. (at least when using there comprehensive overall benchmark formula) this is much better then anything intel offers. One can always just compare to a doom3 bench and say the formula is screwed cause doom3 which likes cache preforms better on a64 3400+ over a64 3500+ which has less cache but higher clock. Overall the 3500+ is faster than the 3400+.

If I glanced at a spilt box of tooth picks on the floor, could I tell you how many are in the pile. Not a chance, But then again I don't have to buy my underware at Kmart.
September 29, 2004 6:01:58 AM

I just noticed something that really got my attention here..

Athlon 64 2800+ 1.8GHz/512K/Single Channel
+200 PR
Athlon 64 3000+ 1.8GHz/512K/Dual Channel

Athlon 64 3400+ 2.4GHz/512K/Single Channel
+400 PR
Athlon 64 3800+ 2.4GHz/512K/Dual Channel

How does going from Single to Dual Channel at 1.8GHz give AMD only 200 points but at 2.4GHz, it gives it double the points? explain that to me? Does the memory controller scale with clock speed? That's the only reasoning I can think of...

AMD Athlon XP-M 2600+ @ 2425MHz (220.5 X 11.0), OCZ Enhanced Bandwidth PC3700 2X 256MB 220.5MHz 1:1 @ 2.5-2-2-8 Dual Channel, ATi Radeon 9600 Pro @ 468MHz/346.5MHz
3dMark2001SE: 14,076
September 29, 2004 3:26:50 PM

It's called, Intel's highest chip speed is 3.6, so AMD needs to beat that so they make up a number (they certainly don't want the 2.4Ghz/512k/Dual Channel to have the name 3600...that would make it look like it competes with the 3.6 instead of stomping on it).

The PR system was a good idea in concept, but I don't think many people like the way it's executing. I think AMD figured people were being blinded by Intel's Ghz difference marketing so they needed to try to use similar blinding skills to become marketing competative. But they may have overdone it. :) 

I'm just your average habitual smiler =D
September 29, 2004 4:35:58 PM

oh no no it wouldn't compete with the 3.6, it has the + at the end so it's better! :)  I really think some of the latter XP ratings should have a - at the end of them because it's really true. It should be like this:

XP Rating vs P4 Willamette: ++ notation, ex: 3000++
XP Rating vs P4 Northwood A: + notation, ex: 3000+
XP Rating vs P4 Northwood B: no notation, ex: 3000
XP Rating vs P4 Northwood C: - notation, ex: 3000-

AMD started out conservatively with their product line and then moved more and more liberally as time wore on. I think they wanted to see how far they could stretch the truth before they would get in trouble for it. I also tend to believe that not only the bus speed, but Hyperthreading, gradual adoption of SSE2 in applications (XP's only have SSE) AND Dual Channel DDR all had big influences on boosting P4 clock for clock. I'm really afraid that AMD is going to repeat their mistakes with the 64. At first the rating is conservative and it slowly and slowly becomes out of whack again. WHich is why I HOPE TO GOD that Intel makes the Pentium-M their desktop processor in the future because that would make the "GHz" war end and AMD wouldn't have to create a performance rating any more.

And the more hilarious thing of it all would be AMD would have the higher clocks again since the K7/K8 architecture can clock higher than the P6 architecture, lmao.

AMD Athlon XP-M 2600+ @ 2425MHz (220.5 X 11.0), OCZ Enhanced Bandwidth PC3700 2X 256MB 220.5MHz 1:1 @ 2.5-2-2-8 Dual Channel, ATi Radeon 9600 Pro @ 468MHz/346.5MHz
3dMark2001SE: 14,076
September 29, 2004 5:44:35 PM

Where are the government watchdogs to be found. Let Intel do this exaggerating, and an investigation into deceptive practices would be launched before you could say Justice Department.



========================
Try everything...
Do not be afraid of failure, for this is how we learn and grow...
Live life to the fullest...
Do not regret what you have not yet done!!!
September 29, 2004 6:41:12 PM

I don't like the confusing marketing practices as much as the next guy, but I don't think this is a job for the government. The benchmarks for speed are too subjective, just as Crashman pointed out. Unless you're referring to something blatantly dishonest on Intel's part that I'm not aware of?

Too bad we don't have a "quarter mile" test for processors. Imagine if we didn't have one for cars. The Japanese companies would brag how their engines rev higher and imply that somehow meant faster. The Americans would brag that their engines have more cylinders, and that meant faster(yes, I know this can *tend* to be true, but it isn't always). The Germans could brag about their advanced engineering and imply that meant faster.

That's kind of what we have happening here. Quarter miles don't lie, but benchmarks can be cherry picked to show a processor in its best possible light. What we need is to have more people in the tech press write harsh articles condemning the practice. Corporations hate bad press like cats hate dogs.

<i>Cigarettes - No cholesterol, high in fiber, low in fat, how could they not be good for you?</i>
September 29, 2004 8:33:20 PM

Everytime someone says more cylindars = faster, I think of a Honda S2000 vs a Camry V6. Which do YOU think is faster? :) 

A standardized set of benchmarks would be good, but that doesn't exist in the car world either. There's top speed, 0-60 mph, 0-100 mph, skidpad, 1/4 mile time, mph recorded at the 1/4 mile mark, gear ratios, etc., etc. And that's just the "performance benchmarks". Even with all that data, you can't tell which you'd rather buy, because you need to see if the car feels good (example, I had a new Nissan Maxima for a couple months and hated it, so bought a new Camry instead...I realize these aren't performance cars, but it still proves my point).

I'm just your average habitual smiler =D
September 30, 2004 12:26:42 AM

Exactly! Intel is always faster!

Abit IS7 - 3.0C @ 3.6ghz - Mushkin PC4000 (2 X 512) - Sapphire 9800Pro - TT 420 watt Pure Power
Samsung 120gb ATA-100 - Maxtor 40gb ATA - 100
Sony DRU-510A - Yellowtail Merlot
September 30, 2004 1:58:10 AM

i think this whole situation is pretty crazy. but its bad on both sides, intel and amd. they got forced itnot his corner becuase of the mhz race, where now thats in limbo and intel wants you to go by numbers instead fo mhz. They even say the model numbers do not reflect performance, so how is anyone suppose to knwo whats better? thats more confusing then ever for the average buyer.

i think the amd track can be just as confusing for the average user, but it seems they have decided to make thier pr numbers reflect performance against its competeing product. for example, the semprons are labeld as competeing against celeron becuase they are the duron replacement. so that 3100+ is against a 3.1 ghz celeron. s754 is suppose to slip down to that level, competeing against the celeron, so amd was in a touch position there, already having released chips that were more advanced then the duron replacement they had in store. by next year, s939 will be the only mainstream choice and the pr numbers will reflect that. i think the big problem was this first transition time, so many product lines being introduced where performance varied so much, overlapping was inevitable at first. im sure amd would rahter have just one line of pr numbers, and they will eventually, they just have to phase out alot of s754 stuff as s939 becomes cheaper and released in greater volume, which is what 90nm helps.

i know its confusing, but i just think it was a bit unavoidable with the way they started s754 out. thier will be growing pains for a time till s939 really takes over. as far as 90nm goes, it loks like its taken off fine, no problems that i see. the whole temp issue im nto sure aobut, ive seen some saying the temps are the same, some say they are a bit higher, so ill wait to see more reports. i ugess this first batch was to just hlep amd bring volume numbers of s939 out, then rev. E will bring the real advancements everyone was looking for.
!