AthlonXP Released

ldopa1 writes "True to form, AMD has released the new Athlon XP today. This article on Tom's Hardware has the full technical specs for the chip as well as a look at the new packaging. Tom's also has a full set of benchmarks for the chip." michael : See also reviews on LinuxHardware.org, Newsforge, AnandTech and AMDMB. Update: 10/09 20:29 GMT by T : gregfortune points out that AMD is giving away quite a few of these in a six-city promotion as well, so if you live in one of the six, perhaps you can snag one.

Comments

Question

[Anonymous Coward]

Do we have to call into AMD to get a number to have the chip activated?

Re:Question

[haruharaharu]

Try running in with a heatsink; your results might improve.

will the trickery work?

[kvandivo]

The question, though, is will their tricky ways work? Cyrix tried the exact same thing a few years back (marking a chip with a model number that represented the 'approximate intel performance' rather than the chip speed itself. Is AMD going to be able to pull this one off, or is it destined to go the same way as the Cyrix chip?

Re:will the trickery work?

[Sir_Real]

Well... The Cyrix chips didn't do well, but it wasn't because of their shady marketing, it was because the Cyrix chips absolutely sucked... (This is MHO, after owning one... I hated the damned thing...)

Donning Asbestos Jumpsuit...

Re:will the trickery work?

[tshak]

This isn't trickery, it's consumer educating. Back in the days of Cyrix, MHZ ment a lot more, and Cyrix was a lot less realistic than AMD is being. AMD could safely call their 1.53Ghz chip the 2000+ instead of the 1800+. With a combination of a conservative "ratin" and a very well performing chip, I think AMD will be successful with their new offering.

Re:will the trickery work?

[Brian Stretch]

This is AMD marketing trickery countering Intel engineering trickery. Intel doubled the pipeline length in the P4 versus the P3 and Athlon (20 stage versus 10 and 11, respectively) in order to crank the clock speed at the expense of performence (see The Megahertz Myth for an excellent overview on the subject). Thus why a 1.13GHz P3 will outrun a 1.4GHz P4 most of the time, and why the upgraded core of the 1.53GHz Athlon XP outruns the 2GHz P4 most of the time (for half the price).

In other words, in marketing, two wrongs make a right!

Re:will the trickery work?

[Telek]

Silly silly question time:

You have 2 competetors, really. The underdawg is going to try to make a new rating system that they hope to get everyone to use to bring back "fairness" to the processor rating game, but when your only real competetor is the one who is going to loose bigtime by using this new system (and therefore they will not), then how do you expect this new rating system to work?

I think what would be better is for now to drop this silly "we're as good as that P4, swear it!" scheme and just call it something else. Call it the AthlonXP G+ or 6P+ or something. That way there is no realative comparison, and when you go to the store to ask "hey, which is better here?" the guy can just tell you "Go with the AMD such and such, because it's as good as this P4 over here", not only does it stay up to date with comparisons, but you also don't get accused of trickery.

[H]ard|OCP Review

[questionlp]

The good ol' guys at [H]ard|OCP have a review of the Athlon XP as well. It can be found here.

Model Numbers

[Derkec]

Despite the fact that there is a new core which yields 3-7% more performance per clock, Tom's points out the Model Number scheme is the most interesting thing. AMD is now not selling thier processors as 1500MHz, but instead as 'equivilant to a P4 at 1800MHz' -> an AthlonXP 1800+. Is this a fair thing to do? It seems to me that it is trying to trick customers into evaluating the processors more fairly. While most slashdotters know MHz != speed, the average joe does not. I am comforted that the AthlonXP 1800+ is able to run with the P4 2GHz. AMD doesn't seem to have overhyped their processors at all.

The next topic for discussion: AMD is trying to bring together a third party instituation to rate processor speeds in some fair way. I'm sure Apple would be thrilled to jump on this bandwagon and our dear friends at Microsoft already have their hands in it.

Re:Model Numbers

[ldopa1]

Is this a fair thing to do? It seems to me that it is trying to trick customers into evaluating the processors more fairly.

Sure it's fair, I just think it's a bad idea. This ties AMD's entire branding effort to whatever Intel does. In other words, if Intel were to take the same tack, in five years we could end up with the "Intel Plentidum XXP++ 1800+++ Equiv4" and all of a sudden, the Cyrus Logic 2.5 GHz chip will look great despite the fact that the latest P7 (or whatever) is running in the 15 GHz range.

If we were to do this with political candidates, our ballot would look like:

George W. Bush (Equivalent to A. Gore)

Al Gore (Equivalent to G. W. Bush)

John McCain (G.W. Bush+)

Ralph Nader (iMac)

Re:Model Numbers

[DivineOb]

The "proper" way to compare processors would be to use the SPEC benchmarks (http://spec.org) (since they are supposed to "define" processor performance). However, AMD won't use these benchmarks use them. If we look at the highest published scores from AMD we see

Integer: 495 base, 554 peak
FP: 426 base, 458 peak

For P4

Integer: 640 base, 656 peak
FP: 704 base, 714 peak

The athlon was Advanced Micro Devic Gigabyte GA-7DX Motherboard, 1.4GHz 1 cpu
The P4 was Intel Corporation Intel D850GB motherboard (2.0 GHz, Pentium 4 processor) 1 cpu

Obviously, this isn't a totally fair match (this is the most recent numbers I could find from both on this page http://www.spec.org/osg/cpu2000/results/cpu2000.ht ml) (there is no CPU2001 benchmarks... we'll probably stick with CPU2000 until 2005 or so... we stuck with the SPEC95 benchmarks for 5 years). The newer althonxps will have improved numbers, but the 1.4ghz part was beaten pretty badly (and those numbers came from AMD themselves, so their setup was optimal in their eyes).

Anyway, there you go

Re:Model Numbers

[silicon_synapse]

AMD is now not selling thier processors as 1500MHz, but instead as 'equivilant to a P4 at 1800MHz' -> an AthlonXP 1800+

Not quite right. AMD is labeling a 1500MHz processor as equivilant to how a Pentium 1800MHz based on the coming Northwood core should perform. The Northwood core will be more efficient than the P4 architecture so an AMD AthlonXP 1800 will easily outperform a P4 1800MHz but should be roughly equivilent (better still but not by as much) to an 1800MHz Northwood.

Re:Model Numbers

[Tumbleweed]

This is actually a pretty awful way to measure the performance of a CPU since it's _highly_ dependent on aspects of the motherboard. The German magazine c't did a test of the Athlon XP CPU recently and tested on two different motherboards (Gigabyte and Asus, I think) - the spec scores different wildly! Neither board tested used the VIA KT266A chipset, which is known to be the fastest (in some cases, by far) of the Athlon-supporting chipsets.

So, that test proves to NOT be a test of CPU, but of the CPU/chipset/RAM/motherboard combination, which is hardly the same thing.

Re:Model Numbers

[startled]

Even SPEC is too easy to manipulate, and too far from what most users do. The only real way to benchmark a processor is by testing it with what you do the most.

For tech-savvy users that might actually notice the speed difference, this means we have to browse the benchmarks at Anandtech, etc., and it's usually pretty easy to find a benchmark for an app identical or similar to what you spend most of your time doing (i.e. Tribes 2).

Of course, it's likely that none of us would notice the speed difference between a P4 2.0GHz or AMD 1800+ in whatever app we're using; and it's certain that a non-power user wouldn't. Which is fine, because the non-geek isn't going to read all the benchmarks.

So what's a computer buyer to do? Simple-- buy from whoever's cheaper. Save yourself $200, and try not to worry too much about remembering if your box is 2% faster or 2% slower than the other one as you surf the web.

Good CPU in spite of AMD marketing

[MtViewGuy]

I think in spite of AMD's awkward marketing plan for the Athlon XP CPU's, you have to admit they are impressively fast.

Both Anandtech and Tom's Hardware show the Athlon XP 1800+ to have pure-CPU performance that exceeds that for the Pentium 4 2,000 MHz CPU (with the exception of any program that takes full advantage of SSE2 instructions, which are still quite rare). This is a tribute to the fact that the Athlon CPU core itself is very fast, particularly the FPU unit.

Once people realize the Athlon XP's excellent performance I think the new CPU will be a good seller.

MP?

[[amorphis]]

From reading the various reviews, the Athlon XP doesn't seem to have SMP capability.

Are the Athlon XP and Athlon MP essentially two lines now? It sucks to see AMD succumb to marketing in order to combat Intel.

Re:MP?

[greenfly]

The Athlon XP uses the Palomino chipset just like the Athlon MP, so yes you can use it in an MP configuration.

For that matter you can use an Athlon tbird in an MP configuration, but the Palominos have AMD's blessing. The HardOCP article talks about this somewhat.

Re:MP?

[counterfeitfake]

At least it's smart marketing

Which will from now on be known as Smarketing.

A chip by any other name...

[Domini]

.. would still run as sweet.

Aparently that are following suit, with NVidia and their DetonatorXP drivers, everyone seems to be trying to get onto the WinXP hype.

They seem to call it Extended Performance (isn't that AthlonEP then?), and sure it has 3-7% more bang for clock than the TB line.

My only question is this, since AMDs are so popular in the linux comunity, what will the change in name do to that support? I for one don't care... :)

Any thoughts on the name's impact?

Anyone know what happened to UserFriendly?

Re:A chip by any other name...

[tshak]

How can this be fraud? It's like selling an Audi A4 and calling it an "A4 150hp" and comparing it to a "Honda Accord 150hp". Now, anybody who knows anything about cars knows that the Audi A4 will easily outperform the Honda even though they both output 150hp. The same goes for the Athlon 1.53ghz... these chips are becoming less and less dependant on clock cycles for performance. Especially when you look at the real world benchmarks - The Athlon 1800+ (1.53ghz) is outperforming the P4 2.0ghz in the vast majority of suites. However, the consumer will be comparing the 1800+ to the P4 1.8ghz, which simply get's crushed by the 1800+. I don't see how this is in any way the least bit misleading.

Athlon "XP"?

[Kreeblah]

Erm, this is getting frightening. First we had nice, normal products. 80286. Windows 3.0. DOS 6.2. Simple to note differences, no? Then we had products which were easier to copyright the names of. Pentium. K6. Windows 95 (OK, that wasn't really for copyright; that was just for misleading people). Now we're seeing a return to the old days, except without the clarity. Office XP. Windows XP. Athlon XP. See, now companies appear to be marching in lockstep. Have the same name, and confuse the customer. I can hardly wait for the "Pentium XP" . . .

XP hype

[nizo]

Since X-( is listed in the smiley guide as: User just died I nominate the new XP smiley to be listed as User just died of confusion (with tongue sticking out) after trying to figure out if Windows XP would run better on a machine with an Athlon XP CPU or a faster(?) Intel cpu (NOTE: Trademarks above are owned by respective companies blahblahblah)

Now what?

[cr@ckwhore]

Ok... so, now I apparently have to stop using AMD processors after my athlon 1.4, because I won't be able to determine the true mHz that my processor is running. I don't necessarily see the reasons why this rating is masked on the XP processors... its probably ok for the average home user, but I'm not average. MhZ ratings mean something to me, because I enjoy tweaking the most from my system.

I stopped using Intel processors a while ago, after learning that AMD's chip architecture was superior to Intel's, the choice was obvious. If you haven't read this document, please do. It'll give you a good technical understanding of performance issues with Pentium processors compared to AMD processors.

So, now what? I guess I'm forced into some hard choices over the specs of my next machine. It may be time to consider Intel again... I just don't know. AMD's new CPU scheme sounds really sketchy to me.

Re:Now what?

[Junks Jerzey]

MhZ ratings mean something to me, because I enjoy tweaking the most from my system.

MHz ratings _shouldn't_ mean something to you in that case. You'd really pick a 2GHz CPU over a 1.8GHz model, even if the latter were 20% faster?

Consumers in general will be fine with this change, but geeks are going to implode. Too many have made a hobby out of tracking MHz and transistor count and other meaningless numbers. Unfortunately, it's about the same as horsepower in cars. More is not necessarily better. And no one who buys a car fixates on horsepower above all else.

Re:Now what?

[aussersterne]

Chip still runs on a clock, so what?

No, clock cycles do not mean anything about performance to "some talented computer users." Here's why, using CPUs other than AthlonXP and Pentium4 so as not to inflame anyone:

The old Intel 8-bit CPU used in PC/XT machines ran at 4.77 MHz (4,770,000 clock cycles per second) but this does *not* mean that it could do 4,770,000 *things* per second, because each time it needed to execute an instruction, it took several (i.e. more than one) clock cycles to do so. Furthermore, the largest numbers it could operate on natively were generally 8-bits long -- a 32-bit calculation, for example, required user code to complete, which of course meant many, many more cycles.

The Hitachi 6309 CPU of the same time period, by comparison, ran at 2.0 MHz (2,000,000 cycles per second), but was **MUCH** faster for the same types of tasks than the Intel 8-bit CPU because it could *often* finish a complete instruction in only one clock cycle and because it had 16-bit registers and a 32-bit register and could thus do MANY types of math *natively*, in just one or several cycles, that the Intel CPU needed user code (and thus, hundreds or thousands of cycles) to complete.

Because of these types of _architectural_ differences, clock cycles have little or nothing to do with the real speeds of different chips performing real-world tasks (which, for gamers, includes things like Quake 3). In fact, clock cycles and MHz are *the same thing*, as MHz on a CPU simply means "number, in millions, of cycles per second."

You will find no statistical correlation between the *actual* clock speed on an AthlonXP and each of the benchmarks vs., say, a Pentium4 at 1800 MHz. Yes, one is running at ~1,500,000,000 cycles/second and one is running at ~1,800,000,000 cycles per second, but that doesn't tell you how many cycles each one is spending doing different types of tasks or (as is often the case) sitting around waiting for data from the rest of the system or from the bus.

The P4 needs a PR rating...

[(H)elix1]

The most irritating thing about AMD switching to a PR rating is most folks miss the fact that mhz vs mhz, the Pentium III blows the Pentium 4 out of the water. It all comes down to what gets used as a normal - using a P3 as the mHz reference point, you get the AMD chip wiping the P3, and the P3 owning the P4. The P4 could use a PR rating as well...

Intel can't make it faster, but we can increase the number of cycles... can marketing do anything with that? Intel killed the PIII because the last thing they wanted was for someone to take a 1.5gHz chip and put the P3 & P4 side by side.

Depending on how you tweak the benching and load things up, you will see strengths and weaknesses in each CPU. Priced the same, the AMD chips are a better deal for my development and gaming needs.

Marketing capitulation

[dgb2n]

Its very sad but AMD is essentially admitting through their marketing ploy that the average consumer is incapable of realizing that the speed of a processor and indeed a system is more than a clock frequency.

Although virtually every reviewer pans the confusing processor labelling, I believe that it was a good business decision. With the success of the Athlon processor, AMD went a long way towards minimizing the marketing impact of "Intel Inside". Now they find themselves "burdened" with a processor which out performs its competitors significantly at a given clock speed. If they label the chip with its clock frequency they invite price comparisons to similarly clocked (but underperforming) Intel products.

I think the new labelling scheme is actually a win for AMD. Smart consumers will buy the chips because of their superior performance, regardless of the name. "Joe 6 pack" will buy it because he can buy the AMD 1700+ system for less than the Intel 1600.

No cheaper than Intel

[Animats]

From CNNfn:
"In 1,000-unit quantities, the Athlon XP 1800 is priced at $252. The most recent list price for a 1.8 GHz Pentium 4 was $256. The Athlon XP 1700 will sell for $190, compared with $193 for a 1.7 GHz Pentium 4. The Athlon 1600 lists for $160, compared with $163 for Intel's 1.6 GHz Pentium 4. The list price for the Athlon 1500 is $130, compared with $133 for a 1.5 GHz Pentium 4."

So AMD doesn't have a significant price edge on this round. That's bad for AMD; they need a price edge to win over vendors.

Without competition from AMD, Intel CPU chips would cost around $1000. We know this because they used to cost that much. Remember when Pentium Pro CPUs cost around $1000? AMD didn't have a high-end offering back then, and Intel could get away with huge markups. That's the difference between a monopoly and competition.

The real test will come when AMD starts shipping the Thunderbird, which is not instruction-compatible with the Intel Itanium.

Re:No cheaper than Intel

[Animats]

("Thunderbird" above should be "Sledgehammer". Sorry. Considering that I'm writing this on a Thunderbird machine.)

Microsoft Does it....

[theAmazingTom]

"1. Motherboards will not pass AMD validation or be posted on the AMD recommended motherboard Web site, if the frequency is displayed by the BIOS during bootup for AMD Athlon Model 6 decktop and multiproccesng processors."

It's one thing to sell it as an 1800+ but I'd still like to know what the MHz is.

AMD's heatsink problem?

[deander2]

Has anyone heard whether or not AMD's heat problem has been solved? Reading Tom's article on what happens if your heat sink falls off really put a kink in my AMD-buying choice. I mean, it wasn't even like you had time to hit your power button - you went from 'snap' to 'smoke coming from case' in less than a second.

No matter how much faster and cheaper they are then Intel, that's a HUGE risk to take on your system.

Re:AMD's heatsink problem?

[FFFish]

I worry about my heatsink falling off about as much as I worry about my dink falling off, i.e.: not at all.

Re:AMD's heatsink problem?

[Xibby]

Umm...if you're heat sink falls off, you have one of the following problems:

You're and idiot and installed it wrong in the first place.

You're system vendor is an idiot and installed it worng in the first place.

You're motherboard is made of cheap materals

You forced something didn't you (see 1)

There is no good reason for your heat sink to just fall off.

Re:AMD's heatsink problem?

[kreyg]

Wow, if my heatsink fell off, it would fall directly on, and likely short out, my video card - which costs as much or more than my processor.

Welcome to FUD.

It's not the end of the world.

[NetJunkie]

You'll still know what speed your CPU runs at. It's not like it'll be a huge secret. Go in to the CPU setup on your Abit board and it'll tell you, they just hide it during boot so normal users don't see it.

Re:Athlon or Athalon?

[Black Parrot]

> "True to form, AMD has released the new Athalon XP today"

> I really hate when it gets spelled that way for some reason.

That's the athaletic spelling. Athaletes need computers too, you know.

Comparison with the old Cyrix-scheme

[GauteL]

Cyrix used to sell their processors with a PR-rating. PR150 which tried to compete with a Pentium 150, was actually a lower MHz.

The difference in the policy is that the Cyrix PR150 was only in _some_ applications the equal of a Pentium 150, at others (gaming) it was truly pathetic.
The AMD Athlon XP 1800+ is in almost every regard better than Pentium IV.

The conclusion is that, even though I wish AMD would market their processors on MHz, they are actually not overhyping their processors when stating in this marketing, like Cyrix did.

Problem with marketing based on mhz

[mcdade]

We know that marketing your product based on MHZ alone will kill it. This was seen right from the early days of the 386's. People looked at the generation of chip and the equivalent clock speed. Teck people would ignore this ask they knew real benchmarket would tell the truth.

Look at recient times, you have a bigger gap in this problem. The G3's and G4's are clocked between 400mhz and 800mhz, but people are put off buying one cause they can get a PC with 1.2Ghz for cheaper. The G4 can be a faster chip with lower clockspeeds but people won't buy it cause all they see is 800mhz vs 1.2ghz. The bigger number in compters means it's better, everyone knows that!

If AMD doesn't start PR rating their chips people won't buy them. They are slower and cheaper (in the mind of Joe Sixpack) so they must not be as good as an Intel.

Re:SMP

[Brian Stretch]

The previously released Athlon MP is the multiprocessor certified version of the Athlon XP. Faster versions of the Athlon MP are expected to be announced next week, with new dual processor motherboards (in addition to the Tyan Thunder and Tiger MP boards already on the market) expected next month.

The Athlon4 notebook CPUs are also equivalent to the Athlon XP desktop CPUs (with the addition of PowerNOW! power management, natch). As notebook and MP-certified CPUs are higher margin parts than uniprocessor desktop CPUs and AMD had no previous MP or notebook Athlon offerings, AMD directed their new Palamino-core fabrication lines to those markets first.

rating the performance of processors

[Tumbleweed]

Well, just because the SPEC tests are awful for CPU performance measurement, doesn't mean there's any single test that's good. :)

I think the only _real_ test of performance is a comprehensive set of real-world (read here: real applications!) tests. That, too, is not a test of _just_ the CPU's performance. I don't know how you'd be able to accomplish that, aside from spouting off some simulated results.

Unfortunately for the consumer, it's not possible to translate real-world performance results into a magic number that they can quickly or easily read to see how fast a system is. That's just life. People need to do research on things. Most people wouldn't be able to tell the speed difference between a 1gHz Athlon and a 2gHz Pentium IV system, anyway, so the point is moot for most people. Those of us who care about such things know where to go and what to look for when researching a computer purchase.

For full systems, a SPEC score might make a small amount of sense - then Dell could advertise their SPEC scores for each system, Gateway could for theirs, etc. But for those of us who buy on a component level, it makes no sense at all. The KT266A motherboard speed improvements over the AMD 760 chipset will probably offset those SPEC scores and let them Athlon XP 1800+ come out on top of the Pentium IV 2gHz CPU. (at least until Northwood comes out).

I guess my main gripe is that SPEC is being bandied about (even by the CPU manufacturers) as a measurement of pure CPU performance, when clearly, it is not. It's unfortunate AMD chose to publish their scores on a platform that's not the fastest. *shrug*

I'm such a nerd in that I even care about this stuff! :)

I'm really waiting for a DDR333 Athlon platform to come out next year. Hopefully there'll be a VIA KT333 chipset and also hopefully the Athlon 'Barton' (0.13micron Athlon platform) will have a 333mHz DDR FSB to mate to it. I've got other purchases in mind until then, assuming I ever get enough money to make them in the firstp lace.