• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

The 2166mhz Barton does not deserve its "3000+" PR

Page 6 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.
MadRat... Oh, well if theinquirer or register said it, then it must be true!
rolleye.gif
You'll need to come up with a better source than that, before anyone is going to believe those accusations.


mechBgon... "Real world performance suggested by the 1.7ghz label"??? Well, I guess the 1.0ghz Duron shouldn't have been called a 1.0ghz then. How about the 533mhz Celeron? There have always been same-mhz chips with different performance. I don't understand why it's such an issue now.

As I said before, many times the AMD PR rating is alright. But in the case of the "3000+", it is more obvious than ever that it's an arbitrary number with no actual methodology behind it.

Take SSE2 apps out? Why would you, if they are very popular applications (such as Photoshop, etc)? I mean if all the reviewers were using some obscure application, then I'd understand.


DynaOne... So, you are saying that the majority of people in the market for a $600 cpu use mostly Office apps? Not photoshop, or 3d rendering, or video encoding, or audio encoding, etc...? I'll have to disagree with you there. I think how fast Photoshop (or video, audio, etc) runs is infinitely more important than how fast Word or Lotus Notes can run.

 
If there was anyway for it to be possible AMD should have released the 3000+ at 2250MHz.

There are a couple of reasons for this:

- The 3000+ would outperform the 2800+ under ALL circumstances.

- The 3000+ would compare more favorably to the P4 3.06.

- Faith in the PR system would be strengthened instead of weakened.

I think these advantages outweigh any losses that might be incurred by not reaching as high a PR rating by the end of Barton's run.
 
Originally posted by: Wingznut
MadRat... Oh, well if theinquirer or register said it, then it must be true!
rolleye.gif
You'll need to come up with a better source than that, before anyone is going to believe those accusations.


mechBgon... "Real world performance suggested by the 1.7ghz label"??? Well, I guess the 1.0ghz Duron shouldn't have been called a 1.0ghz then. How about the 533mhz Celeron? There have always been same-mhz chips with different performance. I don't understand why it's such an issue now.

As I said before, many times the AMD PR rating is alright. But in the case of the "3000+", it is more obvious than ever that it's an arbitrary number with no actual methodology behind it.

Take SSE2 apps out? Why would you, if they are very popular applications (such as Photoshop, etc)? I mean if all the reviewers were using some obscure application, then I'd understand.


DynaOne... So, you are saying that the majority of people in the market for a $600 cpu use mostly Office apps? Not photoshop, or 3d rendering, or video encoding, or audio encoding, etc...? I'll have to disagree with you there. I think how fast Photoshop (or video, audio, etc) runs is infinitely more important than how fast Word or Lotus Notes can run.
Click to expand...

**edit: you know what? This is going nowhere fast, just like all PR debates do. Have fun guys 😀
 
Originally posted by: Wingznut
MadRat... Oh, well if theinquirer or register said it, then it must be true!
rolleye.gif
You'll need to come up with a better source than that, before anyone is going to believe those accusations.


mechBgon... "Real world performance suggested by the 1.7ghz label"??? Well, I guess the 1.0ghz Duron shouldn't have been called a 1.0ghz then. How about the 533mhz Celeron? There have always been same-mhz chips with different performance. I don't understand why it's such an issue now.

As I said before, many times the AMD PR rating is alright. But in the case of the "3000+", it is more obvious than ever that it's an arbitrary number with no actual methodology behind it.

Take SSE2 apps out? Why would you, if they are very popular applications (such as Photoshop, etc)? I mean if all the reviewers were using some obscure application, then I'd understand.


DynaOne... So, you are saying that the majority of people in the market for a $600 cpu use mostly Office apps? Not photoshop, or 3d rendering, or video encoding, or audio encoding, etc...? I'll have to disagree with you there. I think how fast Photoshop (or video, audio, etc) runs is infinitely more important than how fast Word or Lotus Notes can run.
Click to expand...

Wingz, I can't believe it, don't let me down, look at the statement:"As I said before, many times the AMD PR rating is alright. But in the case of the "3000+", it is more obvious than ever that it's an arbitrary number with no actual methodology behind it."

Yes, as I think we agreed, they were too optomistic with that rating, but it was not pulled out of the air. They were off by 66-100 mhz. The rating has no "methodology" per se, its just that they try to pick a P4 that they think it beats, and then name it. They just screwed up this time. At 2900+, it might have been optomistic.
 
Personally, I could care less how close in performance it is relative to a 3.06GHz P4. Most people who buy AMD buy them because they are cheap. You still can't get a good CPU from Intel for under $100. AMD has their place, and so does Intel. The Athlon64 might change that quite a bit (I think SSE2 support is going to really help in certain benchmarks). AMD just needs to get it out the door one of these days.
 
Originally posted by: Bovinicus
Personally, I could care less how close in performance it is relative to a 3.06GHz P4. Most people who buy AMD buy them because they are cheap. You still can't get a good CPU from Intel for under $100. AMD has their place, and so does Intel. The Athlon64 might change that quite a bit (I think SSE2 support is going to really help in certain benchmarks). AMD just needs to get it out the door one of these days.
Click to expand...

You obviously haven't seen the price of the XP 3000+.
 
You obviously haven't seen the price of the XP 3000+.
Click to expand...
Yes, I have. It is about the same as a 3.06GHz P4. Both of them are out of my range, and I am therefore not concerned with their performance at the moment.

After actually reading the review, I have concluded that the CPU is in fact worthy of that performance rating for my purposes. In games and general applications, the rating seems to be very justified. In professional applications and media encoding, the rating isn't so accurate. However, those are the two types of applications I use most. HT did a good job distancing itself in media encoding, but games and apps get the priority for me. However, as I said, I can't afford it anyway so it doesn't really concern me.
 
Well, I read all the posts and it was not as bad as I thought it was going to be.

I spent a few hours reading Anand's review, examining hundreds of bars (or was it thousands?), and mentally calculating the percentages. Before I read the initial premise in this thread, I had concluded that AMDs 3000+ rating was an underestimate as per AMD's usual. The 3000+ beats the 3.06 P4 in "office apps" and it beats it heavily in UT3. In a couple of media creation apps using SSE2, the 3.06 walks away from the 3000+, and in just about all, the 3.06 has a handy lead. It really doesn't matter that web sites are running dozens upon dozens of media creation benchmarks showing how thoroughly the 3000+ is beaten in media creation. It is 1/3 of the score, and AMD doesn't necessarily agree that these are the appropriate media creation benches. The 3000+ is still the winner. If AMD would explicitly say what the numbering system is, reviewer's could run the component benches and get the same numbers. Maybe AMD could just send out a CD along with the CPU wiith it all arranged and the reviewer's could just run it. Right! Would anybody go for that? I doubt it.

Anyway, content creation is Intel's marketing system for their CPUs. There is very little justification for Intel's chosen CPU path otherwise. It is the way Intel approaches selling advanced CPUs, and it gives crippled Celerons the same tint to the eye of the buying public, which is a key to getting a good price. If I were doing content creation, and it was worth the money to me, I would get a 2.4 P4 or better. I guess you can tell I don't. Sure I've created a few dozen MP3s in the last few years, amounting to a total of possibly one hour of CPU time, and that is probably more than 95% of computer users. I've also created a few JPGs, but it seem to be about instantaneous. So in my personal idea of a realistic rating system, content creation would be way overweighted if it was 1/3 of the total, especially the content creation that these dozens of benches do.

There is an old, traditional, and well accepted rating system for CPUs using very old established benchmarks. One of them rates CPUs in flops (floating point ops). Unfortunately this has barely any relevence to 99.999% of computer users, so you can see why no one even brought up the flops for a P4 3.06 or 3000+

I don't know if people remember this, but when the Pentium plain was introduced, Intel had a rating system for their processors, and it went on for years. I forget the name. By their numbers the super-duper Pentium 60MHz beat, almost doubling, the peformance of a 486DX4 100MHz. (It accomplished this by doing more instructions per cycle. How else?) In order to get people to go for a higher priced CPU (something like $1200 back then - that's just the CPU folks), you have to give them a rationale. So Intel had this theory of what people ought to want their CPUs to do. Reviewers at the time didn't think much of the rating because it gave such a high weight to peculiar things.

To the people who wonder why the 2800+ beat the 3000+ on a lot of THG benchmarks: They are the identical CPUs with different amounts of cache. The clock is slower on the 3000+. Just run the appropriate program and it runs faster on the faster clocked CPU. For instance, if the program and data fit in the smaller cache, why would it run faster with more cache? Or if the data is never reused and the program fits in the smaller cache (ie content creation), why would more cache help? It might actually slow things down due to cache overhead. The point of cache is really statistical. How often is a larger cache an advantage? For typical use, a larger cache is almost alway more useful. You can ALWAYS find cases where that is not true. Engineers design by setting some hit rate. In the benches Anand ran, I was amazed at the advantage the larger cache usually had, fully justifying the 3000+ rating.

When the methods of optimization differ so radically between two CPUs, there can be no such thing as a single number which sums up performance. The premise of any CPU rating will be violated if your use system differently than the rating system, so even with a "fair" rating system, you still have to make some evaluation of its relevance. Just because a web site did thirty content creation benchmarks does not mean there is some large significance to this type of program.
 
Benchmarks don't account for what typical workstation users do, which is open about ten applications at one time while they work.

I'm sure extra cache makes a difference in this situation, but we'll never know that by the benchmarks.
 
Benchmarks don't account for what typical workstation users do, which is open about ten applications at one time while they work.

I'm sure extra cache makes a difference in this situation, but we'll never know that by the benchmarks.
Click to expand...

Yeah, and hyperthreading too.
 
Bingo!

Originally posted by: Hulk
If there was anyway for it to be possible AMD should have released the 3000+ at 2250MHz.

There are a couple of reasons for this:

- The 3000+ would outperform the 2800+ under ALL circumstances.

- The 3000+ would compare more favorably to the P4 3.06.

- Faith in the PR system would be strengthened instead of weakened.

I think these advantages outweigh any losses that might be incurred by not reaching as high a PR rating by the end of Barton's run.
Click to expand...

 


AMD explains bizarre Barton 3000+ PR rating

The answer lies in the soi(l)
PRESS REPS at AMD have returned with an answer to a question we posed last week.
That was in response to a roadmap we pixellated last week about future Barton core speeds, and which you can find here.

We wanted to know why the "Barton 3000+" microprocessor was originally slated to launch at 2.25GHz, but was instead released at 2.17GHz.

Said AMD: "One of the advantages of model numbers is that when cores change you don't have the age-old discussions about guessing how much faster the core improvements make the new products. For instance: 'Is a P4 20% or 30% slower that a PIII, clock for clock?'"

In other words - our words these - AMD didn't need, in its own eyes to introduce a higher frequency for the Barton because it believes that the 2.17GHz chip can outperform an Intel 3.06GHz Pentium 4.

The same source pointed out that it's impossible to pick up copies of SPECint and SPECfp at PC World (UK) or Fry's (US).

AMD claims that it has attempted to use applications that real people use everyday, and that's a variety of different applications.

But let's take this a little bit further. Does this mean that if AMD introduces the Athlon64 in September, after Microvole has finished tweaking its famous X86-64 operating system, a lowly clocked 64-bit chip, even if it runs at 1.8GHz, say, will outperform a Pentium 4 3.20GHz?

Apparently, the answer to that is yes. And there's another thing. This is probably the reason that AMD will have to wait for the Microsoft OS.

Because, as far as we remember, those real world applications are all Windows based ones. So it won't be able to pull the PR gig without a proper WinOS to base them on.

We never tire of quoting Jerry Sanders III on this one: "With Intel and Microsoft we are the holy trinity".

Although the "real men have fabs" one is proving to be somewhat of an embarrassment, what? µ
Click to expand...

Flame on. 😛

 
Said AMD: "One of the advantages of model numbers is that when cores change you don't have the age-old discussions about guessing how much faster the core improvements make the new products. For instance: 'Is a P4 20% or 30% slower that a PIII, clock for clock?'"
Click to expand...
Isn't that exactly what AMD just did? They said the core change means the Barton performs faster than the T-bred. AMD gave a definite value to this: 300 PR for the 2.17 GHz chips. So AMD just did what they said the PR system will avoid.

 
Benchmarks lie, benchmarks are for liars.

Intel knew the P4 had some problems so they taylored Benchmarks to make it look good and passed these off as mainstream. They stated this when they were developing the P4.
Bapco shares an address with Intel. Sandra measures Mhz. Others are optomised for Intel.

Bartons 3000+ is a legit as Intels 3.06 P4.
 
Originally posted by: Wingznut
ndee, LeeTJ, and Evan... I could've chosen to remain anonymous about my employer (as a number of other people who post here do), but I decided to be up front with people. It's the way I've been my whole life, and I'm not going to hide behind a keyboard on a msg board. 😉

And because I disclose that fact, it makes posting my opinions more difficult. I don't have the room for error that others have, ESPECIALLY when posting about AMD. If I don't have my facts straight, and present them very clearly, I would be the most flamed person here by far. 😉

That being said, I just hope that somewhere amongst my 8100+ posts I've gained enough respect and the understanding that my career is in making cpu's.... NOT selling them. 🙂


paralazarguer and SexyK... Sure, it helps to have the capital to spend on fabs... But give the talent pool some credit, too. 🙂


Vespasian... All Bartons are being manufactured in Fab30 in Dresden, Germany. I'm not certain where AMD and UMC stand now, after the agreement with IBM/AMD (their 3rd such colaboration in 4 years). I believe the partnership is dissolved, but I'm not positive.


jbond... Absolutely no animosity or implications taken... I just thought it would be a good idea to clarify that I am my own man. 😉


LeeTJ... You are correct. Nobody is truly objective. But I do think that I am less biased than a good number of people on this board. I'll tell you right now that Barton is an excellent piece of silicon. It's just the PR that I don't care for.

Hey, I was a PC enthusiast long before an Intel employee. 🙂
Click to expand...

10 🙂
 
The 2.0GHz Celeron Does Not Deserve Its "2000" MHz
Click to expand...

It has a PR Rating of 1200😉.

But AMD still has a chance to clarify things here. Maybe what they meant with the 3000+ PR rating is comparable to the P4 3.06 without HT or SSE2? Or maybe apps are not optimized yet for the 512kcache? Or maybe, like somebody has mentioned, drivers(he he) or BIOS optmiizations? What they need here is a good lawyer working with a good pyschologist in order to blur the lines further, so there is 'perceived' parity between the P4 3.06 and the 3000+ Barton. The PR rating is pretty much like the 'mysterious' weapon in "Dude, where's my car?" movie (can somebody quote that for me?).
 
But it's really not that complicated...

Had they named it a "2900+", they would have a much harder time hanging a $588 pricetag on it. By upping the rating to "3000+", they can validate their pricepoint.
 
food for thought...

maybe barton @ 2250Mhz compares nicely with northwood at 3200MHz...
and so barton @ 2250Mhz = Athlon XP 3200+ ?!?

---or---

Why hasn't AMD implemented SSE2 on Barton? We already know they are cranking out Opteron cores with SSE2, why not add it to the current Athlon? Man I bet the benchies would be very interesting if they had done so.
 
The true insult would be allowing a p4 with I850E to use PC 1066 since no major computer company ships with it. So if you want to talk about chipset performance choices I'd think id through that at you.

I do think it deserves the 3000+ rating but I very much so wish they would have at least kept it clock speed at 2250 and 2160. I am very upset that AMD chose to lower their clockspeed for the 3000+ as it makes little sense to me. I wish they could have realeased it @ 2300 or above since that would have dominated the p4. A 200 fsb would have been really nice as well and where are those SSE2 optimiztions! Geeze that is what really helps the p4 not HT for sure. Few people here realize that HT has been on the XEON socket 603 chips for quite awhile. It doesnt help the XEON best the Athlon MP in most of the benches and I dont think its that great of a feature since an application must be HT to really use it.

The 3000+ does seem like a healthy competitor to deserve the rating and I am really unsure by Wingnutz figures as they seem off by a significant margin. I have read several other reviews that have conclude AthlonXP 3000+ is the top dog around. I think AMD did put out the best performing chip out there I just wish they would have done it by a much higher margin.
 
You must log in or register to reply here.

TRENDING THREADS

Back
Top