AMD launching Centrino Killer

[dullard]

Originally posted by: dguy6789
Thats where you are wrong. Read reviews and all. Untill the C models came out, the athlon xps PR ratings were right on the money, with the 3000+ being faster then the 3.06ghz(still is).

I think he is considering the timing as well. The 3.06 GHz P4 was launched when the 2600+ and 2700+ were still paper launched. The first store that had either of them listed as available was one day after the 3.06 GHz P4 was available. The 3000+ was no where to be seen. And according to reviews, such as Anandtech's

"In many cases the Athlon XP 3000+ can outperform the 3.06GHz Pentium 4, while in others it manages to tie with Intel's flagship and yet in others it falls behind just as much. The overall performance is close enough to warrant the 3000+ rating in some cases, but there's no question that it is a very close call between the two top performing CPUs."

A CPU which was 4 months behind and just finally came even in performance is what I think Accord99 is taking about.

The P4 was awful - the Athlon XP won. The P4A was quite good and took the performance crown back. The lead has oscillated back and forth since.


As for the Centrino Killer, I have the guess that it is a killer in terms of price, not in terms of performance. We will soon see though.

 

[Accord99]

Originally posted by: dguy6789
Thats where you are wrong. Read reviews and all. Untill the C models came out, the athlon xps PR ratings were right on the money, with the 3000+ being faster then the 3.06ghz(still is).

The 3.06 came out months before the 3000+, at any given time the P4 held the edge and the 3.06 is not slower, it wins most of the benchmarks on Anand's review of the 3000+.

 

[sandorski]

We'll have to wait and see, but don't forge that AMD has many cross-licensing agreements it can source for technology. Some of those technologies deal with low power Processors. AMD has been doing the impossible for the better part of a decade now, don't underestimate them.

 

[formulav8]

Whoever up there that said Cache was cheap to produce is wrong. Not sure where he read that at. I always read that cache was expensive and added tons of resisters to the core (depending on the size cache of course ).



Jason

 

[Accord99]

Originally posted by: formulav8
Whoever up there that said Cache was cheap to produce is wrong. Not sure where he read that at. I always read that cache was expensive and added tons of resisters to the core (depending on the size cache of course ).

Cache is cheap to produce, relative to logic, because it can be manufactured densely, consists of simple, regular blocks of transistors and can be protected against defects through the use of redundancy. Logic on the other hand is much more complex and a single defect could make the entire CPU useless.

 

[formulav8]

Shows me a link please and I can show you links that contradict what you are saying 🙂

Even your buddy Anand says Cache is not cheap and adds alot of transistors which adds cost and complexity and such.

Not that I care, which I really don't. Its just what you said is different than from what I have been used to seeing unless I have been reading wrong (which definitely wouldn't surprise me) 🙂



Jason

 

[Accord99]

Originally posted by: formulav8
Shows me a link please and I can show you links that contradict what you are saying 🙂

Even your buddy Anand says Cache is not cheap and adds alot of transistors which adds cost and complexity and such.

Not that I care, which I really don't. Its just what you said is different than from what I have been used to seeing unless I have been reading wrong (which definitely wouldn't surprise me) 🙂

Yes it does add to the complexity and cost of a chip, but the increase is not that significant because cache doesn't take that much space (less than 12mm^2 per 512KB on Dothan) and redundancies mean that the CPU is still usable even with defects in the cache. It certainly adds less cost to a CPU then integrating a memory controller, and one of the reason why Dothan is much cheaper to produce than any other x86 CPU.

 

[PingSpike]

Originally posted by: LikeLinus

Originally posted by: Kaieye
So even if Intel has several teams working, why is AMD breathing down their necks? Or should I say, when will Intel build a faster gaming or server cpu??

I really think all the AMD fanatics are forgetting one crucial thing.

AMD started a mhz war that Intel answered with the P4. The P4 was built from the ground up to scale very high mhz wise, but because of the length/number of pipelines, it's not as efficient. The whole arcitecture was built for the war that AMD started. Then all the sudden AMD turned around and started their PR ratings bullsh!t because they had no chance of beating Intel at the fight they started.

The Mhz war has been going on since the IBM PC was introduced. Don't act like AMD started it. All that happened was they finally won a battle.

The PR ratings are immensely annoying...but a necessary evil to attempt to market their chips to those with poor understanding of PC hardware.

 

[Ardan]

I agree that you shouldn't sell AMD short. Look at their dual core chips and how it was obivous that, as they said, planned to go to dual cores when they were designing the K8 architecture. How do we know they weren't also thinking of a low-power, high performance mobile option? And it doesn't have to be because of intel, perhaps they were thinking of it as a lucrative item. People said they can't compete in the server market and other markets and they're doing great in them.

I think we should all just wait and see what happens in 2005. Yes, it was the inquirer saying it but at the same time, none of you (including me) have any idea what AMD or Intel may or may not have up their sleeves. Unless, of course, you build CPUs for a living 🙂.

Its going to be a great 2005 for CPUs, and lets just be thankful there is healthy competition in that market. Enough of this fanatical nonsense.:beer:

 

[LikeLinus]

Originally posted by: PingSpike

Originally posted by: LikeLinus

Originally posted by: Kaieye
So even if Intel has several teams working, why is AMD breathing down their necks? Or should I say, when will Intel build a faster gaming or server cpu??

I really think all the AMD fanatics are forgetting one crucial thing.

AMD started a mhz war that Intel answered with the P4. The P4 was built from the ground up to scale very high mhz wise, but because of the length/number of pipelines, it's not as efficient. The whole arcitecture was built for the war that AMD started. Then all the sudden AMD turned around and started their PR ratings bullsh!t because they had no chance of beating Intel at the fight they started.

The Mhz war has been going on since the IBM PC was introduced. Don't act like AMD started it. All that happened was they finally won a battle.

The PR ratings are immensely annoying...but a necessary evil to attempt to market their chips to those with poor understanding of PC hardware.

Ok

1. I never said there had never been a mhz war. If I did, please feel free to show me.

2. IBM has been competing in the desktop market for how many years now? There were no Mhz wars the past 12 years. Once Intel produced the Pentium, no one has been able to battle with them.

3. AMD DID initiate a mhz war with slam ad?s on TV and magazines saying their 1ghz chip was the fastest and yadda yadda. Intel is/was smart enough to know that mhz is what sells. Regardless of what the enthusiast and anyone with a brain realize, Intel is catering to Joe Public. They weren?t about to lose them to AMD. That?s the whole reason behind the P4 chip. It?s created to scale.

4. I?m not sure how AMD finally won a battle. They started a mhz war and as far as I can tell, the Intel chip is far beyond it compared to the XP and 64 line. Of course we all know that the 64 is a superior chip clock for clock, but it?s not like Intel has a bad chip. They also haven?t really lost much market share in the process?which is really where the war is.

Anyways, back to the original post. We're getting off topic. We just really need to wait and see what happens. It could take them sometime to get it perfect. I don't think it'll be a winner right out of the gate. Time will tell

 

[Snoop]

Originally posted by: Pariah

And i dont see why amd cant do it. AMD's Athlon 64 is faster clock for clock then the pentium M, and costs MUCH less to make. All they need to do is reduce power requirements a bit and they will have intel owned.

Might want to check out this article. As always, performance is application dependent. In gaming, the Pentium-M is the better clock for clock than the A64, and for the home user that is where CPU performance comes into play the most. And that's with the M running on an ancient platform. Looked at the overclocked numbers which boost the FSB, at 2.3GHz, the M is running nose to nose and sometimes even beating AMD's best FX-55 which is running 300MHz faster on a vastly superior platform. When the Pentium M's next platform is finally released with a "new" 533MHz bus, the Pentium M will be quite a beast.

From the article conclusion:

"While putting up impressive gaming numbers is one thing, one simply must consider the benefits of the Pentium-M architecture in comparison to these other high-end CPU?s. The Pentium-M can compete typically perform within 5% of top of the line Intel/AMD consumer level processors in gaming while running at one fourth the heat production levels. Power consumption numbers are also far, far less on the Pentium-M compared to other modern processor lines. Basically, you?re getting solid gaming performance without all the nasty side effects of running at high clock speeds, thanks to the efficiency of the Dothan core architecture. "

From GamePC's first article:

100% load

AMD Athlon64 FX-55 2.6 GHz - 53 ° C / 127 ° F
Intel Pentium 4 3.6 GHz (775) - 67 ° C / 152 ° F
Pentium-M 2.0 GHz (Dothan) - 35 ° C / 95 ° F

Those numbers speak for themselves, but it still has to be added that the Pentium M was being cooled by one of these, which wouldn't have a prayer in hell of working with either of the 2 CPU's it was compared to, making the results even more impressive.

Good luck AMD, but Intel is out of your league in this market.

Pariah, something is odd in that GamePC review. When compared to the Anandtech review of similar processors, the performance deltas btw the tested chips are far out of line. It would be interesting to do a statistical analysis (btw the two reviews), but by simple percentages, in the GamePC review the FX performed on average 10% slower versus the EE than in the Anandtech review. This, would most likely equate to three or four speed grades and makes me wonder if there was a problem with the test bed or possibly the specific benchmark routines of one of the two reviews, but is impossible to tell.

IE:All tests done at 1024x768, same video card, ATI x800:
3.4 EE vs Athlon FX 55

Doom3
Gamepc- EE 121.5 - FX 123.3 ---FX by 1.5%
Anand- EE 87.9 - FX 102.5 ---FX by 16.6%

CounterStrike Source
Gamepc- EE 56.86 - FX 61.06 ---FX by 7.3%
Anand- EE 160.7 - FX 190.3 ---FX by 18.4%

Unreal Tournament 2004
Gamepc- EE 234.4 - FX 253.2 ---FX by 8%
Anand- EE 61.1 - FX 74 ---FX by 21.1%

Halo
Gamepc- EE 94.1 - FX 96.9 ---FX by 2.9%
Anand- EE 91.4 - FX 106.6 ---FX by 16.6%

Battlefield Vietnam
Gamepc- EE 175.3 - FX 173.2 ---EE by 1.2%
Anand- EE 240 - FX 240 ---Same
FarCry
Gamepc- EE 130.3 - FX 139 --- FX by 6.6%
Anand- EE 140.9 - FX 158.9 --- FX by 12.7%

Their is something definately wrong with these numbers.

 

[Sunner]

Originally posted by: sandorski

Originally posted by: Sunner

Originally posted by: sandorski

Originally posted by: ViRGE
With no offense to AMD, I don't believe they have the ability to make a chip that can offer the performance/power ratio of the Pentium-M - it's a very well designed chip in a field they don't have much experience with.

I think you could be wrong. AMD "shouldn't" have been able to make the Athlon superior to the P3(and all P4s up to the P4c), but it did. AMD "shouldn't have been able to make the Athlon64 superior to the P4, but they did. When it comes down to it, what does Centrino have that's so hard for AMD to achieve?

Intel can afford to have several teams working on several products at once, I doubt AMD can.

True, yet AMD has been winning in the CPU technology arena.

In the consumer space, technically speaking, yes, in terms of sales, no.
And in the mobile space, I seriously doubt they'll be able to compete for a while, unlike what someone said a few posts up, the Pentium-M isn't just "A reworked Pentium iii".

 

[miketheidiot]

AMD will get close, not quite though.

It will be nice to have a non-bundled option though.

 

[Macro2]

AMDs problem is marketing. When Intel can't rely on high MHz marketing any more...it will help AMD a lot.

Dothan or no Dothan...

 

[IntelUser2000]

Sandorski said:

"We'll have to wait and see, but don't forge that AMD has many cross-licensing agreements it can source for technology. Some of those technologies deal with low power Processors. AMD has been doing the impossible for the better part of a decade now, don't underestimate them. "

Umm, you are wrong. With Intel putting radical measures to protect the technology of Pentium M, I don't think Intel will give low power technology to AMD just like that. I mean you can also see in Itanium family that the technology of Itanium isn't being used(I mean other than the features of the EPIC architecture) for cross-licensing. I think it only applies to small-scale technologies like SSE, x86-64(they don't take much space in the die).

And this is the lowest power Pentium M(granted, this runs at 1.1GHz but still impressive): http://processorfinder.intel.c...Spd=ALL&CorSpd=ALL

5W at 1.1GHz for Ultra Low Voltage

Another: http://processorfinder.intel.c...Spd=ALL&CorSpd=ALL

10W at 1.4GHz for Low Voltage

And you can't ignore Intel's future products for Centrino like Yonah. Because Yonah is what AMD's future mobile chips will compete with, not Dothan. We don't know what tricks Yonah has up its sleeve, everyone thought Dothan was just Banias with doubled L2 cache, but its not, it has some other enhancements to further lower power consumption and increase performance.

Quote from Anandtech's Dothan review: "One by-product of on-time execution was that a number of architectural tweaks that the design team wanted to get into Banias had to be left out, a sacrifice made to preserve on-time execution. As soon as the Banias design was complete, those final architectural tweaks and enhancements that didn't make it into the first Pentium M incarnation were at the top of the list for its successor. As a result, Dothan is best viewed as a more polished evolution of Banias, with higher clocks and more cache made possible by Intel's 90nm process."

So, some mobile processor that's enhanced with leftover optimizations that couldn't be put in on previous generation is doing this well, imagine Yonah.

 

[Rhin0]

It should be obvious to everyone at this point that AMD has left over pieces of the terminator and are using it to create technology that is far ahead of our time. They are releasing it little by little!

 

[SickBeast]

I was a huge Centrino supporter until I spent some time working on Apple iBooks and Powerbooks. They have 6+ hour battery life, just like many Centrino systems. If Motorola can match intel in that regard, AMD surely can. I'm excited to see what they have to offer. A 64-bit chip with hypertransport will be a very interesting low-power mobile part. It's a good thing I haven't bought my mobile CAD workstation yet. 🙂

 

[SickBeast]

Originally posted by: Rhin0
It should be obvious to everyone at this point that AMD has left over pieces of the terminator and are using it to create technology that is far ahead of our time. They are releasing it little by little!

LOL

 

[sandorski]

Originally posted by: IntelUser2000
Sandorski said:

"We'll have to wait and see, but don't forge that AMD has many cross-licensing agreements it can source for technology. Some of those technologies deal with low power Processors. AMD has been doing the impossible for the better part of a decade now, don't underestimate them. "

Umm, you are wrong. With Intel putting radical measures to protect the technology of Pentium M, I don't think Intel will give low power technology to AMD just like that. I mean you can also see in Itanium family that the technology of Itanium isn't being used(I mean other than the features of the EPIC architecture) for cross-licensing. I think it only applies to small-scale technologies like SSE, x86-64(they don't take much space in the die).

And this is the lowest power Pentium M(granted, this runs at 1.1GHz but still impressive): http://processorfinder.intel.c...Spd=ALL&CorSpd=ALL

5W at 1.1GHz for Ultra Low Voltage

Another: http://processorfinder.intel.c...Spd=ALL&CorSpd=ALL

10W at 1.4GHz for Low Voltage

And you can't ignore Intel's future products for Centrino like Yonah. Because Yonah is what AMD's future mobile chips will compete with, not Dothan. We don't know what tricks Yonah has up its sleeve, everyone thought Dothan was just Banias with doubled L2 cache, but its not, it has some other enhancements to further lower power consumption and increase performance.

Quote from Anandtech's Dothan review: "One by-product of on-time execution was that a number of architectural tweaks that the design team wanted to get into Banias had to be left out, a sacrifice made to preserve on-time execution. As soon as the Banias design was complete, those final architectural tweaks and enhancements that didn't make it into the first Pentium M incarnation were at the top of the list for its successor. As a result, Dothan is best viewed as a more polished evolution of Banias, with higher clocks and more cache made possible by Intel's 90nm process."

So, some mobile processor that's enhanced with leftover optimizations that couldn't be put in on previous generation is doing this well, imagine Yonah.

I was thinking of cross-licensing agreements with Transmeta and not Intel in that regard. Intel is not the only one with great power saving features.

 

[IntelUser2000]

That's kinda different though. Transmeta's architectures have different architectures(that was weird saying it). Sure Intel is not the only one with power saving features, but they are not releasing vast majority of their technology either. It also took quite long for Transmeta to get both any performance and low power simultaneously. Transmeta's processors rely on software emulation of hardware to save power? So how can that be put to AMD's architecture? Maybe have a Transmeta variant processor? I don't know.

I think G4 and G5's are using IBM's Power architectures that are cut down from the server variants, not Motorola. Its not first time iBooks and Powerbooks had long battery life, but it was first time for PC, that's what makes Centrino different. I mean G4, G5 is different architecture from x86, so that may have to do with the power consumption. I guess that's the advantage of RISC architecture, good performance/low power consumption.

 

[VirtualLarry]

Originally posted by: JimRaynor
um, the p3 was superior to the athlon, and now only when athlon64 chips are here are the amd chips outperforming p4s. When it was athlon non xp vs p3 the results were neck and neck but the pentiums were narrowly better.

P3 superior to Athlon? When? Are you talking about the later model P3s with internal 256-bit cache busses and hardware prefetch, against early Athlons without? What about later Athlons? For that matter, why not put an Athlon XP TBredB or Barton against a PIII Katmai, with half-speed off-die L2 cache? If you discount the cache effects, the IPC of the architectures are similar, with the Athlon having the edge, due to superior FP capabilities. Integer is similar.

I find it interesting those questioning whether AMD can compete with the Pentium-M, when they already have the low-voltage Athlon family, and the Pentium-M is the only chip that Intel has with an IPC and bus and cache design that can remotely compete with it on a clock-for-clock basis. (The P4 has to clock significantly faster, and in doing so generates a lot of waste heat and draws a lot of power, so the P4 isn't even in competition here, Netburst is a CPU-evolutionary dinosaur.)

I would argue that in terms of high-IPC architecture, AMD actually has a lot *more* expertise than Intel.

 

[VirtualLarry]

Originally posted by: LikeLinus
I really think all the AMD fanatics are forgetting one crucial thing.

AMD started a mhz war that Intel answered with the P4. The P4 was built from the ground up to scale very high mhz wise, but because of the length/number of pipelines, it's not as efficient. The whole arcitecture was built for the war that AMD started. Then all the sudden AMD turned around and started their PR ratings bullsh!t because they had no chance of beating Intel at the fight they started.

"Mhz war" ... that AMD started? I think you need to re-check your history.

Even AMD's 486 chips, used a "PR" rating, it wasn't new. Look up the 5x86-133-PR90, it was an AMD clock-quadrupled "Enhanced" 486 design, with 16KB of write-back L1 cache. (Intel's 486's only had 8KB L1, and it wasn't until very late models that they had ones that supported write-back mode.) The 120Mhz model had a "P rating" of 75, because performance was supposed to be similar to a P54C 75Mhz, and the 133Mhz = P90, etc.

Originally posted by: LikeLinus
Remember this was back when mhz WAS the thing that most people (and they still do, except enthusiast like us) purchase based off.

Yes, but that was due to the *flood* of Intel advertising. AMD never started a "Mhz war" with Intel. Intel was always using that as a core marketing point/strength. Hence why both AMD and Cyrix were using P-equivalent ratings, even back in the 486/586 days, because they designed with better architectures in mind, not just pure Mhz. (Cyrix had some notably advanced features in their "MII" CPU, that Intel didn't have until later - support for a hardware call/return stack, and I think the first implementation of MTRRs. Or maybe that was the K6-2 with CXT core. Don't recall.)

Originally posted by: LikeLinus
If you really want to see how this will turn out, wait for the next totally new core and build up from the ground. AMD is still working off the XP platform and Intel is still working off the P4. It's the next round of chips that will really show you who's more advanced or not.

I don't know how you can say that, AMD is and has been selling a true eighth-generation x86-compatible CPU, that implements both 64-bit addressing modes/extensions, as well as on-die memory-controllers for better performance, and was dual-core ready from the start. AMD is not "still working off the XP platform", if anything, I think that they sincerely wish that they could put Socket-A completely behind them. Yet, Intel has hit an absolute thermal wall with Netburst, which was supposed to scale to 10Ghz (oops!), so instead they were forced to backtrack and produce a hybrid of their prior P3 architecture, with improved branch-prediction and a P4-like bus-interface.

Originally posted by: LikeLinus
As others have said, Intel has a real winner with the Mobile chips. Sure AMD might be able to make something that is decent and close, but the thing you forget is AMD is doing it how long since Intel started Centino? So who had the technology and know-how first. Intel was first in Mhz...and first in low power consumption high performance mobile chips. Thats just how it is.

I think that you are also sincerely forgetting Transmeta, who really fired the "low-power x86-compatible CPU shot heard round the world", with which Intel suddenly invented "SpeedStep" to compete. Intel was not the leader in that arena, and was certainly not the first. Ironic that you should have a username such that you do, and not remember this, considering who Transmeta hired. 😛

 

[VirtualLarry]

Originally posted by: Rhin0
It should be obvious to everyone at this point that AMD has left over pieces of the terminator and are using it to create technology that is far ahead of our time. They are releasing it little by little!

LOL. Sounds like that "area 51/roswell" drivel that ACC has spouted in the past - that the existance, or rather "discovery" of the transistor itself was due to a recovered crashed alien craft. Interesting if true, certainly.

Another little-known fact, both Intel and AMD were started by ex-Fairchild engineers. AMD used to produce Intel-designed CPUs, as a "second source" for them, since mfg's of that day demanded that. Their license ran out around the time of the AMD 486DX/2-66 release.

Oh, and just so that I get credit for calling it first - I wonder if the "AMD K-9", will be a real "dog" of a CPU. harharhar.. 😛

 

[IntelUser2000]

By the way, Pentium 4-M had worse battery life than 24.5W 1.6GHz Centrino and the Pentium 4-M had around 30W TDP and the chipset was made for mobile too.

Oh, I heard K9 is now K8 dual-core and K10 is K9.

 

[Markfw]

Don't put anything past AMD. I'm not saying it is, or it isn;t. I just know that big companies don;t allways win. Xerox bought out Tektronix to "buy the competition" with their 30% vs 3% market share in color network printers (I know I used to work for Tektronix for 20 years) Tek did what Xerox couldn't do. AMD may or may not be in the same boat, but they do have IBM on their side.

Just sit back and watch the fun over the next year.