AMD's Response to Intel Conroe: Energy Efficient Athlon 64 X2 CPU

[Defcon5]

Bang on there locutus12

 

[Biggles 266]

I don't dispute that the Core 2 Duo is king of the hill in terms of performance, but that so far hasn't had an impact on AMDs sales. They are still selling everything they can make. AMD are limited by capacity here. A few months back before Core 2 Duo arrived, AMD made a lot more money on each of their processors. Now their margins have been trimmed because they've needed to use price cuts to remain competitive. But they are still doing better than pretty much ever before.

Things like the energy efficient processors do a few things. Firstly, they cost more, so whilst the production costs for AMD are the same as the standard versions of their chips, they make bigger margins on them. Secondly, it also works in terms of advertising and keeping mindshare. If AMD appear to be releasing new products, well it makes them to still be competitive.

Another thing to consider is that Core 2 Duo makes up a rather small proportion of the market at the moment. Sure, enthusiasts are all over Core 2 Duo like a cheap hooker on a drunk stockbroker, but most computer sales are low end Dells etc. And they're still using Celerons. So if you walk into a nameless purple high street store and ask to buy the fastest PC they sell, you should still end up with an AMD.

AMD are not in trouble.

In the short term, AMD have a die shrink coming up. We should see the processors clock a bit further. To begin with, chances are that they'll enter the line up at similar ratings to what exists just now, but they should have more overclocking headroom. From talking to a few people who work in the CPU industry, we reckon that AMD will be able to go up to around 3.8-4 GHz stock speed on 65nm before they hit a wall. They managed 3 GHz on 90nm, although only a few of those Opterons will have been sold.

In the medium term, K8L is on the way. In terms of performance it should be at least on a par with Core 2 Duo, probably a bit ahead of it. Remember firstly that it's natively a quad-core part and you realise that even if clock-for-clock it's slower than Core 2 Duo, there is still two extra cores. Which means that it's going against Kentsfield really, not Core 2 Duo. The "problem" with Kentsfield is it's basically two Core 2 Duos on the same die, in the way that the Pentium D was basically two Prescott - K8L is a ground up quad-core part and will be a nicer design.

There's a whole bunch of things that AMD will do to K8 when making it into the K8L. First thing they will probably do is give it a 256 bit L2 cache. That'd be a big improvement to start with. I'd expect to see K8L based dual-core processors, possibly just quad-cores with some disabled. It'd improve yields as well. A lot of the K8 is inherited from the K7, so there are half a dozen other things AMD will change. Extra SSE units will help etc. Basically between architectural enhancements and higher clockspeed, K8L will bring AMD back into contention, probably giving them the performance crown again.

Long term there is an entirely new architecture in the works, K10. It'll likely be up against Core 3 Quad or whatever Intel decide to call their next one. Either way, AMD are not sitting around doing nothing.

What you young uns' seem to forget is that AMD and Intel used to trade the performance crown on an almost monthly basis in the past. Intel made some dubious choices, and AMD were ahead for nearly three years, they got lucky with that one. I say dubious choices, although given Intel's sales and profits in those years Netburst can't be called a failure. Now Intel are back on top, next year it'll be AMD. Year after might be Intel again...

It's not just about processors any more though. That's why AMD bought ATI, so they had their own chipsets to offer, to give the complete platform. And also of course it gave AMD some more capacity.

I'm glad Intel finally got their act together. It's meant that I can pick up a really high performance processor for a lot less cash. Now's a great time to be in the market.

 

[Kamakazie!]

I still remember the "." Digital Punk gave me when I first suggested (about Nov 2005 I think?) that Conroe would knock AMD off the top spot

What relevence does that have to anything?
If only we could all remember when someone posted speculation, got a sarcy remark because of it and then turned out to be right. "My Speculation turned out to be right!! Hooray!!".

Unless I'm missing something, "the P4's kept up with the A64's in some benchmarks" is also a true statement... which seems to negate most of the impact this statement was intended to have.

So on both sides, the arguement is pointless.

You would have thought so. But back then AMD were, like pretty much everyone in this very forum - bar the few that knew Conroe really was something special, were blinded by the success of the A64. Remember all those "duels" AMD kept offering up to Intel - effectively gloating in a very public way. They were too loved up with themselves, prancing around like a clown, to realise the sleeping giant's eyes were opening.

No, AMD were too busy taking huge margins out of Intels desktop and server market share. Their engineers have been working on improvements/a new architecture since probably even before the K8 core went into production.
To be honest, from some of your posts i would save the term "loved up" for something else...
As for "the few that knew conroe really was something special", what absolute rubbish. No one knew anything bar the Intel engineers and insiders. All you did was believe the speculation, which turned out to be true.

K8L will almost certainly "match" Conroe on performance. But it's still the aging K8 architecture at its core whose scalability is running out fast. Intel was initially planning to launch Conroe at much higher clock speeds but realised there was no need to and so decided that they'd save those higher clocks for a rainy day (such as the K8L launch day )

K8L is significantly different to the K8 architecture, it is far more than just a refresh and as such its scalability is relatively unknown.

 

[locutus12]

Kamakazie!, well said.

 

[NathanE]

snip

There was a bit more to it than just guessing that Conroe would be good...

What's this 'loved up' remark about?

K8L is significantly different to the K8 architecture

And yet, AMD still refer to it as K8L? I'm sure it'll be great... as I said it should easily match Conroe - possibly exceed it considering the IMC... but my point that you missed is that Intel has plenty of scalability in Conroe just waiting to be unlocked.

As for "the few that knew conroe really was something special", what absolute rubbish

No... there really was a handful of people here that supported Conroe from the moment they knew it was a P6 derived design.

So on both sides, the arguement is pointless.

Sides? It was just intended to highlight the irrelevance of the original statement.

Kamakazie!, well said.

 

[pcoltrane]

And yet, AMD still refer to it as K8L?

I suspect that has more to do with AMD not wanting the new chip to be a "dog" of a chip .... K9 .... than anything else.

 

[Digital Punk]

Hey thanks Kamakazie!

I couldn't have put it better myself, and you saved me the job.

Nathan's been on my ignore list for about 6 months, so I don't get to see his 'posts' anyway until you quoted him!
(Oops there's another rolly for him to remember for a few months)

 

[pegasus1]

Well lets not forget that Intel have produced great chips in the past only to go on and ruin them, lets hope they dont 'tinker' with the Conroe too much.

 

[Kamakazie!]

And yet, AMD still refer to it as K8L? I'm sure it'll be great... as I said it should easily match Conroe - possibly exceed it considering the IMC... but my point that you missed is that Intel has plenty of scalability in Conroe just waiting to be unlocked.

Take a look at the die pics. It's a very significant change. It is called K8L (L being the roman numeral for 50 apparently) meaning K8.50. It has a lot of changes similar to Intel made going from P-6 to NGMA. Out of Order execution, increase is FP units etc.
I didn't miss that point at all, i just chose not to comment on it as there is nothing else to say. Yes, well done, there is a fair amount of scalability left in the Conroe.

No... there really was a handful of people here that supported Conroe from the moment they knew it was a P6 derived design.

There's the difference, people supported the conroe. None of you knew anything. That was my point. Your comments inply that your views should hold more weight than others simply because your swallowing of the Intel hype turned out to be the correct decision. Though i would still say its better just to wait and see.

I think that K8L is going to be very very fast, i still don't know it though. It might fall short of Core2Duo or it might streak far ahead. Either way it will be better than whats currently offered by AMD (not that K8 is in anyway bad).

 

[Lewcifer]

I suspect that has more to do with AMD not wanting the new chip to be a "dog" of a chip .... K9 .... than anything else.

Roffle

 

[locutus12]

And yet, AMD still refer to it as K8L?

its being called K8L to avoid the rather negative marketing that comes with calling it K9, you can find this out for your self if you go read the AMD site or indeed wikipedia. as for the rest of your previous post, im afraid it came across as someone who was a little uninformed.

no one was sitting on their hands, AMD have a good solid architecture for desktops so that side of things has been left to itself whilst AMD stole massive ammounts of market share from intel for its servers and workstations, which in turn gives them the revenue needed to produce better CPU solutions all round in the future.

 

[NathanE]

I've made the K9 point before in another thread. They could have called it K10 - although that's reserved for their next architecture apparently. The fact they stuck with K8 plus an L suggests it is iterative improvements rather than anything substantial. But then, as has been said, K8 is a solid architecture and may only need a few IPC improvements to bring its scalability back up to par.

It has a lot of changes similar to Intel made going from P-6 to NGMA. Out of Order execution, increase is FP units etc.

Yup, hence why it's quite possibly going to match or exceed Conroe.

no one was sitting on their hands

AMD have. If they hadn't then they would have had something to counter Conroe with by now. 4x4 doesn't count as it's not comparing like for like. It's no different to how Intel were sitting on their hands letting their product do the talking in the Northwood days. Turns out they sat on their hands a little too long (read: Prescott) which let the A64 bite the market.

Your comments inply that your views should hold more weight than others simply because your swallowing of the Intel hype turned out to be the correct decision

And as I keep telling you... there was a little more to it than just "swallowing Intel hype". FYI Intel didn't actually acknowledge the existence of Conroe until IDF earlier in the Spring of this year. So any hype before that was solely the doing of the technology media.

 

[Sparky]

Ive seen the cores that Intel have on their roadmap and it does look impressive.

Find it very hard to believe that AMD wouldn't have been developing anything to counteract Conroe.Nothing in business is a secret for long.Isuppose we shall all just have to wait and see.

One thing though,when AMD release a new core/cpu most of the time its not a paper release.Intel may have won the enthusiast market over with the pure speed of Conroe but their bungled release "24th July" was a farse.

 

[Biggles 266]

no one was sitting on their hands

AMD have. If they hadn't then they would have had something to counter Conroe with by now. 4x4 doesn't count as it's not comparing like for like. It's no different to how Intel were sitting on their hands letting their product do the talking in the Northwood days. Turns out they sat on their hands a little too long (read: Prescott) which let the A64 bite the market.

I don't think you fully understand the time it takes to develop a new processor. Conroe first appeared on Intel roadmaps back in early 2004 - that's two and a half years ago. They were working on it before it appeared on the roadmaps at that, meaning they began work on it round about the time the first Athlon 64s appeared.

AMD have spent less than three years extending and enhancing the K8 architecture. In that time they've extended it from single to dual channel memory access, added SSE3, gone through one process shrink (with another in the works), oh and they've also made the first proper dual-core X86 processor. Intel in contrast have spent five and a half years extending Netburst. So really AMD have spent only as long as you would expect on K8. It's not like they've been keeping alive an architecture which is seriously old and uncompetitive.

Sure, from AMDs perspective it would be nice to have K8L ready right now. But it's not because it wasn't scheduled to appear until K8 had been fully extended, which it hasn't yet. But they have been working on it. And K10 at that. AMD are still in fine shape.

AMD have done plenty in the past couple years, and they have more to come. Just because Intel have the performance crown again doesn't mean that AMD are doomed or have failed in any way.

 

[m3csl2004]

I suspect that has more to do with AMD not wanting the new chip to be a "dog" of a chip .... K9 .... than anything else.

iirc there going straight to k10 lol

 

[rayb74]

Sell your Conroes guys !!

The X2`s now run a bit cooler !

 

[aftershxck]

Sell your Conroes guys !!

The X2`s now run a bit cooler !

that's what i was thinking, whoop-de-doo for AMD in the face of conroe