What is AMD's next move?

[Aenslead]

AMD will have something when Intel actually gets their act together. They always have.

Recent example: Pentium D, then, a couple of weeks later, Athlon X2, which is best overall. So we can expect the very same once again.

As shown by Anandtech, memory bandwith does matter for the X2 lineup, so DDR2 will be just fine;

BTW... what does "QFT" mean?

 

[Intelia]

LOL

 

[LT4CAMSS]

Originally posted by: Aenslead
AMD will have something when Intel actually gets their act together. They always have.

Recent example: Pentium D, then, a couple of weeks later, Athlon X2, which is best overall. So we can expect the very same once again.

As shown by Anandtech, memory bandwith does matter for the X2 lineup, so DDR2 will be just fine;

BTW... what does "QFT" mean?

Quoted For Truth

 

[imported_Ged]

Originally posted by: Brian23

Originally posted by: n7
Blah blah blah!

Link to where i can buy a Pentium 5...oh wait, it's not out yet!

Just like M2 isn't out yet...gawd i hate this constant speculation :roll:

but it's fun!

seriously though, AMD has to have plans currently underway. If they started now, there's no way they could design a new, more competitive chip by the time PV comes out next year. They had to be anticipating this move by intel. They knew eventually intel would start making a competitive cpu.

Exactly. If you think AMD has been "sitting on their hands" all this time, you're wrong.

It will be very interesting. AMD stated they would not be going wider than their current design, which has actually been around since before K8 (3 wide), while Intel has gone 4 wide.

M2 will be nice, but I am looking forward to Socket F (1207 pins).

 

[Oemo]

Originally posted by: Brian23

Originally posted by: Duvie

Originally posted by: Brian23
being smaller does not help AMD. It took them quite a while to come out with the Athlon after they bought the Alpha group. They haven't really changed the chip since the original K7. The've just been adding to it slowly. Designing a whole new architecture takes years.

How do you figure this genius??? They just cancelled the tejas architecture about 3Q of last year??? It was a radically different approach then this. Fact is they dont take as long as you think. The fact is many of the chips are made up of several ideas and technologies that are and have been being tested for years.....I dont think they have had as much time in this as you want to believe...

Smithfield was thrown together in world record time.

First of all, I'm not an Intel fanboi. I only said that the chip looks to be "killer" because it's something new and the specs look good. I agree that there is no proof that it's a good chip until Anand puts up the benchies.

The reason why I say it takes years to develop an architecture is because it really does. Look at how long they milk each architecture before coming out with a new one. Incremental updates do not count as a new architecture. What they do is add a little more FSB here, lower latency there, tweak this, add a little more SSE there. Those tricks improve performance in the shorter run, and in the mean time they're working on new architectures behind closed doors. The Athlon X2 is basically the same chip as the first slot A chip, it just has a lot of add-ons like L2 cache, memory controller, Hyper Transport, tweaked pipeline etc. As far as intel goes, they used the same architecture from the original pentium pro all the way up through the P3. Of course they switched when they came out with the P4 and the PM, but they were working on those for years while people were still using Ppro, P2, and P3 processors.

The reason why major redesigns are so few and far between is because of one thing, $$$. It's in both AMD and Intel's interest to slowly tweak their chips because it gets us to buy more. How long it takes to develop something new who can say, but I'm sure either company could come out with something fairly quick that would totally blow away what we've seen to date, if they had to. There's just little reason for them to do so now, what does the typical computer user do that really stresses out the hardware of today? I'm not saying that this is necessarily fair, I love to see crazy new tech, it's just business.

 

[Topweasel]

For people who think that the K8 isn't a new design or that PM is just a P3 in drag and that every new design has to be completely different YOU ARE WRONG. Do auto manufacturers build cars with 6 wheels 2 in the center just to make sure they don't look to much like the car that it was replacing. Isn't funny how every new car looks like every other car when its released. You don't throw away well designed parts just because you are making something new. The things that AMD changed on the K8 are huge, who cares if they felt they didn't need to change the FPU very much, They completely changed the bus it uses, incorporated a very difficult memory controller onto the die, designed it as a 64bit CPU, this is just the tip of the iceburg but are the easiest to identify changes and any one of these would qualify a new generation in architecture. The P-M is almost completely a new dog remember when the P3 ran into the same problems as the P4 is running into now. They kept the Design going have tweaked it to the point where they are only 600MHz-800MHz behind AMDs only CPU to worry about, I would keep refering to it as a tweak on the P3.

So I say again is a Car considered not to be a new car even if it uses the same engine and nothing else from another car.

 

[Keysplayr]

Originally posted by: n7
Blah blah blah!

Link to where i can buy a Pentium 5...oh wait, it's not out yet!

Just like M2 isn't out yet...gawd i hate this constant speculation :roll:

Obviously, a great deal more people than not, love it. Else these forums would dwindle away to just a newsgroup.

 

[ZobarStyl]

Originally posted by: Avalon
Well, if Intel beats out AMD in the performance wing next round, we might see AXP prices from AMD again.

QFT - Even if Intel manages to pull a magic rabbit out of it's hat this time, they still never compete on a price/performance basis. I never understood the Intel strategy of basically giving their chips to OEM's for free then charging out the nose for their retail boxed parts.

 

[clarkey01]

Originally posted by: Bona Fide

Originally posted by: AMDCrazy
Nothing can beat AMD...Period!!!!

Fanboyism is at an all-time high, I see.

Well if M2 is the best that AMD can do, Intel is going to win this round. The real spectacle will be Intel's answer to K10.

M2's a platform not a new chp. Jesus. We have dual core K8's at the moment and the K9's are due in 2 years time, so just after Intel's second coming of crhist.

 

[Viditor]

Originally posted by: Duvie
Again....They just cancelled Tejas not even a year ago so where did this radical design come from??? They probably had it on the backburner simmering but it was clear Intels cpu roadmap did not include a cpu like this just 1 year ago....A bunch of technologies are played with like a black ops program but may have actuakllky come together as quickly as under 1 year. They probably had already focused their attention on this before the Tejas announcement, but it is not the multiple years you are alluding to....

I believe that part of the Tejas cancellation announcement was that they were accelerating their already existing Dothan based projects...
In general, it takes 5 years to go from concept design to finished product on CPUs...

 

[Hacp]

AMD says their keeping K8 for 2006. Looks like they will be relying on increased clock speeds from 65 NM and also DDR2 to help provide the small performance boost they think will be necessary to compete with Intel's early offerings of their new architecture.

 

[Viditor]

Originally posted by: Brian23
PV looks to be a killer chip. What does AMD have in store to battle it?

To be fair, so did Itanium and so did Netburst...even Prescott looked to be killer!
The truth is that we just won't know till it gets here at the end of next year...
As to AMD's response, take a close look at the AMD roadmap...see how blank it is for next year's end? IMHO, AMD is sandbagging Intel at the moment...but we shall all have to wait and see, cause while it's fun, it really doesn't mean much.

 

[Diasper]

Re-examining M2 I have some faith for the design. I'm not expecting it to eb revolutionary performance over S939. Certainly, less than 10% and probably more like 5% difference (this is just guesstimate speculation). As someone said the performance scale will be more akin to S754 to S939 albeit I think a little more significant at least certainly for dual-core chips.

What I believe is that Intel's chips will be good but not AMD killers. I don't think we'll be seeing +10% over current AMD chips except in certain fields (eg like Intel did with content creation for a long time).

Reasons:
While Intel's next chip will be with four executing units etc it is hard to keep those filled and leads to lower inefficiency per pipe. Now while Intel will no doubt have an improved branch prediction and it's shorter pipe length will all enable significantly greater performance than current P4, bar the extra pipe this is all what we currently see and what AMD will be refining for M2.

What Intel won't have compared to AMD is they still won't be having hyper transport meaning slower access to the memory. Now with Intel pursuing a policy of using large super fast cache they can limit the performance disadvantage somewhat. Nevertheless, hypertransport will boost AMD's current performance. Moreover, AMD current S939 using DDR can have lower memory timings given it another comparative performance boost compared to Intel.

Moreover, in dual-core while sharing the cache is a good move AMD's implementation with more direct core-to-core communication is superior. Again, it seems another move to catch up with AMD as much as possible compared to P4 but again this will be where AMD will have an architecural performance advantage. Also, with AMD moving to DDR2 667 the chips will benefit with it compared to the current DDR400 and so we may end up seeing upto a 10% performance jump in heavily loaded, memory intensive tasks. In all, AMD will be competetive in dual-core yet.

My prediction, is that Intel's single core chips will exceed current S939 performance but not by much. In mhz to mhz they will be competetive as remember Intel will have a slightly higher pipeline (2 extra) perhaps lowering the IPC value ever so slightly while allowing to clock slightly higher than AMD. Either way I don't think we'll be seeing massive performance advantages over current AMD (except maybe media encoding). Faster yes, but not by much. In dual-core the situation will be less clear. For those operating dual-core with faster RAM speeds eg DDR500, AMD will be very competetive (except perhaps encoding again depending on Intel's single core performance) - or at least more competetive than AMD's single core.

Again all these scenarios are where some extra mhz couldn't make up.

As for M2, where we'll see advantages is that the core will again be slightly revised and tweaked. This means we'll see some extra % increase out of it. What will be the bigger winner out of the move will be AMDs dual-core as they see some benefit out of the extra bandwidth while DDR2 timings and cost are coming down. AMD single core M2 I think we'll see rather similiar performance to current S939 ie within a few % given tighter RAM timings helping to make up for some of revision improvements.

What AMD could do is implementaer much imrpoved cache - both in latency and in bandwidth as for ages now AMD has been using far slower caches than Intel. While AMD having hypertransport will not allow them to find the same performances than Intel see from using it, it will still offer them some increased performance. If anything this is perhaps my greatest hope for M2 and if they do implement it I think we'll see very comparative performance.

In all, I think motherboards will be much more significant in the future. Performance will be more similiar and so features will matter more. However, also as we have seen with the new ATI (and ULI) chipsets for AMD, they have managed to get significant performance increases over current nforce4 boards of upto 5% over the fastest (refer to Anandtechs article on Saphires board) and so in getting the absolute performance potential of either chip, the maturity and efficiency of the motherboard will be significant. In this, given Intel designing their own they might have the edge.

 

[TuxDave]

Originally posted by: Aenslead
AMD will have something when Intel actually gets their act together. They always have.

Recent example: Pentium D, then, a couple of weeks later, Athlon X2, which is best overall. So we can expect the very same once again.

As shown by Anandtech, memory bandwith does matter for the X2 lineup, so DDR2 will be just fine;

BTW... what does "QFT" mean?

I dunno... but I don't consider the Pentium D a good example of getting your act together.

 

[Bona Fide]

Originally posted by: Aenslead
AMD will have something when Intel actually gets their act together. They always have.

Recent example: Pentium D, then, a couple of weeks later, Athlon X2, which is best overall. So we can expect the very same once again.

As shown by Anandtech, memory bandwith does matter for the X2 lineup, so DDR2 will be just fine;

BTW... what does "QFT" mean?

I should correct what you wrote.

"As shown by Anandtech memory bandwidth does not matter for the majority of single and multi-threaded applications." So, DDR2 will not help too much unless it's a straight jump to DDR2-800. And then you have to factor in the problem of high latencies at such high frequencies.

Also, Intel admitted that the Pentium D was a rush job, just to beat AMD to the dual-core desktop. AMD took the time to think out its K9 structure, hence their chips beat the Pentium D like it was nothing. But hopefully Intel is taking their time on the new chip. Remember back when Intel dominated with the early Pentium line? I'm talking Pentium II and III. Not all that new stuff. Actually, the Northwood cores were pretty good as well.

QFT - Quoted for Truth

 

[Fox5]

Originally posted by: Topweasel
For people who think that the K8 isn't a new design or that PM is just a P3 in drag and that every new design has to be completely different YOU ARE WRONG. Do auto manufacturers build cars with 6 wheels 2 in the center just to make sure they don't look to much like the car that it was replacing. Isn't funny how every new car looks like every other car when its released. You don't throw away well designed parts just because you are making something new. The things that AMD changed on the K8 are huge, who cares if they felt they didn't need to change the FPU very much, They completely changed the bus it uses, incorporated a very difficult memory controller onto the die, designed it as a 64bit CPU, this is just the tip of the iceburg but are the easiest to identify changes and any one of these would qualify a new generation in architecture. The P-M is almost completely a new dog remember when the P3 ran into the same problems as the P4 is running into now. They kept the Design going have tweaked it to the point where they are only 600MHz-800MHz behind AMDs only CPU to worry about, I would keep refering to it as a tweak on the P3.

So I say again is a Car considered not to be a new car even if it uses the same engine and nothing else from another car.

Only 600-800mhz behind? That's quite a bit....

Anyhow, I'd say K8's primary design focus seems to be multicore support and performancee, with better single core performance just coming about because the integrated memory controller works well for a single cpu, but works even better in multicpu configurations.(and why intel initially plans to put it on their server cpus)

PM's focus was obviously maximum general purpose performance with the lowest amount of heat generated/power used, and to do so Intel really worked over the cache. The faster memory bus isn't as important in what they did to the design.

AMD's K9 will be further the multicore idea, but will have a focus on enhancing single core performance as well. It's possible Intel could catch up or surpass AMD during this time, but the K10's introduction isn't supposed to be long after.(though I wouldn't be too surprised to see K10 scrapped entirely for a new design if it turns out to be unworkable, I don't think K10 was planned to be a multicore monster, yet that's where the industry is heading..but since K8 was designed with multicore in mind, I can't imagine that they would go away from that with K10)

Also, Intel admitted that the Pentium D was a rush job, just to beat AMD to the dual-core desktop. AMD took the time to think out its K9 structure, hence their chips beat the Pentium D like it was nothing. But hopefully Intel is taking their time on the new chip. Remember back when Intel dominated with the early Pentium line? I'm talking Pentium II and III. Not all that new stuff. Actually, the Northwood cores were pretty good as well.

The X2s aren't K9, K9 isn't out yet.
And let's see...
Intel's P1 and P2 dominated(though K6-II with 3dnow could catch up, and K6-III was many times the fastest performing chip available), but athlon beat P3 pretty much start to finish. Initial P3s were worse in mhz per mhz performance, later ones were better, but they never caught up to Athlon in clock speed, and after AMD also integrated the cache on die and finally had DDR memory support, the Athlon had a healthy advantage over P3 in mhz per mhz performance and top speed. And it took 3 revisions of Northwood(up to northwood C) before it finally had a decisive advantage over Athlon XP.(the AXP 3200+ actually does beat the 3.06ghz P4 B in many tests, but AMD's Athlon 64s were like a year late)

 

[BitByBit]

Firstly, I highly doubt Conroe will be called the 'Pentium 5' - there's just too much irony there.
Secondly, Conroe isn't out yet: it was unveiled yesterday, not released.

It's very tempting to speak of AMD's impending doom in the face of Intel's seemingly impressive new architecture. But I'm sure AMD's executives are quite tired of hearing the same thing whenever Intel releases details of a new architecture.

Don't assume AMD has no response to Conroe in the works. The K8 will either be retired by the time it is released, or will be heading that way. We've had a brief glimpse into future processor architectures, but how about next year?
2006 will see high-clocked K8's going against Presler and subsequently Yonah. I can foresee AMD holding onto the performance crown for most of 2006, provided it can scale past 3.0GHz, which looks plausible.
Given AMD's traditional secrecy, we can only speculate on K10's architectural details, although Anand's interview with Fref Webber told us that the Athlon's triple-issue width is as wide as it will go.
Other than that, K10 is vague at best, but we can be sure it will atleast be a match for Conroe.

 

[OneOfTheseDays]

look i don't honestly think AMD is dumb enough to just sit on there hands here. i expect both companies to be pretty much neck and neck in 06'.

 

[bunnyfubbles]

Its probably been said before, but PV isn't something AMD has to step up to, the PV is Intel stepping up to AMD to match the A64/X2. However they'll need something better in terms of marketing, because now intel can brag they're just as fast

 

[Hacp]

Yes, AMD has no new arcitecture for 2006 as I previously stated. They are relying on DDR2, AND the 65nm process will probably allow AMD to take their chips to 3.2 GHZ or 3.4 if they're lucky. Performance SHOULD be similar compared to Intel's chips. Intel would probably win out since their chips are based on a new design, but AMD will still be ableto compete, and if it prices things right, will still be able to do OK. 2007 is when AMD will probably change to a new platform and architecture.

 

[Diasper]

Single core = Intel with a slight win but <10% (except maybe encoding?)

Dual-core = Neck and neck or even AMD with a slight edge (except maybe encoding again?)

IF AMD implements a better CPU-cache system including just faster cache we'll see a much more interesting scenario emerge.

 

[Diasper]

One thing if Intel = DRM forced apon you which it does (unlike AMD) I don't care if it's 50% faster.

 

[Hacp]

WTF? AMD has DRM, and I don't think Intel will force anything on its consumers.

 

[Duvie]

Originally posted by: Diasper
Single core = Intel with a slight win but <10% (except maybe encoding?)

Dual-core = Neck and neck or even AMD with a slight edge (except maybe encoding again?)

IF AMD implements a better CPU-cache system including just faster cache we'll see a much more interesting scenario emerge.

OK someone is smoking something...


Single Core = pretty much a tie....AMD is much better at gaming for liked price chips but behind 10% in encoding and such. AMD is faster inmath apps but behind in CAD apps that can take advanatge of HT...

Dual Core = not even close....Similar priced cpus are anywhere from 20-30%+ over Intel and even in multimedia....Wake up!!!

Since I seem to prove what I state start pulling some links to prove what you stated...

 

[meson2000]

I agree with Hacp. That is what I have been reading on the net too. 2006 will just bring DDR2 and the 65nm process. If AMD releases the 3ghz FX 59 on the 90nm process in Qtr 1 2006, then I think they will be around 3.4 by the end of 2006/1st Qtr 2007. No significant changes to the core in 2006. I think the reason why no one has heard anything about K10 (K9 has been skipped to bring in dual core K8's) is because I think AMD's new push is just adding more cores to the CPU. 2007 will bring a quad core Athlon with probably a shared L3 cache and maybe (but highly unlikely) a next gen DDR3 memory controller. I really think that AMD thinks they have found the sweet spot with the K7-K8 core and are only going to make minor tweaks going forward and just increase clock speed and add more cores. I seriously doubt the 'mythical' K10 will be a major redesign considering the small amount of changes they made to the core going from K7 to K8.