What will AMD do between now and bulldozer?

[Idontcare]

Originally posted by: aigomorla

Originally posted by: Idontcare

So why do we have 4GHz 150W Nehalems instead of 0.1GHz 1kW Nehalems?

i thought the .1ghz 1kw nehalems were called itaniums.

Pretty sure you are getting that confused with larrabeast now

Originally posted by: Soleron
Well, AMD's technology development is behind Intel's for sure, that's not a technical decision. That's likely because they have less capital to throw at it, and less volume to make up the capital costs.

What I'm trying to say is that AMD could launch a 32nm chip now. It would have sub-10% yields, be leaky, poorly-clocking and all-around inferior to 45nm, and it would cost several times as much per good die to produce. But it could be done. So there must be a point at which it goes from being uneconomic to do a 32nm chip to a good decision. And I think when GF talks about Q3 2010 as "risk production" and "volume ramp" they are referring to a technical position, which is unlikely to be when retail chips are viable. Look at when TSMC announces processes - 40nm was "ready" in late 2008, but the first retail chips in any quantity are in September this year.

I'll make another point against 32nm Deneb, which is that every dollar and man-hour you spend on it is less to spend on Bulldozer. And Bulldozer better be perfect and on time.

...we should know for certain either way on the November Analyst Day.

--

That said, I have no insight on process technology beyond what tech reviewers, news reports, and JF say. If you know that AMD's process technology can launch 32nm Denebs profitably in Q3 '10 then I'd love to know about it.

Soleron your points are valid, I hope you aren't getting the feeling that I am in violent disagreement with them or some such.

I've long held the perspective that a 32nm Deneb shrink would just be a needless distraction of AMD's already diminished R&D resources and that they needed BD sooner than later. But look at what they did at 65nm, the last time they introduced a major architecture change, they dragged the Athlon X2 into 65nm a year before Phenom came out.

BD is set to make the same 1yr stagger release, so the question is why would AMD not plan for a Plan B by shrinking Deneb/Thuban? I had ruled it out as an unlikely drain on R&D resources until I heard them confirm they did shrink the Deneb core logic for Fusion...well if they shrunk it for Fusion and got the layout and everything tweaked then it is truly a minimal cost adder at that point to hack off the GPU in the layout and just spin the reticles for a stand-alone 32nm Deneb shrink.

At this point I'm not trying to argue that they should, I am arguing that I can't rationalize how they could justify not going that one extra last step at minimal cost and by doing so have a backup plan to any snafu's or delays in 2011 with BD.

If we think AMD is in for some rough times in 2010 with their 45nm Stars Core based chips fighting off Intel's Westmere based 32nm chips then what do we think 2011 is going to be like?

Absolutely implausible to me that AMD would expect their shareholders to find it OK that AMD's decision makers decided 45nm K10.5 cores would be good enough for 1H 2011. Those decision makers have something planned for 1H 2011, and if its not BD then it only stands to reason that it will be a deneb shrink with that fusion IGP lopped off (or perhaps just disabled if it isn't too big of a diesize hit).

 

[Kuzi]

Good points IDC. I mean if Bulldozer was only a few a few quarters away from release, then I'd say AMD would be better of to just concentrate on releasing bulldozer and it's derivatives and skip a K10.5 shrink. But as you say AMD needs a plan B especially if Bulldozer is that late (2H 2011).

 

[Soleron]

I agree. There's too many variables to be certain, the main ones being when 32nm will be launchable and when Bulldozer arrives. If the former is Q3 '10 and the latter is Q3 '11 then it would be a good idea; if the former is Q1 '11 and the latter is Q2 '11 then it isn't.

I don't see Westmere as a threat though. Look at what each model is up against - which one would you pick?

2.8GHz/3M ($87) vs. X3 435 [2.9GHz] ($87)
2.93GHz/4M/HT ($123) vs. X4 630 [2.8GHz] ($122)
2.93GHz/4M/HT ($143) [no matchup]
3.2GHz/4M/HT/Turbo ($176) vs. X4 945 [3.0GHz] ($169)
3.33GHz/4M/HT/Turbo ($196) vs. X4 965 BE [3.4GHz] ($195)
3.46GHz/4M/HT/Turbo ($284) [no matchup]

 

[richierich1212]

Originally posted by: Soleron
I agree. There's too many variables to be certain, the main ones being when 32nm will be launchable and when Bulldozer arrives. If the former is Q3 '10 and the latter is Q3 '11 then it would be a good idea; if the former is Q1 '11 and the latter is Q2 '11 then it isn't.

I don't see Westmere as a threat though. Look at what each model is up against - which one would you pick?

2.8GHz/3M ($87) vs. X3 435 [2.9GHz] ($87)
2.93GHz/4M/HT ($123) vs. X4 630 [2.8GHz] ($122)
2.93GHz/4M/HT ($143) [no matchup]
3.2GHz/4M/HT/Turbo ($176) vs. X4 945 [3.0GHz] ($169)
3.33GHz/4M/HT/Turbo ($196) vs. X4 965 BE [3.4GHz] ($195)
3.46GHz/4M/HT/Turbo ($284) [no matchup]

AMD is pretty much screwed if it comes down to these choices.

 

[Dark4ng3l]

The problem right now is that no matter what AMD comes out with Intel is far enough ahead that they are sitting on "untapped potential" that they can easily counter anything AMD comes to market with. At this rate even if AMD starts catching up today it would take years to get to where Intel is.

 

[Idontcare]

Originally posted by: Dark4ng3l
The problem right now is that no matter what AMD comes out with Intel is far enough ahead that they are sitting on "untapped potential" that they can easily counter anything AMD comes to market with. At this rate even if AMD starts catching up today it would take years to get to where Intel is.

This is true, all that overclocking headroom on the Intel chips exists for a reason, the process technology is churning out chips that could be binned for stock clocks at much higher clockspeed SKU's than the current ones but Intel has no financial incentive to do that at this time as it would result in marginally lower parametric yields.

 

[Concillian]

It seems Intel is happy to give the bottom end of the market to AMD. They could cut the legs out from AMD by pricing their low end cheaper, bringing the E5xxx series into Celeron prices and their E6 and E7 series to E5 series prices, etc...

However doing so would serve no purpose and would likely lose money. Lots of people buy those procecssors even with AMD being price and performance competitive (in that segment of the market). They'd be cutting the legs off AMD, but they'd also cut the legs off their own profit. You have to remember that the bulk of sales are through Compaq, Dell, and such and the bulk of that is to businesses who mostly aren't considering AMD at all. If they drop prices on their low end, they pick up a few retail consumers who are choosing AMD now, at the expense of losing profit on the millions of PCs that are sold without concern for their competition.

Remember when AMD had a clearly superior part in terms of price, performance and performance per watt... yeah, that time when Intel was still pricing their processors higher than AMD and still making profit? AMD undercuts Intel because it has to. There's no reason for Intel to undercut AMD or even compete at price parity in the low-end or mid-end market.

 

[cbn]

Originally posted by: Idontcare

Originally posted by: Dark4ng3l
The problem right now is that no matter what AMD comes out with Intel is far enough ahead that they are sitting on "untapped potential" that they can easily counter anything AMD comes to market with. At this rate even if AMD starts catching up today it would take years to get to where Intel is.

This is true, all that overclocking headroom on the Intel chips exists for a reason, the process technology is churning out chips that could be binned for stock clocks at much higher clockspeed SKU's than the current ones but Intel has no financial incentive to do that at this time as it would result in marginally lower parametric yields.

Yep, I have noticed comparing stock speed AMDs to stock speed Intels for the purposes of determining "value" is somewhat misleading if OCing of any sort is planned. (re: Intels have much more OC potential)

 

[cbn]

Originally posted by: Concillian
It seems Intel is happy to give the bottom end of the market to AMD. They could cut the legs out from AMD by pricing their low end cheaper, bringing the E5xxx series into Celeron prices and their E6 and E7 series to E5 series prices, etc...

However doing so would serve no purpose and would likely lose money. Lots of people buy those procecssors even with AMD being price and performance competitive (in that segment of the market). They'd be cutting the legs off AMD, but they'd also cut the legs off their own profit. You have to remember that the bulk of sales are through Compaq, Dell, and such and the bulk of that is to businesses who mostly aren't considering AMD at all. If they drop prices on their low end, they pick up a few retail consumers who are choosing AMD now, at the expense of losing profit on the millions of PCs that are sold without concern for their competition.

Remember when AMD had a clearly superior part in terms of price, performance and performance per watt... yeah, that time when Intel was still pricing their processors higher than AMD and still making profit? AMD undercuts Intel because it has to. There's no reason for Intel to undercut AMD or even compete at price parity in the low-end or mid-end market.

Good point. I was wondering why Intel didn't lower prices more on the dual cores? In fact, Intel's die size for dual core is smaller than AMDs ATHLON X2 die size so they definitely have the room to do it.

More efficient architechture for AMD in the future?

 

[nyker96]

Originally posted by: Just learning

Originally posted by: Concillian
It seems Intel is happy to give the bottom end of the market to AMD. They could cut the legs out from AMD by pricing their low end cheaper, bringing the E5xxx series into Celeron prices and their E6 and E7 series to E5 series prices, etc...

However doing so would serve no purpose and would likely lose money. Lots of people buy those procecssors even with AMD being price and performance competitive (in that segment of the market). They'd be cutting the legs off AMD, but they'd also cut the legs off their own profit. You have to remember that the bulk of sales are through Compaq, Dell, and such and the bulk of that is to businesses who mostly aren't considering AMD at all. If they drop prices on their low end, they pick up a few retail consumers who are choosing AMD now, at the expense of losing profit on the millions of PCs that are sold without concern for their competition.

Remember when AMD had a clearly superior part in terms of price, performance and performance per watt... yeah, that time when Intel was still pricing their processors higher than AMD and still making profit? AMD undercuts Intel because it has to. There's no reason for Intel to undercut AMD or even compete at price parity in the low-end or mid-end market.

Good point. I was wondering why Intel didn't lower prices more on the dual cores? In fact, Intel's die size for dual core is smaller than AMDs ATHLON X2 die size so they definitely have the room to do it.

More efficient architechture for AMD in the future?

I was wondering about the same thing, why Intel would just leave the low end market completely to amd. they are currently using dual cores against even quads. but I think for the sake of competition Intel may just want to keep the antitrust guys off their backs by giving amd the low end market that doesn't have too much profit margin anyways and just concentrate on mid to higher end. but i wouldn't rule out the possibility of them introducing a really cut down quad to compete in the sub 150 range for 1156 next year if they feeling the need to.

 

[IlllI]

the low end has less room for profit though, right?

 

[IntelUser2000]

Originally posted by: Just learning
Yep, I have noticed comparing stock speed AMDs to stock speed Intels for the purposes of determining "value" is somewhat misleading if OCing of any sort is planned. (re: Intels have much more OC potential)

Concerning clock speeds, lots of presentations indicate clock speeds won't increase much in the future, with yearly gains being around ~8%, at least for Intel processors. They'll be TDP-limited anyway.

If you look beyond air-cooling, and compare exotic cooling, AMD isn't behind at all for OC headroom. No Core 2's or Core i7's can reach 7GHz that the Deneb's can reach, except the Pentium 4 based cores.

Even their 32nm SRAM prototype doesn't clock that high, which seems to be the vehicle for determining frequencies of upcoming CPUs(roughly speaking).

the low end has less room for profit though, right?

Depending on how low end. Maybe under-$100 is different, but Clarkdales should be Intel's cash cow that the high end 4 cores can't be. Margins might be somewhat lower, but Clarkdale will sell far more than Lynnfield/Bloomfield will.

I'm also assuming Clarkdale should be competitive with Athlon II X4's(or even some Phenom II X4's) that's clocked similarly in most apps if we look at the results from this article and apply the gains from Nehalem: http://www.anandtech.com/cpuch...ts/showdoc.aspx?i=3663

 

[Concillian]

Originally posted by: IlllI
the low end has less room for profit though, right?

Assuming AMD isn't losing money on their die sizes and Intel is charging more for smaller dies that all overclock very well (meaning yields are quite high), I think Intel is doing fine in terms of profit on the low end.

Sure it's less profit than an e8xxx series which is the same die size at $50-100 more, but the sheer numbers make up for less profit per die

 

[OCGuy]

Honestly GF is the only thing that will make substantial money. The graphics division is lucky to net money at all averaged over 4Q......

They need better marketing. When is the last time you saw an AMD ad on TV? Exactly.

 

[IntelUser2000]

You know what's interesting guys? It's that despite all their differences and being direct competitors to each other, how the high-level architectural decisions so far have come out to be strikingly similar. Who'll argue that Core 2 isn't a drastic architecture difference from Athlon X2's and its just basically an expanded version? How almost eerie is to compare between Deneb and Nehalem and see they are so similar?

Not just CPUs either.

Take a look at this: http://www.techreport.com/articles.x/17732/2

What I want you guys to see is the "CPU-aiding the iGPU part". Some might think the Intel iGPU is so weak it needs the CPU to assist. No, that's not my point. It's the assist part. There INDEED is a benefit to placing the iGPU closer to the CPU(even if its not on die). Clarkdale, with iGPU being close to the CPU without actually being on die, should be able to assist each other quite nicely.

Not only power will go down because of the package-based power control, performance will go up due to the GPU being able to use CPU resources better. Simply, it'll communicate faster. What does this have to do with AMD? Well, when Fusion comes along, it might use the resources similar to Clarkdale. The execution units on even the most powerful iGPU probably isn't that powerful, and can gain nicely from sharing CPU resources and its aid.

Remember AMD's claims that Fusion will show the advantages of being the only decent CPU and GPU manufacturer? Well, I presume similar resource sharing on currently very pitiful G45 will happen on the Fusion-based iGPU. Similar acadamia conclusion.

 

[brybir]

Originally posted by: Concillian
It seems Intel is happy to give the bottom end of the market to AMD. They could cut the legs out from AMD by pricing their low end cheaper, bringing the E5xxx series into Celeron prices and their E6 and E7 series to E5 series prices, etc...

However doing so would serve no purpose and would likely lose money. Lots of people buy those procecssors even with AMD being price and performance competitive (in that segment of the market). They'd be cutting the legs off AMD, but they'd also cut the legs off their own profit. You have to remember that the bulk of sales are through Compaq, Dell, and such and the bulk of that is to businesses who mostly aren't considering AMD at all. If they drop prices on their low end, they pick up a few retail consumers who are choosing AMD now, at the expense of losing profit on the millions of PCs that are sold without concern for their competition.

Remember when AMD had a clearly superior part in terms of price, performance and performance per watt... yeah, that time when Intel was still pricing their processors higher than AMD and still making profit? AMD undercuts Intel because it has to. There's no reason for Intel to undercut AMD or even compete at price parity in the low-end or mid-end market.

Intel also has to be careful to avoid the perception (or actual practice) of dumping its processors below cost in order to wipe out its competition. There are significant legal implications if it could be proven that this was or were to occur.

Whether it would come to pass or not in the U.S. is uncertain but you can bet the E.U., Korean and Japanese competition authorities (all those who have fined Intel or found them in violation of the nation's competition laws) would be watching with a sharp eye.

Besides, Intel is quite profitable in any event. Why not let AMD have the lower profit segment while Intel can keep on making substantial profits in its markets without having to worry about anti-trust laws?

 

[tommo123]

they wouldn't have to sell below cost would they? aren't intels chips much cheaper to produe than AMDs? if they sold them at cost they'd blow AMD out of the water wouldn't they?

 

[brybir]

Originally posted by: tommo123
they wouldn't have to sell below cost would they? aren't intels chips much cheaper to produe than AMDs? if they sold them at cost they'd blow AMD out of the water wouldn't they?

I would venture to guess that Intel's manufacturing abilities are better than AMD's overall and that they indeed get better economies of scale in the long run. But there are also other concerns.


Imagine if Intel tomorrow cut the price on its Core2Duo and Core2Quad's in half. Suddenly, the value proposition of its Core i5/i7 is significantly undermined. Sure, some will still buy the Core i7/i5 for $200+, but if you can get a Core2Quad 9300 for $105....suddenly there is a lot of self competition in that segment. Not to mention that Intel would then perhaps have to increase manufacturing capability in a segment (Core2Quad) that it is shifting away from.

I think in the end a drastic price cut on Intel's bottom line stuff would hurt Intel's own bottom line quite a bit, enough to raise the eyebrows of shareholders and perhaps the aforementioned competition folks.

But, I have no special knowledge beyond that if this marketplace were truly a free market in the most academic sense, Intel could hemorrhage money for a year or two and completely wipe out AMD. $99 Core i7's anyone? Then, once AMD is gone, Core i7 = $450 and the profit is recouped. Saved R & D if you dont have to keep innovating all the time too.

 

[tommo123]

yep, short term it would be awesome but long term we'd be screwed

 

[Phynaz]

Originally posted by: Concillian
It seems Intel is happy to give the bottom end of the market to AMD.

Word is on Jan 17th dual core 2.6Ghz Wolfdale Celerons are debuting at $53. Not much low end left after that.

 

[cbn]

Originally posted by: Idontcare

Originally posted by: KingstonU
The current spin is (RB-C2? CPU-Z) has helped Phenom a lot but has been in production for how long now? It seems to have a thermal wall since the 965 is rated at 140W TDP. Is there at least new-spin of Deneb scheduled to take it 3.6-4.0+ GHz at stock and lift the overclocking ceiling of 4.0GHz to something like 4.5+GHz?

The only other things on the horizon are desktop version of the 6-core Istanbul and Opteron's 12-core Magny-Cours. Then I guess HD6000 series GPU's by 2H 2010.

There is a C3 waiting in the wings, rumored to do for PhII what the G0 stepping did for the Q6600. Supposedly enables at 140W 3GHz Magny-Cours, cut that in half for a Thuban and watch those clocks climb as you eat up the rest of your 70W budget.

The TDP vs. clockspeed vs. # of cores analogy you are making here is interesting.

If multithreaded programming was more common place could someone make the argument that increasing the number of cores and lowering clock speed/L3 cache was more efficient from both a die size and energy consumption standpoint?

So at X die size having more cores running at slower speeds (without L3 cache) would be a more efficient use of both energy and silicon (provided multithreaded programs were available to make of all the cores.....not just half, one third or one quarter of them).

...................................................................................................................................

Good example is Phenom II X4 vs Athlon II X4. Phenom II has 50% more silicon and uses 30% more electricity at roughly the same clock speeds.

If AMD made a Athlon II as large as the Phenom II quad core.....it would have six cores (but no L3 cache)

 

[Idontcare]

Originally posted by: Just learning

Originally posted by: Idontcare

Originally posted by: KingstonU
The current spin is (RB-C2? CPU-Z) has helped Phenom a lot but has been in production for how long now? It seems to have a thermal wall since the 965 is rated at 140W TDP. Is there at least new-spin of Deneb scheduled to take it 3.6-4.0+ GHz at stock and lift the overclocking ceiling of 4.0GHz to something like 4.5+GHz?

The only other things on the horizon are desktop version of the 6-core Istanbul and Opteron's 12-core Magny-Cours. Then I guess HD6000 series GPU's by 2H 2010.

There is a C3 waiting in the wings, rumored to do for PhII what the G0 stepping did for the Q6600. Supposedly enables at 140W 3GHz Magny-Cours, cut that in half for a Thuban and watch those clocks climb as you eat up the rest of your 70W budget.

The TDP vs. clockspeed vs. # of cores analogy you are making here is interesting.

If multithreaded programming was more common place could someone make the argument that increasing the number of cores and lowering clock speed/L3 cache was more efficient from both a die size and energy consumption standpoint?

So at X die size having more cores running at slower speeds (without L3 cache) would be a more efficient use of both energy and silicon (provided multithreaded programs were available to make of all the cores.....not just half, one third or one quarter of them).

...................................................................................................................................

Good example is Phenom II X4 vs Athlon II X4. Phenom II has 50% more silicon and uses 30% more electricity at roughly the same clock speeds.

If AMD made a Athlon II as large as the Phenom II quad core.....it would have six cores (but no L3 cache)

Excellent question, the answer has something to do with what is called "interprocessor communications" in the area of computer science that deals with performance scaling with cores/threads.

I made this simplistic graphic for another thread, can't find that other thread now but the point to be made here is that any given multi-threaded application invokes a certain amount of mandatory communication between its sibling threads processing on the other cores and depending on the quantity of this interprocessor communication (which includes data sharing) the performance can become hamstrung by lack of shared cache and so on.

So a 6 core L3 cache-less Athlon II X6 might just well outperform a 4 core L3 cache-containing Phenom II X4 for those multi-threaded apps which involve very little interprocessor communications but it would be expected to underperform the PhII X4 for any mutli-threaded app which relies on large amounts of data sharing and communication between the threads.

 

[KingstonU]

Originally posted by: Dark4ng3l
The problem right now is that no matter what AMD comes out with Intel is far enough ahead that they are sitting on "untapped potential" that they can easily counter anything AMD comes to market with. At this rate even if AMD starts catching up today it would take years to get to where Intel is.

I wish I didn't agree with you. Untapped potential due to lack of competition is bad for us consumers.

 

[Idontcare]

AMD to axe jobs in Q1

 

[Eeqmcsq]

Originally posted by: Idontcare
AMD to axe jobs in Q1

That's around the time AMD is supposed to restore reduced salaries also. It's going to be a tough 2010 and 1st half of 2011. It'll be Bulldozer or bust for them.

What adds to the sad story is that Hector Ruiz, the guy most people blame for AMD's downfall, is linked to the recent insider trading story. Ex-Chief of AMD Is Linked to Galleon