New AMD CEO readies strategy shift

[Arkadrel]

afaik most of the 9-18w TDP on Bobcat is from the power hungry GPU.

Thats on 40nm though,... and with lots of otherstuff in the chip that can be stripped away because tablets dont need them.


(Brazos-T):
AMD have a 4.5watt TPD bobcat thats fully as fast as the 9watt TPD one
(and all they did to get lower TPD was remove extra stuff that tablets dont need in the chip).

Now if AMD can do a dual core 1ghz bobcat (I think it is?) @40nm tech, with 4.5w TPD, imagine what that ll be like on 28nm? Would a 28nm bobcat be able to hit ~2.5watts (TPD)? less?

Question is if thats enough, is a dual core 1ghz bobcat (fast) enough for a tablet?
How does it compair with a ARM chip that uses 2w TPD?

 

[ShadowVVL]

IDC's secret weapon...Thank god he hasn't used this on me yet!!

If amd can do good in ARM im all for it, they need the money and id like to see how they do.But if they don't I hope they wont keep dragging it on.

btw what are they going to be using fab8 for cpu,apu,gpu or arm?

 

[Arzachel]

I think getting an ARM license would be the dumbest thing AMD could possibly pull, small margins, lots of competition. What AMD, in my opinion, should instead do is create a way to hook ARM cores onto Bobcat or reverse, and get their Radeon cores onto ARM SoC's.

Think about it, a low power ARM core provides basic functionality while using barely any power, but once you open a x86 exe, the Bobcat core gets to work. Something similar to this might actually be possible with Windows 8.

 

[Medu]

Thats on 40nm though,... and with lots of otherstuff in the chip that can be stripped away because tablets dont need them.


(Brazos-T):
AMD have a 4.5watt TPD bobcat thats fully as fast as the 9watt TPD one
(and all they did to get lower TPD was remove extra stuff that tablets dont need in the chip).

Now if AMD can do a dual core 1ghz bobcat (I think it is?) @40nm tech, with 4.5w TPD, imagine what that ll be like on 28nm? Would a 28nm bobcat be able to hit ~2.5watts (TPD)? less?

Question is if thats enough, is a dual core 1ghz bobcat (fast) enough for a tablet?
How does it compair with a ARM chip that uses 2w TPD?

The 4.5watt Bobcat might well be a binned chip, just as AMD/Nvidia current laptop chips are. AMD won't be able to supply much of the market if that is the case. I have no doubt that they can get a 28nm chip to a suitable TDP for tablet, but will it's be able to keep up with quadcore ARM chips- I doubt it.

 

[Rezist]

I think getting an ARM license would be the dumbest thing AMD could possibly pull, small margins, lots of competition. What AMD, in my opinion, should instead do is create a way to hook ARM cores onto Bobcat or reverse, and get their Radeon cores onto ARM SoC's.

Think about it, a low power ARM core provides basic functionality while using barely any power, but once you open a x86 exe, the Bobcat core gets to work. Something similar to this might actually be possible with Windows 8.

Well that's all Tegra really is at the end of the day. ARM's chip with there SoC around it. Basically there graphics.

 

[waffleironhead]

Changing focus? So say they started today, how far out would a ARM product be?

I'm imagining the previous CEO was ousted when he told the BOD the reality of the situation.
BOD: We need this ARM thingy like nvidia has.
Dirk: OK, Give me 3 years and maybe 6 billion
BOD: Uh, 3 years? Any way we can get it quicker?
Dirk:Well we may be able to cut it back to 2.5 years but we'll need 9 billion
BOD: uh, about that money, what can we get for $5
Dirk: I resign.

 

[Arkadrel]

The 4.5watt Bobcat might well be a binned chip, just as AMD/Nvidia current laptop chips are. AMD won't be able to supply much of the market if that is the case. I have no doubt that they can get a 28nm chip to a suitable TDP for tablet, but will it's be able to keep up with quadcore ARM chips- I doubt it.

This is just from looking at the Coremark video nvidia did, showing the Tegra3 scoreing like 11k marks, and looking up how a D525 scores ~10k. Then compaireing a D525 to a C-50.
So guesstimateing (take it with a grain of salt, or whatever that saying is)

CPU wise.... I think we re talking in the 20-30% range or so.
A C-50 will probably be like ~20% slower than a Tegra3 (cpu wise).

Power wise... I think we re talking about C-50 TPD ~4.5watts
vs Tegra3 ~2watts TPD?

So yes the Tegra3 is a stronger chip than a C-50, and at lower power use, or atleast it looks that way from just watching that video of coremark. We ll have to see how things turn out, in more benchmarks once Window 8 is released, and the chips for the tablets.

 

[Fox5]

This is just from looking at the Coremark video nvidia did, showing the Tegra3 scoreing like 11k marks, and looking up how a D525 scores ~10k. Then compaireing a D525 to a C-50.
So guesstimateing (take it with a grain of salt, or whatever that saying is)

CPU wise.... I think we re talking in the 20-30% range or so.
A C-50 will probably be like ~20% slower than a Tegra3 (cpu wise).

Power wise... I think we re talking about C-50 TPD ~4.5watts
vs Tegra3 ~2watts TPD?

So yes the Tegra3 is a stronger chip than a C-50, and at lower power use, or atleast it looks that way from just watching that video of coremark. We ll have to see how things turn out, in more benchmarks once Window 8 is released, and the chips for the tablets.

ARM still can't compete in performance per core. Tegra3 is a quad core, and on the desktop we struggle to find apps that use quad cores, the situation is likely worse on phones/tablets. Atom falls into the same problem though, as it's reliant on hyperthreading to attain max performance. AMD's Z-01 tablet chip is unexpectedly strong in the current landscape of shipping products for tablets, and in actual performance will probably still outperform a tegra 2. The 28nm followup should beat a tegra 3 pretty handily I'd imagine.

To put it another way:
Phenom Quad core versus core 2 duo. Core 2 duo was better in almost any real situation. Per core performance will likely matter MORE in the mobile world than it does on the desktop. Price matters even more though, otherwise intel could just sweep the market with ULV 22nm core chips.

 

[krumme]

Were AMD not talking about a very mobile solution for 2013?

AMD hydrabad team who is responsible for bobcat is around 60 persons i think. Its a very small number. Ofcourse they use the key competences and tech from other places. But we are not talking huge R&D ressources - 60 persons!.

They look to work even more efficient than the Intel Haifa team. The Bobcat 2 is long ready and waiting for the tsmc 28nm, and i guess if there is a more mobile solution than bc, they must have worked on it for 1 year+. And i guess the hydrabad team are the guys for it. The new CEO can take his tiger and predetors to the zoo, if the products is not there, even a good strategy doesnt help. But it looks as if the products is comming, and that the team is there to do the hard work. And the CEO is going to take the bragging rights. Very visible for the BoD, so will be a huge success

This Hydrabad team looks to be the crown jewel in AMD right now. Executing before time. And just delivering results in spades. As with the Intel Haifa team, we can hopefully expect lots of good stuff from them in the comming years.

 

[busydude]

This Hydrabad team looks to be the crown jewel in AMD right now.

:thumbsup:

That is very good to hear.. because I am from Hyderabad. :awe:

 

[Fox5]

Were AMD not talking about a very mobile solution for 2013?

AMD hydrabad team who is responsible for bobcat is around 60 persons i think. Its a very small number. Ofcourse they use the key competences and tech from other places. But we are not talking huge R&D ressources - 60 persons!.

They look to work even more efficient than the Intel Haifa team. The Bobcat 2 is long ready and waiting for the tsmc 28nm, and i guess if there is a more mobile solution than bc, they must have worked on it for 1 year+. And i guess the hydrabad team are the guys for it. The new CEO can take his tiger and predetors to the zoo, if the products is not there, even a good strategy doesnt help. But it looks as if the products is comming, and that the team is there to do the hard work. And the CEO is going to take the bragging rights. Very visible for the BoD, so will be a huge success

This Hydrabad team looks to be the crown jewel in AMD right now. Executing before time. And just delivering results in spades. As with the Intel Haifa team, we can hopefully expect lots of good stuff from them in the comming years.

Depending on the power efficiency of a Bulldozer module, perhaps Bulldozer's cores could eventually filter down into AMD's low power chips. How well would a 1 module, low clocked, L3 cacheless Bulldozer perform, and what kind of power efficiency would it have? I'd be surprised if AMD didn't try to achieve maximal efficiency/mm^2 with Bulldozer.

 

[Idontcare]

Changing focus? So say they started today, how far out would a ARM product be?

I'm imagining the previous CEO was ousted when he told the BOD the reality of the situation.
BOD: We need this ARM thingy like nvidia has.
Dirk: OK, Give me 3 years and maybe 6 billion
BOD: Uh, 3 years? Any way we can get it quicker?
Dirk:Well we may be able to cut it back to 2.5 years but we'll need 9 billion
BOD: uh, about that money, what can we get for $5
Dirk: I resign.

Yup, the project management triangle, you can have any two of the three but not all three.

 

[mrSHEiK124]

Apple should buy AMD.

 

[sonoran]

Yup, the project management triangle, you can have any two of the three but not all three.

Yep, it's not a new saying. "Better, faster or cheaper...choose two."

 

[ultimatebob]

If I ran AMD, I think that I'd develop a better strategy for shipping product ON TIME.

Seriously, guys... they really suck at that. Sometimes I think that AMD really stands for AlMost Done after reading the news about them.

 

[LOL_Wut_Axel]

This is just from looking at the Coremark video nvidia did, showing the Tegra3 scoreing like 11k marks, and looking up how a D525 scores ~10k. Then compaireing a D525 to a C-50.
So guesstimateing (take it with a grain of salt, or whatever that saying is)

CPU wise.... I think we re talking in the 20-30% range or so.
A C-50 will probably be like ~20% slower than a Tegra3 (cpu wise).

Power wise... I think we re talking about C-50 TPD ~4.5watts
vs Tegra3 ~2watts TPD?

So yes the Tegra3 is a stronger chip than a C-50, and at lower power use, or atleast it looks that way from just watching that video of coremark. We ll have to see how things turn out, in more benchmarks once Window 8 is released, and the chips for the tablets.

CoreMark is a synthetic benchmark that loves to throw out inflated numbers that aren't true. ARM is a complete dog when it comes to floating point performance, including Tegra 3. That's what they don't tell you.

Also, it needs twice the cores and around twice the clock speed to reach comparable integer performance--in a synthetic benchmark.

 

[CHADBOGA]

Intel might reach the point where, power use isnt a issue anymore (thanks to be better process nodes, or better battery technologies). How many hours of gameing/working does a laptop/tablet really need? I think once you reach the 12hours+ of avg use, with gameing included ect, the need for faster cpu/gpu > longer battery life.

*IF* that happends, then it becomes just a matter of who makes the faster chips. 86x has a clear edge in the performance department.

Well said.

 

[senseamp]

AMD would be a hell of a lot more competitive in ARM than they are at x86, as long as Intel doesn't decide to do ARM. But, I don't think they have the resources to do both. So they have to commit. But if they do, it's game over for x86 in the long term.

 

[CHADBOGA]

AMD would be a hell of a lot more competitive in ARM than they are at x86, as long as Intel doesn't decide to do ARM. But, I don't think they have the resources to do both. So they have to commit. But if they do, it's game over for x86 in the long term.

LOL. Don't be silly.

It would be game over for AMD if they walk away from x86, that is the only thing they have going for them.

 

[LOL_Wut_Axel]

LOL. Don't be silly.

It would be game over for AMD if they walk away from x86, that is the only thing they have going for them.

This.

 

[senseamp]

Lets say "Intel/AMD" with 86x have 12hours web surfing.
Now they used "twice" as much, so ARM has 24hours web surfing.

However the Intel/AMD can do more demanding things than the ARM tablet, at faster speeds.

Which do you think people will pick?
At some point... you have long enough use, you dont need anymore.

Lets just say that the daily rutine becomes to recharge it once daily.

With arm, you ll never run into a "awww im outta power" situation, but you ll have to live with haveing bad performance in certain things vs 86x that dont.



Do you really think if such a situation comes up, that power consumption will matter in the end to the buyer's/user's?

Once you reach the critical "long enough battery life" threshhold, the payoff in more effecient power designs drops drastically (is my assumption), and Im assumeing consumers would rather have speed.

Power consumption will still matter. If ARM can run for 24 hours something that x86 runs for 12 hours, the ARM device manufacturer can cut the battery in half, make a product that is thinner, lighter, more desirable, yet cheaper to manufacture.
Intel, with its Integrated Graphics, Atom, etc, has taught people to live with and accept limitations in processing power. Hundreds of millions of Intel's own customers buy computers that can't even play modern games, and they don't care. But try selling those customers a tablet that is twice as thick as the iPad and/or needs a fan. The need for more CPU performance is nothing but a pipe dream. CPUs have been good enough since Core 2 Duo, once they could play 1080p video. That is pretty much the most CPU intensive task an average user does. The only thing an average consumer needs a powerful CPU for is video encoding and decoding, and that can be done in dedicated hardware block much faster and more efficiently.

 

[BlueBlazer]

AMD would be a hell of a lot more competitive in ARM than they are at x86, as long as Intel doesn't decide to do ARM. But, I don't think they have the resources to do both. So they have to commit. But if they do, it's game over for x86 in the long term.

The fact is Intel abandoned ARM long ago (with the sale of XScale), and pursuing x86 for smartphone and tablet markets (which has not fully come to fruition yet). AMD on the other hand does not have the R&D resources to pursue this, and adopting current ARM designs would be an easier route (just like NIVIDIA does with Tegra). :hmm:

 

[krumme]

Depending on the power efficiency of a Bulldozer module, perhaps Bulldozer's cores could eventually filter down into AMD's low power chips. How well would a 1 module, low clocked, L3 cacheless Bulldozer perform, and what kind of power efficiency would it have? I'd be surprised if AMD didn't try to achieve maximal efficiency/mm^2 with Bulldozer.

Bocat is synthesizable, meaning it can easily be portet to new processes, without time and cost consuming manual optimations. The drawback is lack of speed. In your formula it is less efficiency/mm2. But you can have have it faster to the market, and more derivatives for less cost - very fast.

If we take BD, and hopefully some more technical knowledgeable can sheed light on this, i would guess that it is comparable larger to have better modularity. If we look at the die, there is lot of area that seem to be unused. I guess that is to make it cheaper and faster to have derivatives, (different number of modules, APU configurations, future south bridge whatever). So less efficiency again for mm2.

For both bobcat and BD AMD seems to sacrifice efficiendy/mm2 for time to market and cost for modularity.

From a strategical perspective AMD is trying to compensate for its lack of ressources compared to Intel, paying perhaps 15% efficiendy, to be able to reach the same product portfolio, with 10% of the cost. Its a solution as described by idontcares project management figure.

As for low power module BD i dont think so, the front end is very beefy, and having it running in idle must cost a lot. But i guess you are right we will se very small 30mm2, 1 module sans L3 BD ! plus gpu perhaps 80 shaders - it will rock for mm2 on the low cost market. Having the ooo running on bobcat is also costly for power. Perhaps the future very low power, will have some completely different solutions to that dilemma.

Excactly therefore knowing the Hyderabad team is there, is good to know. They seem to act very fast, and with good energy.

 

[Arzachel]

CoreMark is a synthetic benchmark that loves to throw out inflated numbers that aren't true. ARM is a complete dog when it comes to floating point performance, including Tegra 3. That's what they don't tell you.

Also, it needs twice the cores and around twice the clock speed to reach comparable integer performance--in a synthetic benchmark.

This.

Also, Brazos-T will have two Bobcat cores on a 40nm process within 4.5w. Better than any ARM offering in FP and with a far better GPU.

 

[zebrax2]

I could see AMD getting an ARM license; use ARMs design for the core then design the GPU themselves. They could market their ARM line for smartphones while using brazos and its derivatives for the tablet market. It will earn them extra money and increase their brand recognition as well as develop relations to mobile manufacturers.

If I ran AMD, I think that I'd develop a better strategy for shipping product ON TIME.

Seriously, guys... they really suck at that. Sometimes I think that AMD really stands for AlMost Done after reading the news about them.

You can bet that AMD want to release BD more than anyone but a delayed product is better than an under performing one that get labeled as a bad/uncompetitive product by everyone.