XBitlabs: Advanced Micro Devices Set to Unveil New Strategy Next Week

[StrangerGuy]

Why not "Bin" Stock design ARM Cores?

Right now, AMD financial guys might look at the ARM market and say, "I don't like the fact we have to pay a royalty per chip resulting in a thin margin". <----Couldn't "Binned" SOC change that?

As big caveat might be this would only work for Tablets. As I see things the contract smartphones are controlled by only a few device makers scattered over four major networks. I'd imagine a company would have to be pretty large in order to have enough "high end bin" left over to fill the spec for a special contract phone.

My Conclusion (FWIW): AMD should skip the Contract phone market (where there are too many competitors anyway) and instead focus any ARM based efforts on the non contract devices (Tablets and entry level Smart phones).

Whatever AMD do there it's way too late to the mobile phone party IMO.

 

[Cerb]

Why not "Bin" Stock design ARM Cores?

Right now, AMD financial guys might look at the ARM market and say, "I don't like the fact we have to pay a royalty per chip resulting in a thin margin". <----Couldn't "Binned" SOC change that?

As big caveat might be this would only work for Tablets. As I see things the contract smartphones are controlled by only a few device makers scattered over four major networks. I'd imagine a company would have to be pretty large in order to have enough "high end bin" left over to fill the spec for a special contract phone.

My Conclusion (FWIW): AMD should skip the Contract phone market (where there are too many competitors anyway) and instead focus any ARM based efforts on the non contract devices (Tablets and entry level Smart phones).

But, how would they be much better off? If they improved Brazos' performance where it is weakest (caches, complex integer memory work, FPU), starting with the easiest to improve for the next gen, and make a LP version, they could get close enough to ARM for business tablets and ultra-portable business notebooks, that the great inertia of x86 would make people gloss over the rough edges. Entry-level smartphones are already good and cheap, and there are excellent chips in that market, already (Qualcomm's MSM7000 series, FI). If they were going to do that, I don't doubt they could have done it 3-5 years ago, but it would be a huge risk, now.

IMO, they have been out to lunch regarding embedded markets, as well. Low-power chips with ECC and good Windows and Linux support would be awesome, and would have been easy to integrate into Brazos, the first time around (same with Atom--it seems to me that both Intel and AMD have this mentality that mass consumers with deep pockets, and data centers without enough rack space, are the only potential customers in the universe, and other markets with crap hardware/software to choose from can keep choosing crap hardware/software).

If they can get into consumer phones and tablets via ARM, then what? Try to make it on razor-thin margins against companies that make real profits and have cash to back them up? Right...

I won't say AMD shouldn't have a "mobile strategy," but if AMD is trying to get into saturated ARM markets, if they survive it, they won't be the AMD we've known. While they haven't exactly been rolling in dough, AMD at least had been successfully recovering from Ruiz, even if Dirk Meyer may not have been the best CEO (one of several things I wonder about, given that situation, is why he chose to leave, instead of dropping down to a technical management position--was it ego, or is the board really that out there/greedy?). If they started pursuing mobile graphics more, along with well-integrated platforms (have GCN come out with a decent LP variant on TSMC, and multi-platform OpenCL/DirectCompute/Brook/etc. compilers, FI), focusing on making the GPU division somewhat profitable, instead of just pumping blood in one arm while the CPUs bleed it out the other, they could probably have been on a nice bumpy road to making money. It wouldn't be flashy, but Intel not wanting to serve too many low-cost, low-margin market niches gives AMD not only large gaps to exploit, but fairly predictable ones.

If they are going to take a new direction that pits them against companies already entrenched, I can't see good coming from it. Either they partner with one or two hardware vendors, succeed, and have bad blood with others that have worked with them in the past, or they fight and lose, or they bribe Chinese officials and businessmen, and succeed until somebody else can do it better.

 

[jpiniero]

If the Board thinks AMD is being stretched too thin, then abandoning the desktop (Bulldozer is a dud) and discrete GPU (not making any money) markets makes sense. Both of them are in decline too.

I don't think ARM is a part of the strategy; in fact, ARM being so slow gives them an opportunity to attempt to get Brazos into tablets and ultrabooks. Whether they can execute is another story.

 

[cbn]

Entry-level smartphones are already good and cheap, and there are excellent chips in that market, already (Qualcomm's MSM7000 series, FI).

Entry level smartphones chips are terrible! Good example is the 624 Mhz single core ARMv5, going into Lenovo's new Chinese Mass Market Smartphone announced Oct 26th, 2011. Yes, I did write ARMv5! This is especially egregious in today's world of dual core optimized browsers. What China needs is a modern entry level smartphone SOC built around a couple of Sparrow (Cortex A5) or Kingfisher (Cortex A7) cores. I just fear Nvidia is beating everyone to the punch with "Grey" and its successors.

 

[Elixer]

How AMD announces that they are splitting the company up, with the ATI division going their own way.
Most of the OEMs want the all in one GPU+CPU, and AMD has that ready, they don't really have a use for high end video cards, since that market is shrinking, and it also makes BD look really bad.

All kidding aside, I wouldn't doubt they have been thinking about this.

 

[cbn]

I like the idea, but compared to Nvidia's and Intel's resources and headstart into that market segment, as well as the momentum and resources of the existing stalwarts (Texas Instruments, Qualcomm, etc), I have no idea how AMD could hope to gain traction in that space while simultaneously making up for lost time.

Their ONLY advantage in the x86 race is they have superior GPU tech to bolt onto their chips. That advantage does not exist in the ARM race owing to Nvidia's presence (and lengthy headstart).

I just can't convince myself that AMD is actually going to chase after Nvidia, but nothing about AMD adheres to conventional wisdom so they may well pursue it.

Of all the existing ARM companies you mentioned I feel that Nvidia has the best chance of being the "Giant Killer".

Here is why:

1. Nvidia has CUDA HPC. This gives them good profits to fuel the other strategic goals they have.
2. Nvidia has "Denver": This gives Nvidia a shot and headstart at becoming the king of ARM Single thread performance. Focusing on ARM single thread performance really makes sense here considering Nvidia GPU Computer Power is going to increase without end in sight! Developing the very best ARM CPU makes sense to keep their Growing HPC GPU fed!
3. Nvidia has a good road map for Superphones/Tablets/Tablet-like Laptops.
4. Nvidia has a roadmap for mass market Smartphones with "Grey"<----This looks like the company's future Attack strategy to me.

So what is Nvidia's true goal and why do they bother doing things that don't make sense (like entry level smartphone "Grey") to other companies?

Well, I have seen some writing suggesting they have the plan to "merge the GPU and CPU instruction set into one instruction set". It will take a long time, but I am sure this goal involves Nvidia becoming the next big thing, the "Intel" of the future. Achieving "Mass Market Smartphone" and "Lowest Common denominator" makes sense if "Nvidia proprietary language" really is the plan.

So can AMD and friends stop this? Not sure. I just wish AMD and OPEN CL HPC would make some progress! That would be start!

 

[soccerballtux]

It's actually not really possible for a company to purchase ARM.

It's not really that expensive (It is still worth more than 3 times as much as AMD presently ), but it is largely owned by it's customers that have bought it as a strategic investment to ensure supply. Even if you show up with *much* more cash than it's market value, I find it doubtful that Samsung, Apple, nVidia, Nokia, TI, and many others would be willing to sell their stakes unless they felt that their supply would be guaranteed. Basically, if you wanted to buy out ARM you'd have to buy out majority of the tech sector that depends on it. No-one has that much money.

huh? what supply? Or you referring to the "supply" of intellectual property?

 

[formulav8]

My prediction...They will instead focus on Trinity and future APUs while ignoring the high end CPU business. Low power APUs for laptops and small form factor PCs and work on modifying Bobcat to be even lower power for tablets. Possibly even work with ARM like Nvidia is doing to get into the cellphone market.

That part above


Also to the people who loves Intel so much and like to see them walk on AMD, you just might get your wish permanently. Corporate fannism I never could understand... See how happy you'll be in a couple years if it ends up being true that AMD is no longer focusing on the already declining mid/higher-end desktop market.

 

[piesquared]

That part above


Also to the people who loves Intel so much and like to see them walk on AMD, you just might get your wish permanently. Corporate fannism I never could understand... See how happy you'll be in a couple years if it ends up being true that AMD is no longer focusing on the already declining mid/higher-end desktop market.

Agreed. Forum memebers will have voted for this with their wallets and mouths. Here's the rub for people like Nemesis though(and AMD continues on, likely much healthier than before). If AMD leaves the high end desktop market, it will shrink quickly due to lack of interest, and then the innovation stops. Should check the hardware specs for gamers playing BF3. I bet Trinity plays it. Speaking of Trinity, as it turns out AMD doesn't need to waste capacity on Bulldozer desktop dies when Trinity is coming in almost 6 months ahead of schedule. Cede the shrinking high end PIB market and concentrate on the growing mobile and emerging markets. Afterall, Read is apparently pretty pumped about Fusion.

 

[drizek]

Desktop dies are server dies.

Trinity is not going to replace Zambezi.

 

[cbn]

If AMD leaves the high end desktop market, it will shrink quickly due to lack of interest, and then the innovation stops.

If x86 was able to achieve higher single threaded performance AMD would have done it.

Maybe it is time for the company to look at other architectures?

It would be particularly interesting to find out how much CPU IP AMD could bring to ARM processor design. (Or do clauses in AMD/Intel CPU Cross licensing agreements prevent this?)

 

[Revolution 11]

If x86 was able to achieve higher single threaded performance AMD would have done it.

Maybe it is time for the company to look at other architectures?

It would be particularly interesting to find out how much CPU IP AMD could bring to ARM processor design. (Or do clauses in AMD/Intel CPU Cross licensing agreements prevent this?)

What is inherent about ARM that would increase single-threaded performance? (genuine question) I thought performance/watt doesn't scale perfectly at higher power levels.

 

[podspi]

If x86 was able to achieve higher single threaded performance AMD would have done it.

Perhaps (probably) I misunderstand you, but Intel did it already :biggrin:


If you are saying that AMD can't attain higher ST performance than Intel w/ x86, but might be able to beat the ARM hoards, I concur.

 

[Cerb]

Entry level smartphones chips are terrible! Good example is the 624 Mhz single core ARMv5, going into Lenovo's new Chinese Mass Market Smartphone announced Oct 26th, 2011. Yes, I did write ARMv5! This is especially egregious in today's world of dual core optimized browsers. What China needs is a modern entry level smartphone SOC built around a couple of Sparrow (Cortex A5) or Kingfisher (Cortex A7) cores. I just fear Nvidia is beating everyone to the punch with "Grey" and its successors.

But how cheap are they, and how long will it take to get those to market?

You generally expect a cheap smart phone to run a fairly modern OS, all the same applications, just not run them quite as fast. Old ARM ISA and all, Qualcomm's run Android 2.2 and 2.3 just fine (though 2.3 has been delayed for a few phones, due to the need for custom porting), and if you don't play games or multitask, you won't notice a bit of performance difference. Maybe they should get to using A5 and A7 CPUs, and maybe they will, but the old CPUs they are using now work just fine.

 

[cbn]

But how cheap are they, and how long will it take to get those to market?

According to this marketing graph from ARM, Cortex A5 is actually the same size as ARM9 (ARMv5) if they were built on the same manufacturing process. Therefore it should be the same price to make (all things being equal)

However, I have been under the impression Cortex A5 only comes on 40nm....which is a more expensive wafer. Maybe when the price of these wafers drop enough, we will see more companies upgrade to the Cortex A5 "Sparrow" and Cortex A7 "Kingfisher" later for their mass market SOCs.

You generally expect a cheap smart phone to run a fairly modern OS, all the same applications, just not run them quite as fast. Old ARM ISA and all, Qualcomm's run Android 2.2 and 2.3 just fine (though 2.3 has been delayed for a few phones, due to the need for custom porting), and if you don't play games or multitask, you won't notice a bit of performance difference. Maybe they should get to using A5 and A7 CPUs, and maybe they will, but the old CPUs they are using now work just fine.

I was thinking mainly about internet browsing, which is optimized for dual core.

 

[cbn]

Entry-level smartphones are already good and cheap, and there are excellent chips in that market, already (Qualcomm's MSM7000 series, FI).

I looked up the spec of that Qualcomm MSM7000 series. It is a 800Mhz Scorpion (Cortex A8) on 45nm.

Apparently that is not cheap enough.

The mass market smart phone Lenovo plans on selling use a Marvell single core ARM9 (ARMv5) clocked at 624 Mhz for China Mobile's 3G network.

More information found here: (dated October 27, 2011)

http://www.tmcnet.com/usubmit/-marv...rtphone-china-mobiles-/2011/10/27/5887225.htm

 

[Cerb]

I looked up the spec of that Qualcomm MSM7000 series. It is a 800Mhz Scorpion (Cortex A8) on 45nm.

Not all, like the MSM7225 and MSM7227. Here they're in $200-350 phones. You won't mistake them for having a higher-end CPU if you do stress them, but with similar non-CPU processors and memory similar to the faster SoCs, they feel just fine until you do, even using ARM11s.

An even slower chip would be just fine, IMO, at the right price; and price is what they are going for (<250 to start; expect price drops over time). That it can run a version of Android that the vast majority of Android apps will run on, is far more important than CPU performance.

 

[AtenRa]

The Bulldozer Architecture was made for TLP and DLP. Increasing IPC performance through ILP is more difficult and and it doesn't scale much anymore.

Leaving the High End x86 Desktop automatically shut down Server as well and AMD have invested a lot of time, resources and money in the Bulldozer Architecture to simple let it go now.

Bulldozer arch is the beginning of the true Fusion that AMD wants to create, so for the next few years i believe they will continue to pursue that Fusion road in all platforms (Server, Desktop and Mobile)

 

[Tuna-Fish]

huh? what supply? Or you referring to the "supply" of intellectual property?

Yes. Even if you build the chips yourself, you still need a license from ARM.

 

[Ancalagon44]

I'll be interested to hear this new strategy but I'm not holding my breath. Judging by what AMD has done recently, I'm expecting a disaster.

I'm also thinking that they will abandon high end desktop, but I dont think they will abandon the server market. We will have APUs for the mobile/desktop market and CPUs for the server market. Possibly a special APU built for high floating point throughput for servers (like Cell in a way).

I dont think they will have much to do with ARM - it screams "me too". But maybe.

 

[zsdersw]

Seems to me they need to find a way to advance beyond their competition, in whatever market that is possible, and stick with it. It's time to put the "advanced" back in Advanced Micro Devices.

Leverage the strengths, sell off the weaknesses.

 

[-Slacker-]

So uh, it's "next week" already. Let's see it.

 

[Idontcare]

The Bulldozer Architecture was made for TLP and DLP. Increasing IPC performance through ILP is more difficult and and it doesn't scale much anymore.

If Sandy Bridge didn't exist then your point would not be so easily refutable.

But Intel has clearly shown there is a path forward, at least up to the IPC performance point that Sandy Bridge currently fields.

AMD either threw in the towel early, rather than trying as hard as Intel's engineers did to improve IPC, or they guessed and guessed wrong about the value to come from prioritizing TLP and DLP over ILP.

 

[Rifter]

I dont know what they have planned but:

1. You cant abandon high end desktop without also abandoning server and i dont think AMD will abandon the server market.

2. They have no where near as much cash as they would need to break into the ARM/mobile market with players like samsung and qualcomm and Nvidia already in the door.

So hopefully they wont do something stupid but with there recent track record im sure they will anyways.

 

[Ancalagon44]

1. You cant abandon high end desktop without also abandoning server and i dont think AMD will abandon the server market.

To be honest I think in some ways thats exactly what BD is. The only difference is that they tried to market it as a high end desktop part, which it is not. If it had been released as server only, it would do fine.