Will AMD design ARM SOCs? (Late 2011 thread)

[cbn]

Place your vote.

 

[theeedude]

No, even if they wanted to, they'd be dead before they made inroads in the ARM space.

 

[cbn]

No, even if they wanted to, they'd be dead before they made inroads in the ARM space.

A lot of people have written the same in various Anandtech threads.

I think most people expressing this sentiment are thinking AMD would adopt the classic "Me-too" strategy....which is understandable.

However, I have to believe AMD is working on a plan that is more creative and innovative if they are indeed pursuing an ARM license.

 

[theeedude]

I don't think they have the will to abandon x86, and they don't have the resources to fund parallel ARM and x86 efforts.

 

[cbn]

I don't think they have the will to abandon x86, and they don't have the resources to fund parallel ARM and x86 efforts.

I don't think they will abandon x86 either...that is AMD's high powered product line-up for desktop, virtualized server and HPC.

However, I do think they are looking for the resources to enter ARM at some level.

Maybe at the level of the Asian Smartphone SOC market?

 

[RussianSensation]

I don't think they will abandon x86 either...that is AMD's high powered product line-up for desktop, virtualized server and HPC.

Well AMD already spent 5 years developing Bulldozer and it was unsuccessful. Does it make sense to invest more $$ into desktop/server chips when 5 years of that resulted in failure? You'd be hoping that your next generation chip in 5 years from now will somehow match or beat Intel's again, despite Intel's ever present node manufacturing advantage. Intel can simply add 10-15% more IPC every 2 years, keep shrinking 12 months faster than AMD, and this course of action means unless Intel makes a mistake like Netburst, AMD will never catch them.

So what can you do?

Since AMD has laid off four vice presidents, including Patrick Moorhead (corporate vice president of strategy), it's pretty obvious that they are going to completely change their strategy.

You can't have a completely new strategy if you continue to focus on desktops/servers and graphics - because that's your current strategy. Something will be de-prioritized.

This article by Xbitlabs reveals some info:

"AMD wants to put more efforts into consumer-oriented product designs, another person, who also asked not to be identified, with knowledge of the plans revealed. In particular, the code-named Bobcat micro-architecture, which powers the company’s popular Brazos platform, will play a much more important role in the company’s future that it was previously planned. In approximately one year the company shipped 20 million Brazos (Zacate, Ontario, etc.) chipsets for netbooks and low-power notebooks, a large number. Going forward, the firm will attempt to introduce flavours of Brazos aimed at tablets, consumer products and other devices.

What is unclear is what AMD plans to do with its high-performance and server offerings. While it is hard to expect the company to completely get rid of them, it is believed that the company will change its corporate targets and rebalance investments."


Based on that info, AMD might just spend a lot less $$ on new desktop architecture following Bulldozer and pour a lot more $$ into Brazos or another mobile/tablet CPU architecture. I see AMD completely abandoning the >$200 desktop CPU market, at least until their new strategy works and they can go back with a lot more $$ and actually try to design a competitive desktop CPU.

 

[Cerb]

That makes sense. Bobcat and Zacate, as it is, is pretty much anti-Atom. But I think they could improve it greatly, as long as they aren't going for a performance crown. It has quite a few obvious weaknesses, most of which were likely chosen for perf/W, or development time reasons.

Improve efficiency, where easy to do, instead of improving total performance, then either make the L1 cache big (64K/64K?), or give it a large and fast L2 (128-256K?), backed by a victim buffer (shared?) that is slow and dense like the current L2 caches. Then, basically rework the entirety of the FPU, to be higher-performance. Tada, Bobcat+1, with much better multimedia and multitasking performance, yet still packing all that x86 goodness inside.

In the mean time, however, ARM should be improving their offerings, and IMO, if they know what's good for them long-term, they will begin removing many integration options, either completely (optional floating point needs to die ASAP, ECC caches should also not be optional), and let any that may get in the way for certain users be turned on or off by the OS.

 

[cbn]

This article by Xbitlabs reveals some info:

"AMD wants to put more efforts into consumer-oriented product designs, another person, who also asked not to be identified, with knowledge of the plans revealed. In particular, the code-named Bobcat micro-architecture, which powers the company’s popular Brazos platform, will play a much more important role in the company’s future that it was previously planned. In approximately one year the company shipped 20 million Brazos (Zacate, Ontario, etc.) chipsets for netbooks and low-power notebooks, a large number. Going forward, the firm will attempt to introduce flavours of Brazos aimed at tablets, consumer products and other devices.

The Deecan platform was originally slated to ship the end of 2011, but it has been pushed back to June 2012. http://www.tomshardware.com/news/ultrabook-Deccan-Kerala-Kabini-Wichita,13800.html

I feel that is delay in 28nm Bobcat may be a telling sign of another strategy brewing in AMD....a strategy that likely points to ARM.

 

[podspi]

I said yes, but this is more of a hope than something I am sure of. They honestly don't need even need to design their own ARM core, just use one off-the-shelf, and pour the dev money into the SoC.


Like NVDA, AMD's competitive advantage would be their strong GPU technology.


Bobcat is a great product, it will be interesting to see what BC++ looks like. I don't see it transforming into something that will end up in a smartphone or low-power tablet, though. It should do great in x86 Win tablets, but those (at least so far) have been geared more towards businesses, not consumers.

 

[Kevmanw430]

I think, If AMD could get close with ARM and be the first company with an A15 design, coupled with a great 28nm GPU, then they could be VERY sucessful as an ARM chip maker. Problem is, its probobly too late for them to be the first with an A15 design, or they'd have to get the ball rolling NOW. Now, who's to say they aren't already planning/doing this, but it is unlikely.

 

[quest55720]

I said yes, but this is more of a hope than something I am sure of. They honestly don't need even need to design their own ARM core, just use one off-the-shelf, and pour the dev money into the SoC.


Like NVDA, AMD's competitive advantage would be their strong GPU technology.


Bobcat is a great product, it will be interesting to see what BC++ looks like. I don't see it transforming into something that will end up in a smartphone or low-power tablet, though. It should do great in x86 Win tablets, but those (at least so far) have been geared more towards businesses, not consumers.

Wait till next fall with the Launch of windows 8x86. There will be a lot more consumer geared windows. If AMD can get the next gen 28NM bobcat APU Z's out there by then. I can see them selling a lot of chips and W8x86 tablets being a hot item next Xmass.

 

[cbn]

I think, If AMD could get close with ARM and be the first company with an A15 design, coupled with a great 28nm GPU, then they could be VERY sucessful as an ARM chip maker. Problem is, its probobly too late for them to be the first with an A15 design, or they'd have to get the ball rolling NOW. Now, who's to say they aren't already planning/doing this, but it is unlikely.

When you mention Cortex A15....I think this is why a lot of people vote "No".

(Like you said) AMD would be too late with Cortex A15 and there are too many competitors in the existing strong markets with A15.

My guess is that they would choose a smaller ARM processor that does not overlap with their existing product line-up (particularly Bobcat on Windows). This smaller ARM processor would also probably be a better match to Asia's slower wireless networks.

So which ARM processor would be best for AMD's line-up ? How about Cortex A7 dual core to round things out?

From higher power to lower power:

Bulldozer CPU
Bulldozer APU
Bobcat APU
Cortex A7 Smartphone SOC<----A much need improvement over the existing ARMv5 and ARMv6 single cores in the Asia Mass market phones.

Sure Asia isn't a short term fix to AMD's problems. But It could be something very good for AMD in the long run. It will be interesting to see how AMD plans to make this happen......if they are indeed planning a strategy like this!

 

[quest55720]

When you mention Cortex A15....I think this is why a lot of people vote "No".

(Like you said) AMD would be too late with Cortex A15 and there are too many competitors in the existing strong markets with A15.

My guess is that they would choose a smaller ARM processor that does not overlap with their existing product line-up (particularly Bobcat on Windows). This smaller ARM processor would also probably be a better match to Asia's slower wireless networks.

So which ARM processor would be best for AMD's line-up ? How about Cortex A7 dual core to round things out?

From higher power to lower power:

Bulldozer CPU
Bulldozer APU
Bobcat APU
Cortex A7 Smartphone SOC<----A much need improvement over the existing ARMv5 and ARMv6 single cores in the Asia Mass market phones.

Sure Asia isn't a short term fix to AMD's problems. But It could be something very good for AMD in the long run. It will be interesting to see how AMD plans to make this happen......if they are indeed planning a strategy like this!

The issue is they are years and years behind everyone else at ultra low powered GPU technology. It would take them years to even catch up to NVIDIA and many others. It is pointless for AMD to go this route they have only so many RD dollars. Those would be better spent on just about anything else.

 

[RussianSensation]

We only have to wait until Wednesday to find out AMD's new strategy.
http://www.brightsideofnews.com/new...er-from-amd-ceo-rory-read-to-the-company.aspx

AMD will announce its new course, codenamed "Project WIN"on November 9th, 2011.

"A lower cost base allows us to be more competitive today and to invest back into the business to fuel our "attack" strategies in- low power, emerging markets and the cloud."

 

[gevorg]

Even if AMD can magically catch up, ARM devices have much thinner profit margins than desktop/laptop PCs. As a pure chip manufacturer, AMD has little to gain in that market. ARM is profitable for companies like Apple who have multiple revenue streams from their ARM-based devices. In order to make money on ARM, AMD *really* needs to innovate in it, otherwise its just a cheap commodity.

 

[Cerb]

Even if AMD can magically catch up, ARM devices have much thinner profit margins than desktop/laptop PCs. As a pure chip manufacturer, AMD has little to gain in that market. ARM is profitable for companies like Apple who have multiple revenue streams from their ARM-based devices. In order to make money on ARM, AMD *really* needs to innovate in it, otherwise its just a cheap commodity.

At the same time, the ARM SoCs typically have the problem of a 1000 cuts. ARM gets a few cents, the GPU guys get a few cents, the radio guys get a few cents, the crypto guys get a few cents, the memory controller guys get a few cents, and so on and so forth. The SoC vendor is deigning how those are all glued together. AMD and Intel pay much less of that, so might be able to get substantially higher real margins. Not on a cheap phone, but on a business tablet, I could see it all day long (it will be hard for them to scale down, but equally hard for everybody else to scale up).

 

[BD231]

Cortex chips are in their nineth generation ... I don't see how AMD could cover that kind of ground.

 

[cbn]

Even if AMD can magically catch up, ARM devices have much thinner profit margins than desktop/laptop PCs.

x86 desktops have better margins, but I am worried this will decrease over time.
ARM has lower profit margins, but the volume is poised to increase more than x86

In order to make money on ARM, AMD *really* needs to innovate in it, otherwise its just a cheap commodity.

Well AMD is a CPU company with a slew of patents.

Maybe at some point (after taking the appropriate baby steps) AMD will start to integrate their engineering expertise into some kind of custom ARM CPU (via an ARM architectural license)?

 

[cbn]

What I think is very interesting is the possibility of a stronger AMD and Lenovo alliance due to the acquisition of Rory Read.

What makes this even more interesting is Lenovo's recent interest in Mass Market Smartphones for China!

So how could AMD, ARM and Lenovo all help each other?

Well.....In certain blogs I have read ARM was very concerned and disappointed their entry level Cortex A5 wasn't licensed.

http://armnews.wordpress.com/2010/12/06/cortex-a5-may-have-to-wait-for-its-big-day/

ARM News
Cortex-A5 may have to wait for its big day

With smartphone demand showing no sign of slowing, there seems little reason for manufacturers to introduce less expensive and less powerful handsets. This is great news for ARM as the royalty rate per core is far higher for Cortex-A8 processors than its predecessor, the ARM11. Unfortunately for many consumers the end price is too high and simply not affordable.

The Cortex-A8 and soon-to-launch Cortex-A9 are powerful smartphone Application Processors, but we are still waiting for the first smartphone with a Cortex-A5 to debut. The Cortex-A5 is the baby brother of the highly successful Cortex-A family, and has been designed so it scales from single-core to quad-core if necessary. It was announced on 21 October 2009 and no manufacturer has announced a product that will utilise it.

The Cortex-A5 is fully software compatible with the Cortex-A8 and Cortex-A9 so there is no technical reason why smartphones powered by the processor cannot be launched within the next 12 months. Although we love to see ARM technology continuing to march into Intel territory with more powerful processors, we also want to see less expensive smartphones on the market during 2011.

Is there a market for the Cortex-A5 in the smartphone space? Western consumers expect a certain level of performance when accessing the internet . The Cortex-A8 delivers what we believe to be the minimum performance that many would deem acceptable. Although the Cortex-A5 royalty rate is significantly less than the Cortex-A8, does it have a place in a low-end smartphone?

We want to see cheaper smartphones become available, but we don&#8217;t want less powerful handsets. And we don&#8217;t think we&#8217;re alone. Will the Cortex-A5 usher in a time of smartphones at stupid prices, or has the mobile internet revolution (using the Cortex-A8) killed the Cortex-A5? What are your views?

(They really wanted very capable $99 smartphones to protect themselves from other entry level competitors. Unfortunately for ARM the major companies were only interested in Cortex A9, insisting on old and existing design ARMv6 and in some cases ARMv5 SOC for emerging markets )

With AMD all that could potentially change! AMD has no existing ARM entry product line-up that conflicts with these new low power cpu cores (Cortex A5 and Cortex A7) ARM wants to push!

With a stronger foothold in China (and the rest of Asia) It seems believe AMD could experience some good growth potential over the next 10+ to 20+ years.

 

[wahdangun]

no, I don't think AMD will even considering to enter ARM market, it will be suicide attempt, especially after the fact that BC APU sell so well, and have higher margin, not to mention AMD didn't have competition in there and let face it ATOM is suck.