ARM & The Future

Thankfully, Rory isn't HPing the company. AMD will continue to build its own x86 CPUs and GCN (and future) GPUs. The difference is that AMD will now consider, where it makes sense, using other architectures. AMD didn't come out and say it, but it's clear that the other ISA under consideration is designed by ARM. In the markets where it makes sense, AMD might deliver an ARM based solution. In others it may deliver an x86 based solution. The choice is up to the market and customer, and AMD is willing to provide either.

What's most interesting is that AMD was very clear about not wanting to be in the smartphone market. It believes, at least today, that the smartphone SoC market is too low margin to make financial sense. With smartphone SoCs selling for under $20 and given how hard it has been for Intel and NVIDIA to break into that market, I don't blame AMD for wanting to sit this one out. However, smartphones have been a huge success for ARM. If AMD is to offer ARM based SoCs coupled with their own CPU/GPU IP in other markets, it's unclear what the reception will be. The flexibility is definitely appreciated and it's a far more defensible position than saying that all future products have to use x86, but simply embracing ARM isn't a guarantee for success.

Rory Read presented a vision of the future where a large, vertically integrated device manufacturer may want to deliver custom silicon for everything from tablets to notebooks to TVs. AMD's goal is to be able to provide silicon to companies like this, while differentiating based on its own internal IP (x86 CPUs, GPU cores). One current example would be Microsoft's Xbox 360. AMD designed much of the silicon for that console, although it's using 3rd party CPU IP. In other words, should a customer want an ARM based solution mated with an AMD GPU, they could have one. If a customer wanted a strange x86/ARM APU, that would be a possibility as well.

AMD did a good job outlining that it would be more agile and flexible, however it didn't outline what specific products we'd see that implement this new architecture agnostic mentality. I suspect AMD's lack of specific examples is a result of the simple fact that the new management team has only been in place for a handful of months. It will take a while to develop outlines for the first products and a clear roadmap going forward. Until then, it's all about executing on the APU, GPU and server CPU fronts.

The New Focus: Client Mobility
Comments Locked

84 Comments

View All Comments

  • Sabresiberian - Friday, February 3, 2012 - link

    The market barely exists??

    Won't be spending much on R&D??

    You people just make this stuff up as you go along. The enthusiast market is flourishing. Dozens of companies do billions of dollars worth of business in the world selling components to enthusiast builders.

    http://www.eetimes.com/electronics-news/4229845/In...

    http://techreport.com/discussions.x/22373

    http://www.xbitlabs.com/news/multimedia/display/20...

    Even where the market share for enthusiast builds (somehow separated from "performance" builds in the xbitlabs article) is predicted to decline, it shows a stable number of computers being built in that category because the market overall increases. The research company also offers the opinion that the high-end market will "always" be a good market.

    Is the high-end consumer market the bulk of Intel sales? Of course not, but make no mistake - they DO make good money off of high-end CPUs.

    Would you throw away a billion dollars worth of business in one aspect of what you because you do 10 billion in another area? Somehow, I don't think that's what Intel has in mind.
  • wumpus - Saturday, April 14, 2012 - link

    I suspect that intel merely wanted to deny AMD the market. Starving AMD means they won't have the R&D resources needed to attack the high-margin server market.

    This seems to have been intel's strategy since the first celeron (no enthusiast would buy a K6 over an overclocked celery. No corporate customer would be caught dead buying a celery. Win-win!).
  • Malih - Thursday, February 2, 2012 - link

    since the low-end server CPUs uses AM3+ socket, we'll have to wait and see how the CPUs perform when used as desktop system, hopefully AnandTech will review them for high-end desktop use.
  • Master_Sigma - Friday, February 10, 2012 - link

    "Desktop loyalists" gave AMD the finger a long time ago when they got drunk off of the bullshit Intel benchmarketing Koolaid. Enthusiasts decided that completely synthetic marketing aids that have nothing to do real world performance were all that mattered, and unfortunately AMD doesn't have the chops to tailor their CPUs around those the way Intel can.

    AMD has finally figured that out and is giving "loyalists" exactly what they want. Have fun with your Intel monopoly. You won.
  • arjuna1 - Thursday, February 2, 2012 - link

    They are just going to leave the desktop market out in the cold??

    S*** people, get ready for sky high Intel cpu prices, developing at a crawl pace and working in a locked and limited environment.
  • bill4 - Thursday, February 2, 2012 - link

    You wish.
  • Impulses - Thursday, February 2, 2012 - link

    Prices might go up some, but Intel doesn't really gain much by squeezing a stagnant desktop market that's barely growing... They'd just give AMD an opening to jump back into it, Intel's smarter than that.

    AMD has been largely irrelevant on the desktop since the A64 and prices haven't really gone up for the mid and high end parts... They haven't gone down either but we've been getting new architectures from Intel faster than ever (comparing the last 5 years vs the previous 5-10).

    Besides, Intel still develops new architectures in a top down fashion, introducing them on the desktop first and then optimizing them for mobile. Until that changes I'm not gonna cru that the sky is falling...
  • wumpus - Saturday, April 14, 2012 - link

    Look at the GPU market. They use fabs a generation behind intel's and cost almost as much as the last generation for the same performance. Moore's law may allow you to get twice the transistors on tomorrow's chip, just don't expect to afford it.

    You will see sky high intel prices and slow growth with or without AMD. Of course, I can only hope it will only be as bad as the last 5 years (GHz holding ... holding .... holding ...).

    Finally, why would anyone expect a public corporation to act anything like a psychopath is beyond me. Simply assume they will slit your throat for a buck regardless of "what you did for them" and you won't be disappointed. Fanboy all you want, but they could care less about you.
  • Schmich - Thursday, February 2, 2012 - link

    So the reports of AMD leaving Global Foundries were false?

    "Intel is doing something similar with Haswell."
    What a missed opportunity! You can have said "Intel is doing that as well with Haswell" =D
  • bleh0 - Thursday, February 2, 2012 - link

    It just isn't viable for AMD to attempt to compete with Intel within the consumer high end desktop x86 market. More studies are showing day by day that the average consumer is moving towards a more mobile lifestyle and AMD is doing what is can to move the company in that direction. Why should AMD waste the resources and manpower on high end x86.

    Also, Intel has to compete against itself in pricing and people just can vote with their wallets if the prices get too high.

Log in

Don't have an account? Sign up now