AMD CTO: Core wars a mistake

AMD CTO: Core wars a mistake

Summary: AMD CTO Phil Hester said the semiconductor industry made a big mistake by entering megahertz wars at the expense of power management and how applications were actually used. And now the industry could be headed for a similar mistake by playing a game of "I have more cores than you do.

TOPICS: Processors

AMD CTO Phil Hester said the semiconductor industry made a big mistake by entering megahertz wars at the expense of power management and how applications were actually used. And now the industry could be headed for a similar mistake by playing a game of "I have more cores than you do."

Hester, speaking at AMD's analyst meeting, said it's a battle the company doesn't have to fight. Instead, AMD's road map focuses on accelerated processing to trump Intel. If the idea gets mindshare, you'll have CPUs, GPUs for graphics and APUs. 

In a nutshell, accelerated processors are AMD's way of bridging enterprise and consumer semiconductors and derive value out of the ATI acquisition. The general theory: Vista's adoption will make 3D graphics the norm across all markets. Another wrinkle: AMD won't produce one general purpose chip for everything. Instead, you'll have building blocks that can be combined and customized based on applications and how they are used.

"The industry made a mistake by getting into a megahertz war. The industry could also make the same mistake by getting into core wars," said Hester. "The accelerated processing choice is better suited to applications and how they are used every day."

These accelerated processors are coming together under a "fusion vision" that should come together in 2009. Here's a look at AMD's processing priorities.

These chips are designed to accelerate computing at the systems and silicon level.

So why do we need a new chip category? Hester said new applications with digital media, high-definition and DRM will require a rethinking of current CPU designs. Meanwhile, the multi-core chip architecture will become inadequate by the end of the decade.

Below is what the accelerated processor components look like. Under AMD's scenario, accelerated processors will be building blocks that can be customized by market.

Add it up and AMD's processing lineup will contain a set of building blocks and intellectual property to mix and match processing capabilities depending on the end market.

How AMD's vision ultimately plays out remains to be seen, but the company has been ahead of the curve on the power management front. If AMD is on target it could keep Intel on its toes through the end of the decade.  

Topic: Processors

Kick off your day with ZDNet's daily email newsletter. It's the freshest tech news and opinion, served hot. Get it.


Log in or register to join the discussion
  • Ahh.... Some one has brains.

    When in the forums of George Ou's article on Multi-Core Computing, I stated that there will be a point when throwing cores at the problem will no longer be the solution, same as throwing more hertz at the problem.

    I was curious to see what AMD was going to do to prevent these 80 Core monsters from being developed. Now I know.

    Dual Cores are a wonderful thing and I think we should continue to expand, but I find AMD's Stance impressive because of Intel's new "Call to Arms" in the fact that they are rushing their CPUs to market in hopes that they can recapture what AMD took from them. I really do hope that AMD will give us more of a preview of what they have under the hood in the next couple of years, and hopefully it will kick Intel into another competitive streak.
    • Yes, indeed....

      For a few months now, it's been all about how Intel has flipped the table on AMD. Maybe now we're not so sure - AMD has surged technologically the last few years, and it's only to be expected that another wave of innovation on that scale will take a bit of time. So while some have been quick to write AMD off as having "awoken the giant and now ready to be killed by it"...I see the opposite: AMD has taken a breath and is back with more. It has the advantage of not being the heavyweight that Intel is - that's what helped AMD think out of the box and do more than just ramp up clock speeds, and that's the factor that is still present now.
      • *IF* Amd has something to offer

        I will certainly look. But as it stands *today* they are a long ways back from Intel. (In Quad-Core, energy, performance, and die size until they eventualy get to 65nm. Of course Intel is already moving to 45 nm so we will have to see if AMD can catch up.)
        • Heh...

          They're not so far back as you seem to think. AMD is producing on 65 nm *now*. What's this "eventually" stuff? oversimplify - AMD's move to 45nm is not the same as Intel's. AMD is using immersion lithography (joint venture with IBM), whereas Intel didn't go that route. Getting that immersion process to work will allow AMD to move to smaller sizes than 45 nm, whereas Intel took the shortcut and will have trouble. The immersion technique allows for more accurate lithography and will mean better yields...which will translate into something related to oh...lower costs...which just might mean...say, larger margins. Just some small business-side trifles.

 is not so clear who is ahead just by who gets to *a* 45 nm process first.
          • 65 nm and 45 nm availability

            The current schedules I see are that Intel is currently shipping the majority of their chips in 65 nm today, while AMD has just reached the ability to provide samples of their first 65 nm chip.

            AMD expects to go to volume 65 nm use in 3Q2007.

            Intel expects to be shipping 45 nm in 2Q2007.

            AMD has not even put a date for when they might get to 45 nm, and when they get there they'll STILL be at best catching up with where Intel will already be.

            Give us a ballpark for when you expect this litho advantage to have AMD shipping smaller-faster-cooler parts than Intel?
        • AMD Does Offer

          Amd has a 45nm out....Read the news!!!!
          • you sure about that?

            The news I read says that AMD has just BARELY reached the ability to supply X2 processor samples at 65 nm, not 45.

            What processor do you believe AMD has at 45 nm and what fab is it coming from?
        • AMD will have something to offer

          You have two ways of going here.

          If you are buying a new PC today you have two choice. You can buy with intentions to replace or buy with intentions to upgrade.

          Now if you buy with intentions to replace I'd go with Intel as they have the best product out there right now for both performance and price. (The most common choice)

          If I was planning of buying to upgrade I'd go with AMD as AMD if preparing to leap frog Intel. So while you might pay more now you pay less later. Meaning that upgrading might cost you less. Still if it means a whole new Mobo chipset to run AMDs new stuff you'd still be better off going Intel for right now anyways.

          Still with AMD buying ATI that might mean gamers could get better performance from AMD, maybe.

          Still I'm not a fan of Intel of Intel Mother Boards, I find the AMD one seem to be better. So that leans me to AMD for the processor.
      • Apple should've partnered with AMD

        Their chips, to my understanding, are more reliable.
        but oh, well.
        Graham Fluet
        • More reliable?

          When was the last time you had a processor fail from any manufacturer? If you did it was probably overclocked.
    • Knew that for a while....

      Otherwise I can't explain why an AMD CPU holds up longer than an Intel CPU... :-)
      Arnout Groen
      • Deluded by AMD

        Check the failure rates. Intel has a much lower failure rate on their CPUs.
        • So give us some objective data.

          You wanna give us something to back up your claims? I could just as easily come on here and say "Check the failure rates. AMD has a much lower failure rate on their CPUs." Back it up.
        • During the first 90 days, maybe.

          But not long term. Not since the release of Athlon.
          Dr. John

        AMD's CPU's doesn't cause laptop batteries to explode..lil humor!!! But could be true......
    • Say huh???

      "but I find AMD's Stance impressive"

      News flash, AMD is trying their best to out a Quad-Core together, they just arn't doing so good at it.
    • The Sony/IBM/Toshiba group has already started this with the Cell...

      The Sony/IBM/Toshiba group has already started this with the Cell processors. Instead of having one giant (even multicored) generic CPU for handling everything, you'll have many specialized coprocessors to handle specific tasks. The Cell allows you to set each individual cell to do a specific task.

      AMD is looking to do that as well, but rather than creating an entirely new architecture like Sony, IBM, and Toshiba did, they'll be bridging existing technology. CPUs, GPUs, and APUs will be composited on the same die versus separate components on a computer (i.e. CPU, video card, and sound card.) This will allow for unparalleled throughput and communication between all three processors.
      • But...

        With all the Cores (7 of the 8 flotation, which isn't useful in gaming computations) it
        is very hard to code games for the PS3, and the only thing good about the PS3 is that
        it is the cheapest blu-ray player, which is good unless HD-DVD wins the format wars.
        Graham Fluet
        • CELL processing

          actually that 8 sub cores, each with its own Vector Processor and a small amount
          of Cache. They are all tied to one 64bit G5 class Silicon Germanium processor core
          which maintains the Order, a transmission if you will. The 8 Sub cores are the

          The cell was found to NOT be good for graphics. But this is not to say its bad for
          other applications such as Advanced Mathematics.

          You must remember the Graphics Companies ATI and Nvidea, have invested allot
          of Time and Money to create Core structures that handle Graphics data in a very
          special way. They are masters of their arts.

          SONY found only to late that the CELL would need a GPU from some one outside
          the loop to make the PS3 work. They ended up with Nvidea. But Nvidea only
          because im sure of the ATI/Microsoft deal which had taken place while CELL was
          still in development. Originally they thought they could use two CELL processors
          to make the game work... To bad. I think it still needs two, even with the RSX
    • The war will be about feature-size

      We've been in a period of leap-frog, where AMD took a nice lead from Intel and used it to gain market share. They had a better chip design, and were able to hit superior price/power/performance tradeoffs against netburst.

      Now Intel has leaped forward, core-2 is at least as good as AMD's chip design (which shows no sign of further improvement in the near future), and Intel's original strength is making the hay: They have better fabs. They'll be at 45 nm when AMD is still trying to get to 65 nm in any meaningful volume.

      AMD sees they can't compete any time soon on that dimension, so their plan is to move laterally and start bringing in GPU functionality to the main chip.

      That's a longshot. I see Intel's current fab advantage as being something which will be hard to buck. It's going to give Intel an advantage on power, on cost, on heat, and on speed - and Intel can spend that advantage anywhere they want, including pushing graphics work through the cpu regardless of the exact GPU architecture.

      AMD is making the best move it has. I don't think it'll bear fruit anytime soon, and meanwhile Intel is gonna eat their lunch due to the fab advantage.