The AMD Embedded G-Series platform being introduced tonight is the world's first Accelerated Processing Unit (APU) for embedded systems. AMD has had quite a bit of history of supporting x86 based embedded systems. Starting with the Geode processor in 2003 (obtained from National Semiconductors and used in the OLPC project), AMD went on to introduce AMD64 technology into the embedded markets with the AMD Opteron processors in 2005. In 2007, the addition of graphics and other chipset options by AMD enabled comprehensive embedded solutions. In 2009, AMD introduced BGA (Ball Grid Array) packaging to meet customer demand.

At CES 2011, they gave us a sneak peek into the Embedded G-Series platform based on Brazos. AMD has increased performance and features in every generation while bringing down the power, area and price barriers for x86 in the embedded market.

The embedded market space is dominated by SoCs based on RISC processors such as ARM and MIPS. For most power sensitive embedded applications, PowerPC and x86 based solutions do not make the cut. x86, in particular, has been the dark horse due to the excessive power consumption for systems based on that architecture. Process shrinks have helped lower the power consumption numbers. However, we are still a few nodes away from when the x86 based solutions can really compete with RISC based solutions on the power front.

In the meantime, solutions like what we are seeing from AMD today integrate premium graphics capabilities within power envelops similar to what x86 used to consume in the previous generation—so you get CPU+GPU instead of just a CPU. RISC based embedded solutions may still be winning on the power front; however, for applications where slightly higher power consumption is not a concern, the x86 threat from the AMD embedded G-Series platform can become a cause for concern. MIPS is usually popular in such applications (set top boxes, digital signage etc.) and they will be facing credible opposition with AMD's integrated graphics capabilities.

The AMD Embedded G-Series
Comments Locked

76 Comments

View All Comments

  • Shadowmaster625 - Wednesday, January 19, 2011 - link

    Most embedded devices do not have a display controller. At most, they will have a simple serial port that can be used to send text to a terminal. They do not require a great deal of processing power. Usually they need just enough cpu power to keep the I/O buffers filled. Thus they use ARM.

    For some, the added convenience of being able to develop and execute code on the same platform vastly outweighs the cost of a display controller. I for one know it is a hell of a lot easier to write and debug code for an xp-embedded system than it is to code for some little piece of garbage that I might or might not have ever worked with and might or might not ever work with again! You can get a project done in half the time by using as much of your standard development platform as possible. And yes if necessary, later port it over to a more power efficent platform if the volume warrants it.
  • milkylainen - Wednesday, January 19, 2011 - link

    Yes, you are right that most embedded devices don't have a display controller such as a complex framebuffer with raster-ops etc. Most embedded devices reside in washers, elevators and buckedloads of stuff most of us take for granted. That's the core of embedded. But the analogy of using ARM devices are still just plain wrong. Look for example at the Cortex A9. SMP dual core, up to 2GHz class cpu's with up to 4M L2. Full display controller, security engines etc etc etc.
    They will still develop, in a normal configuration, their max performance within _MILLIWATT's_ of power envelope, typical PEAK. (something like 250-500mW perhaps). That's what will be residing in complex portable devices like whatever-pads and telephones. x86 in anything that has a week of uptime? Not as far as the current showings from AMD and Intel goes.
    Good luck integrating something with a fan in a ultrathin pad of some sort.
    x86 is still confined to rather large devices with large battery packs and abysmal uptime. That's why I don't like the frivolous use of "Embedded".
  • milkylainen - Wednesday, January 19, 2011 - link

    Just an addon. The Cortex A9 is a CPU complex. They are typically integrated as hard macros in a modern process with display controllers etc. It's not a SoC.
  • Muhammed - Wednesday, January 19, 2011 - link

    I am sorry , but this is not like the usual quality articles I always read on Anandtech , this just a regurgitation of marketing statements and flashy banners , The article added zero information to my knowledge which seemed very odd to me considering my solid experience with this site .

    Worse , what's being presented in the article is not at all revolutionary , or even exciting , may be for some corporate folks , but not for the general population , the new platform delivers solid price/performance ratio for sure , however on the performance front and/or the power consumption front , it's nothing special .. Yet the way it is presented in the article may make it sound like the new "Core i7" .

    Some of the paragraphs really shocked me , it made me wonder whether I am reading an article on a different "news" site , I know Anand and it's staff can do better than this , and I hope no to see this kind of press ever again on this well respected place .
  • ganeshts - Wednesday, January 19, 2011 - link

    This piece has been specifically categorized as a 'news piece', in which case we don't have any hardware to test out / characterized / determine performance or suitability for any particular application. As such, we can just do analysis of the press release provided to us with the information available at our disposal.

    Please do keep a lookout for our reviews where you will find extensive analysis and performance testing of the gadgets / chips under the scanner.
  • krumme - Wednesday, January 19, 2011 - link

    Damn it was fast. It must be the synthesized methology working big time.

    For me it just looks like the same stuff but with added integration - under the name embedded :), - but guess its what the OEMs wantet. Next is probably the southbridge :)
  • zodiacfml - Wednesday, January 19, 2011 - link

    ........ go AMD!!! :p
  • blowfish - Wednesday, January 19, 2011 - link

    You write that the AMD APU may be competitive with the ARM solutions from a power perspective when it's built on a smaller node (I think you suggested a couple of generations from the current 40Nm) Is that comparing with ARM still built on the current node though? Wouldn't the inherent power efficiency of ARM processors mean that when built on similar node size, they would always be more power efficient?
  • Shadowmaster625 - Wednesday, January 19, 2011 - link

    The embedded market is huge. And it is not as concerned with power use as you might think. Atom with XP-embedded has made huge inroads. Brazos with 7-embedded could do even more. They have to get this out there in as many segments as possible so that people can start developing unique APU code. There are some very niche functions in the embedded world that could be done much faster with SIMD arrays. AMD needs to get that 100th monkey effect as quickly as possible.
  • SlyNine - Wednesday, January 19, 2011 - link

    "In the meanwhile, solutions like what we are seeing from AMD today integrate premium graphics capabilities within power envelops similar to what --x86-- used to consume in the previous generation."

    Should that read

    "In the meanwhile, solutions like what we are seeing from AMD today integrate premium graphics capabilities within power envelops similar to what --RISC-- used to consume in the previous generation."

Log in

Don't have an account? Sign up now