You are currently viewing SemiWiki as a guest which gives you limited access to the site. To view blog comments and experience other SemiWiki features you must be a registered member. Registration is fast, simple, and absolutely free so please, join our community today!

  • ARM sets up quagmire-free ecosystem for IoT

    Wandering around DAC this week, I found much of the discussion focused on the EDA community being at an inflection point. How do we get more design starts from new places with new ideas without jeopardizing existing business? It’s not as simple a transition as it sounds.

    From its very beginning as an independent company under the leadership of Robin Saxby, ARM has been about collaboration. At first this was a direct exercise with its customers, typified by the consulting operation Warren East set up when he joined the company. As the timeline unfolded, attention turned to building an ecosystem around its processor IP to ease the entire process from design to foundry, and lately including software and development platforms.

    A couple months ago I covered the joint ARM and Open Silicon webinar on the DesignStart program, offering a free look at the ARM Cortex-M0 core and a streamlined commercial license for $40k when ready to move to production. In that post, I said:

    Access to processor IP does not a chip make.
    I’m positive ARM saw that train coming, but at the time they did not quite have all the ecosystem agreements in place.

    It’s a different world we’re seeing open up. Sure, there are still the big chip design houses with their huge billion-gate designs targeting mobile and infrastructure. There are design teams with hundreds or thousands of skilled engineers to handle the design and verification, and the big EDA tool vendors are set up to serve those needs.

    At the other end, something else is happening. The IoT is dotted with open source creatives and makers galore, able to take a maker module and a lob of code and cobble together a prototype. These are people who by and large have never taken a chip through a foundry, or programmed in RTL, or seen an EDA tool. As Daniel Nenni pointed out to visitors during our book signing event at DAC, to become a startup these folks need to have a packaged design to move to the funding stage, even with crowdfunding. Established companies with vertical expertise are also getting into the IoT game – again, with ample software but less hardware design experience.

    For custom SoC projects, processor IP is a necessary but not sufficient component. ARM is finally ready to describe the “path to silicon” in IoT ecosystem terms:

    Article: What's Next For Emerging Memories-arm-path-iot-silicon.jpg

    This is starting to capture the steps and associated costs in making even a simple chip. We’ll talk about the Approved Design Partner piece shortly, but the intriguing part of this for me are two out of three of the big EDA vendors represented on this slide.

    Mentor Graphics’ participation is a bit easier to explain. Many of these designs are mixed-signal since they interface to real-world sensors and actuators. The Tanner EDA biz unit of Mentor has stepped up with Tanner AMS in an affordable package for IoT designers. They are offering a no-cost virtual lab, a sandbox hosted in the Mentor cloud for experimenting with possibilities around a Cortex-M0 core in a reference design.

    Article: What's Next For Emerging Memories-cadence-hosted-design-solution.jpgCadence has also stepped up and created a hosted design solution for the entire ARM Cortex-M family (a good move BTW, many designs need more horsepower than the Cortex-M0 provides). They also have a mixed-signal flow in these tools, as well as an ARM-developed test chip showing off the tool flow and some Cadence IP targeting IoT designs.

    Software-as-a-service (SaaS) for IoT-centric EDA is a remarkable development, not because of the technology involved but because of the economic implications. It’s a different model, albeit compartmentalized with complete IT but limited technical design support, from what Cadence deploys for its larger customers. As we look at it right this second, it’s a model Synopsys is not entirely comfortable with yet – at least in an IoT context with uncertain EDA tool volumes.

    I asked Synopsys CEO Aart de Geus about that in a fringe moment in the DAC aisle, so I wouldn’t consider his response “official”. He was at best leery of this idea given current conditions, although he clearly believes there is substantial opportunity on the IoT (see: ARC, and that may be why they aren’t excited about an ARM IoT partnership like this just yet).

    In my words, I’d agree that there are some significant challenges in setting up an economically viable EDA model that doesn’t get eaten alive in support costs from chip design noobs (again, my word used there). I’ve been there, it’s a difficult marketing problem. The last thing you want to do is scale customer counts way up and revenue per customer way down unless you can fully support a self-sufficient, incremental effort that doesn’t take away from anything else. That also explains why Cadence is bundling IP for the IoT in their model, something Synopsys could also do in theory. We’ll see if Synopsys jumps in on this program down the road.

    That drama aside, the other piece of this is the ARM Approved Design Partners. Chip design noobs, or even those with a little more experience, need exactly what these folks provide. ARM put four companies – Open Silicon, SoC Solutions, Sondrel, and eInfoChips – through the wringer of a full audit of business practices to gain approval for this program.

    Article: What's Next For Emerging Memories-arm-approved-design-partner-audit-areas.jpg

    This is a clever move – read my comments above again. It offloads support of the stuff around the ARM core to an experienced group of smaller firms in a position to deliver what we in marketing call customer intimacy. (ARM used to resemble that remark, but times change. The number and type of potential IoT customers here could overwhelm even them from a support standpoint.) They have experience in smaller design starts leveraging multi-project wafers to keep costs down. Several of these firms also have IoT-specific integration expertise in taking chip designs from the drawing board through verification into packaging and foundry readiness.

    What you should be getting from this is more than just an ecosystem announcement where companies are working together on a big picture. There is some substantial thought and vetting going on, trying to set up an ecosystem that helps IoT customers become successful while avoiding getting bogged down in a support quagmire. It sounds sinister, but it’s important for customers to understand: you want vendors to provide support, but you also want them to succeed and be behind you for the long run. At least, you should. ARM has a track record of success in setting up this type of collaboration, and the business benefits of these partnerships perhaps outweigh the technical benefits of their IP for IoT projects.