32/64-bit Embedded CPUs: The Third Era Begins

Article By : Woz Ahmed

From proprietary to open source architectures, the embedded CPU and its associated business models have evolved — and so has customer expectation.

The first two eras of 32/64-bit embedded processors were defined by proprietary architectures. The third won’t be.

During the First Era of 32/64-bit embedded CPUs, which spanned the 1980s and 1990s, semiconductor companies developed and maintained their own proprietary CPU architectures.  And there were many of them. The cost of maintaining these architectures became increasingly burdensome, and third-party operating software vendors were  — for reasons of their own, involving costs, complexity and ROI  — unwilling to support multiple unique CPU architectures.  The cost crunch and lack of third-party software support encouraged companies to abandon their proprietary architectures and license processor intellectual property (IP).

This led to the Second Era, during the late 1990s and 2000s, of licensing proprietary processor architectures.  Companies such as Arm, ARC, Andes, MIPS, Tensilica, and others offered licenses to their proprietary processor cores. Arm offered an architecture license, which gave customers the right to develop their own  implementation using the vendor’s “blueprints.”

ARC, MIPS and Tensilica occupied specific niches. Arm’s licensees didn’t consider how much power they handed to their supplier.  Various Arm CPU cores proliferated across customer roadmaps; the third-party software and tools ecosystem coalesced around Arm and abandoned or rejected support for other embedded CPU architectures. ARM is now the dominant player in CPU IP.

Over time, customers were unable to resist paying high license fees for what soon became incremental innovation and unwanted features  — an engineer wouldn’t get fired for licensing Arm, but executives increasingly winced at the eye-watering prices. Customers were caught in a trap of their own making, with no one vendor strong enough to bring an architecture to market with the requisite third party software and tools support, and no credible third-party alternative existed.

Until, that is, 2010, the start of the Third Era.

The first two eras involved proprietary architectures.  The Third Era is that of the industry standard, open-source architecture.  At the vanguard of this movement is RISC-V.  With RISC-V, licensees are not tied to one vendor; they can move between different vendors, each licensing RISC-V processor IP.  This decreases the chances of an entire market being held hostage by one supplier and helps to keep licensing and royalty rates relatively lower (at a time when chip companies are fighting to preserve margins against all-powerful OEMs, where buyer concentration is at an all-time high).

On the other side of the coin, there is scope for RISC-V vendors to differentiate from each other on multiple dimensions, including performance, size, power, customizations (a source of lock-in and stickiness), and vendor capability/reliability.

Arm’s responses to the RISC-V phenomenon could have come straight out of the dominant player’s playbook, under the chapter heading “When you’re spooked.” Firstly, it released marketing collateral attempting to generate fear, uncertainty and doubt (FUD) about RISC-V, which merely served to help inform the industry of the existence of RISC-V (IBM’s FUD about minicomputer vendors achieved much the same effect).

Secondly, Arm waited to see if the RISC-V start-ups ran out of money.  To Arm’s surprise, the emerging RISC-V vendors were beginning to win customers with low-end processor IP cores (as customers finally felt they had an alternative to Arm, at least at the low-end), and more venture capital (and corporate VC) investment flowed towards RISC-V.  Companies like Western Digital heavily backed RISC-V.

To make matters worse for Arm, Softbank seemed to demand that Arm raise prices for its low-end M class processors.  This apparent misstep drove further business away from Arm and towards RISC-V. Now, more Arm customers are reviewing the value for money offered by their long-term supplier.

What does this mean?  A de facto monopoly usually doesn’t collapse overnight, but the fault lines become apparent, and customers switch one project at a time… yet Arm has a well-deserved reputation for execution and delivery, though it is becoming enveloped in the “fog of war.” A troubled Softbank is leaning on its portfolio firms to increase profits.  Arm’s internet services group (ISG) has yet to become a money-making business, with many target customers spoilt for choice in a largely standards-based sector.

The likely next tactic from that playbook would be for Arm to assert its CPU patents as the RISC-V vendors gain revenue traction.  That could stall or starve RISC-V start-ups, who likely possess few patents. It would likely annoy Arm’s customers and encourage them to increase their support for RISC-V – but dominant players don’t often think about the second or third order consequences of their actions, as the leadership mindset is usually not attuned towards potential disruption.

What about the RISC-V vendors?  Some will go bust, some will be acquired; a few may thrive.  Offering the same generic product as your competitor is not a path to riches.  Yes, some will say they have an “AI story,” others will have a “security story.”  But to me, at any rate, that is all unadulterated nonsense  — a meaningful and differentiated USP, and hence value creation and capture, comes from the intersections of these areas.

Yet still, that isn’t enough.  In IP licensing, you can provide your customer with the building blocks (IP cores) to develop a solution, or you can provide more of a solution to them  — namely, optimized and pre-verified subsystems, integrating other high-value/complex hardware IP and middleware and apps, targeted at solving a really tricky customer problem (so you get paid a premium).

Such a differentiation strategy around a standard processor architecture can potentially yield higher than average royalties, but it can only be executed by an IP vendor with a broad portfolio of high-value IP  — and the fundamental patents underpinning it, to counter the potential actions of a dominant player that is beginning to run scared.

Dominant players rarely die, they simply fade away, as others rise to prominence, accelerated by new demand drivers, i.e. new applications with no legacy requirement.  But there is an existential question for the RISC-V start-ups, namely “What else have you got?”  Their answer to that is critical to encourage customers to switch to RISC-V beyond the low-end.

Woz Ahmed, is chief strategy officer and chief of staff, Imagination Technologies.

Leave a comment