Exponential Growth of Embedded Processor Innovation

Blog

Driven by open standards, and an ecosystem built on co-operation

The amazing thing about exponential growth is that you cannot see it when you are looking at it, you can only see it in hindsight. An unrelated, but very interesting reference to this is comes from a blog on the topic of AI http://waitbutwhy.com/2015/01/artificial-intelligence-revolution-1.html. The blog has nothing to do with EDA or semiconductor IP, but is a very worthwhile read.

I believe that embedded design – like AI and linked to it since intelligence is rapidly moving to the embedded devices – is about to enter an exponential growth phase. This growth however is not going to be driven by the ARM-centric market dynamics of the last decade.

Let’s face it – embedded processor innovation is about as active as desktop processor innovation. We have begun to celebrate single and low double digit improvements in power and performance, while in the past we celebrated multiples or orders of magnitude.

Lack of competition drives complacency, and in desktop era once AMD stopped challenging Intel, improvements slowed to a crawl. However, even there, AMD’s newest processors look to reignite the fire of competition in the desktop market- leading to a better situation for all. The new Ryzen processors are delivering similar performance at 1/2 the production cost (I just bought one for my workshop and it is awesome).

But why is it that they can compete? Well, for better or worse (for Intel), AMD has a license to the Intel x86 ISA and can create a processor that is able to leverage the existing rich x86 infrastructure (it’s also a cool tidbit that it is the AMD 64 bit standard that is used in all Intel processors). Intel allowed AMD to be competitive by giving them access to the ISA – AMD helped Intel by providing a working 64 bit ISA extension, a typical win-win. So, while not “open”, the x86 ISA allowed multiple companies to compete.

 

As they say – a rising tide lifts all boats.
However, in the embedded world, with ARM in such a dominant market position, what will be the driver that allows new processor technology to flourish and to drive innovation? A company can be an architectural license of ARM, and while that is somewhat x86’ish in term of openness we can and should do much more.

There has been no shortage of potential embedded innovation over the years with powerful new technologies regularly trying to make the jump from “interesting” to “viable”. This jump though is extremely difficult. Just to compete against the likes of ARM, Synopsys (ARC), Cadence (Tensilica) you need to invest hundreds of man years to establish a base-line of functionality. It is not uncommon to see new startups die on the vine just trying to establish the needed infrastructure. When 95% of investment goes to establishing the baseline, how can you manage to stand out? Your new technology could be 10x greater, but if it requires a new infrastructure (as they all have in the past) then it is dead in the water.

This is why the open RISC-V ISA is such a big deal. RISC-V enables innovation and competition on every level of embedded design – while preserving investments.

 

Let’s begin with business model.
There are now viable open source processors built on the RISC-V standard. If you want to build a company on these open designs and support them to your customers you have the opportunity to do so, just as RedHat did with Linux.

You can also build new SoC/Chips-on-demand consulting models and customization as SiFive (www.sifive.com) have done by leveraging open source designs to deliver new SoC devices with lower per-part cost since they do not need to pay established vendors (e.g. ARM).

Finally, you can choose to use the open implementations internally to develop you own design on which you have complete control (with associated costs). It’s ultimately a cost of ownership decision – spend with a vendor who you can rely on for support, or build that support infrastructure yourself?

 

But an open ISA enables so much more
Innovation only really happens when companies are forced to compete on merit. RISC-V removes the traditional lock-in associated with ARM, ARC. Etc. and makes sure that everyone producing a RISC-V processor needs to compete on the merits of their design for any given application.

Just at Intel and AMD need to compete on x86 – but extended by orders of magnitude with 10’s or 100’s of companies trying to build the best possible processor for your potential application.

 

Extensible to the core
A key feature of RISC-V is its extensibility – it encourages people to change and modify to suit their needs – while ensuring the core functionality and ecosystem leverage is maintained. You can add application specific instructions and acceleration while still being able to take advantage of a rich ecosystem to software and OS providers.

A single instruction or minor core enhancement will deliver improvements measured in multiples with only hours or days of effort – compare that to traditional embedded providers that delivers fractions of a multiple per generation.

Customization is no longer a dirty word – you can have your cake and eat it too. Customize a processor, yet leverage a rich infrastructure.

 

The potential is there – time will show the result
We at Codasip are just one provider of RISC-V technology, and we already see customers using our tools and technology to do way more than was possible with traditional embedded design without limiting themselves to a single IP provider.

Thanks to RISC-V we cannot keep a customer happy just by delivering “a” solution, it need to be “the” solution for them. This will not only make life harder for us, but for every one of our competitors. If we are not delivering, there will be others that are.

 

Investing in the 5% that gives 10x not the 95% that provides the baseline
We at Codasip have unique technology that allows customers to easily customize RISC-V based processors to enable them to compete on the global market. That alone is not enough, as an early RISC-V supporter we will continue to work with the foundation to provide a solid foundation (the baseline) that we and our competitors can build on – so that our customers can deliver on the key differentiations that give them the 10x they need to succeed,

RISC-V creates a symbiotic relationship between its members – within the foundation it is not only the collaborators and users that we need to be our best, but our competitors. 

 

It’s just the beginning
Many cool processor technologies have not managed to make the jump to viable and have spent most of their funding building a baseline infrastructure just to be considered – before ultimately being deemed too risky. In the past, some companies also considered Codasip technology as risky (no pun intended – ok it was), but with the advent of RISC-V that technology now allows our customers to quickly build a unique RISC-V processor variant that gives an order of magnitude improvement in performance while able to build on a common RISC-V software infrastructure.

We however are likely one of many in the future, and we will live and die by our ability to outperform other RISC-V implementations. We, and our competitors, are going to be forced to compete on the merits of our solutions – but more importantly we get to invest in those differentiations.

Will some of our investments help our competitors? Of course. Will some of their investments help us? Of course.

 

A rising tide lifts all boats…