Timing, opportunity and geographic location matter in life – when we decide to do important things and where we decide to live and work. It turns out these factors are also important in solving the biggest compute challenges our industry is facing at this time. I am personally very thankful for joining the MIPS leadership team at a pivotal time of transition and growth. There is not a better company as storied and as interesting as MIPS. Like many of you, I studied MIPS ISA in college and still am nostalgic about it. MIPS ISA based processors have powered some of the most iconic computing hardware from SGI workstations to our favorite game consoles, to the automotive systems that protect us and the networking equipment that connects us all together to this day.
RISC-V is Inevitable: Open and Modular
In 2021 when MIPS made the decision to transition from the MIPS ISA to RISC-V, this was a recognition of both the incredible cost of developing and maintaining an ecosystem around an Instruction Set Architecture (ISA) as well as the business model of RISC-V. As Calista Redmond eloquently put it, RISC-V is inevitable. Regardless of the differences between the MIPS and RISC-V architectures, both share in the philosophy that Reduced Instruction Set Computing (RISC) is the future. For those of us who lived through the early days of RISC-V, the ecosystem was a weakness. Every meeting was spent describing RISC-V and why it was significant. Ecosystem maturity was always the first customer concern. Nowadays, the script has flipped, and the established ecosystem and openness are the reasons RISC-V has become as successful as it has. When given a choice, an open ecosystem always wins… Just ask the Linux folks. That is not to say that RISC-V has more market share than incumbents in every market, but the inertia at this point is unstoppable. One thing the RISC-V ISA does very well is its modular approach to the architecture. Instruction extensions build on top of each other and allow for specialization without sacrificing ecosystem compatibility. This is a key feature/capability that MIPS plans to make use of.
So what are we doing at MIPS? MIPS has a rich history of solving hard problems with innovative solutions. First and foremost, we are focused on creating world class processors. I know that is a very broad statement as processors are everywhere in all shapes and sizes. There is no shortage of compute problems to be solved. Obviously, we can’t tackle all of these problems. My goal is to be very deliberate about the problems we are solving, and to solve them better than the rest of the industry by wide margins. No different than MIPS of the past. Today our target markets are Automotive, Datacenter, and Embedded AI. These in and of themselves are broad markets of which we will be targeting specific segments. But how will MIPS solutions be better than other solutions?
MIPS Solving Compute Where it Happens
One of our MIPS’ unique advantages is how we leverage multi-threading (link to white paper) to address certain applications like packet-processing. It is this willingness to do the hard work and build truly differentiated solutions that we are looking to carry forward into the new MIPS. We will continue to build multi-threaded solutions across our IP Portfolio, but the path forward goes beyond this. We are establishing deep understandings of our customers’ problems, and our customers’ customers’ problems, with the goal of truly solving their issues. What I have learned through my career in the processor industry is that it is not enough to be incrementally better. Being 10% or 20% better is not enough to convert customers from incumbent solutions. At MIPS we are looking to be orders of magnitude better at specific problems. To put it succinctly, MIPS is looking to meet compute where it happens. The growth of new technologies like AI coupled with the slowing of Moore’s law, has resulted in an explosion of accelerators and specialized compute solutions to meet the growing demand. Any time data moves, I see an opportunity for MIPS. First our products have been the best data movement CPUs in the world. Second, what if we can eliminate the need to move data altogether by moving the specialized compute to the CPU, improving performance and reducing power in the process not to mention reducing the complexity of the system? This is the type of disruption that is needed to move the industry forward and that I am confident we can deliver.
Again, I am proud to be in the position to build on the legacy of MIPS and in moving the conversation beyond micro-benchmarks to real problems that our customers truly care about. Stay tuned for more details on what we are working on in this space. I am looking forward to solving the most complicated compute problems where they happen.