New exclusive interview with James Prior for the blog AT. He is the Head of Global Communications for SiFive, a innovative fabless company. Their designs are based on RISC-V ISA, with the target of creating the next generation of innovative new products, from AI and enterprise to healthcare, automotive, and consumer tech (PCs? 😈 Tremble x86 world!).
And if you want to know more about James and SiFive, I invite you to read the full interview…
Architecnología: I always ask at the beginning: Who is James Prior? (Describe yourself, please)
James Prior: My current role is Senior Director of Product Marketing Communications at SiFive. I have previously held roles at AMD, as Senior Product Manager, and UAB as a Systems Programmer, and Rage3D.com, as a journalist. I hold a BSc. Electrical & Electronic Engineering from the University of Portsmouth, U.K.
AT: When and how did you start being passionate about technology?
J.P.: I have been inquisitive about technology from an early age, disassembling and reassembling my toys to try to understand how they work. I began programming in Atari BASIC on the family home computer, and progressed from there.
AT: Do you have a reference? Someone who has inspired you?
J.P.: My Dad inspired me as a kid, he worked for IBM. I learned from him about computers, programming, and technology.
AT: Does it make sense to discuss about ISAs? Let me explain: sometimes the implementation (microarchitecture) is more important than the ISA itself. In addition we move towards heterogeneous architectures, GPGPU, domain specific accelerators, etc., where it seems that the concept of the classic CPU is losing its importance… Or Am I wrong?
J.P.: As the type of workloads we want to process change, the hardware being used needs to evolve with the data and requirements. As we better understand the data we’re generating, and how to use it to make decisions, the general purpose flexibility approach of classic CPUs is being augmented by fixed function accelerators. This will also be replaced with more tightly integrated systems designed with the software being developed and run in mind. Efficiency and energy use requirements, as well as manufacturability concerns drive new architectures. The ISA becomes a critical part of the discussion as the hardware design needs to be flexible, easily extended, and configurable without imposing legacy limitations or overly-burdensome licensing.
AT: RISC-V ISA: Do you think it could be predominant in the future?
J.P.: RISC-V should be the last ISA that matters. The instruction set architecture is modern with key attributes that permit the development of a huge range of design points. The true RISC approach encourages the commercial companies, academic institutions, and researchers to work together to develop extensions. This open and collborative approach ensures that RISC-V can be used as the basis for any computing platform, targeting any workload or performance level.
AT: Dominate consumer sector should be the key the success of HPC as well, especially to accelerate the development of software for this platform. What is your opinion about that?
J.P.: Software is key to the adoption of hardware platforms. RISC-V is well supported today, from both commercial software vendors and open source projects. This Will only continue to improve as more hardware and projects that use RISC-V are completed.
AT: Is adapting to the x86 ecosystem the only way for you (and other RISC-V processor developers in general)? I mean, make RISC-V based chips compatible with UEFI, hardware developed for x86 such as graphics cards, interfaces, standars, etc.
J.P.: The big opportunity for innovation today is in areas that lend themselves to software-first hardware design. Chips designed with a specific purpose and function in mind, with a range of workloads, but not necessarily a general purpose computing platform. Standards help to ensure compatibility for platforms overall, and are key to fast time to market for different products, but are not the only consideration.
AT: What does SiFive need to develop CPUs with performance similar to Intel or AMD? Do you have a roadmap marked out to reach these performance numbers?
J.P.: SiFive recently previewed a next-generation SiFive Performance core architecture that increases performance by 50% from the existing SiFive Performance P550. This is a direction we will continue to follow, increasing performance – we want to offer high-performance cores, with modern features, to be used in many markets. More details about this will be shared at the upcoming RISC-V Summit.
AT: SiFive is provider of IP cores RISC-V-based too… Are you focusing only on licensing business as Arm in the future? And what kind of licences can your customers find?
J.P.: SiFive’s focus is to be the leader in high-performance computing IP, based on RISC-V. We license our IP to companies who want to design and build SoCs for a wide variety of purposes. SiFive uses a ‘relentless innovation’ approach to our core design, releasing updates for our IP in a software-style manner twice a year. This approach enables customers to get the latest features and performance updates through our portfolio subscription model.
AT: Has it been a challenge to gather talent for the microarchitecture design? Have you recruited experienced personnel from other companies (Intel, AMD, Qualcomm, NVIDIA,…) or have you opted for young talent preferably?
J.P.: SiFive has a blend of industry experience and leaders from high profile technology companies, as well as engineers early in their careers. Our leadership includes veterans of the semiconductor industry with experience at Qualcomm, Marvell, Google, Apple, Tesla, and more. SiFive employs over 100 PhD engineers and scientists to help build our class-leading products, and we’re growing fast.
AT: Using Intel Foundry Services is a good choice (considering the innovation of TSMC or Samsung)?
J.P.: SiFive works with all leading foundry companies.
*AT: Sorry. Right, my mistake. I didn’t check this. 👉 Partners site
AT: And finally… Do you plan to launch more motherboards like the HiFive Unmatched in the near future? It is an incredible product and it would be great news to have more…
J.P.: No comment.
AT: Hahaha. Sometimes silences have more revelance than words…
J.P.: Ok, how about ‘I can’t speculate about unnanounced products’.