SiFive, a RISC-V processor design agency, unveiled two new chip designs geared toward bringing a high-performance computing resolution to varied industries.
The announcement of the SiFive Efficiency P870 and SiFive Intelligence X390 befell at an in-person press and analyst occasion in Santa Clara, California. SiFive designs chips primarily based on the open RISC-V structure, which competes with architectures from Arm and Intel.
The SiFive Efficiency P870 and SiFive Intelligence X390 provide low energy consumption, compute density, and vector compute functionality. These processors purpose to offer a major efficiency increase for data-intensive computing functions within the fields of client electronics, automotive know-how, and infrastructure.
SiFive executives highlighted the corporate’s dedication to driving high-performance RISC-V innovation and bridging the hole with different instruction set architectures. They emphasised the speedy growth of the RISC-V ecosystem, which is fueled by the rising demand for semiconductors in sectors reminiscent of client electronics and infrastructure.
SiFive CEO Patrick Little mentioned in an announcement, “The pliability of SiFive’s RISC-V options permits corporations to deal with the distinctive computing necessities of those segments and capitalize on the momentum round generative AI, the place we now have seen double-digit design wins, and for different cutting-edge functions.”
The SiFive Efficiency P870 is designed for high-performance client functions and will also be used alongside a vector processor in information facilities. It units a brand new customary for RISC-V efficiency by way of instruction set structure availability, throughput, parallelism, and reminiscence bandwidth.
With a 50% peak single-thread efficiency improve over earlier SiFive Efficiency processors, the P870 incorporates a six-wide out-of-order core and a shared cluster cache that allows as much as a 32-core cluster.
The SiFive Intelligence X390 builds upon the success of its predecessor, the SiFive Intelligence X280, in combining AI and ML functions with {hardware} accelerators for cellular, infrastructure, and automotive functions.
The X390 presents a four-fold enchancment in vector computation with its single-core configuration, doubled vector size, and twin vector ALUs. This results in quadruple the quantity of sustained information bandwidth. The SiFive Vector Coprocessor Interface eXtension (VCIX) permits corporations so as to add their very own vector directions and/or acceleration {hardware}, offering unprecedented flexibility and efficiency enhancement.
SiFive’s mixed resolution of the P870 and X390, together with a high-performance NPU cluster, offers product designers with an agile and programmable resolution for generative AI functions. This resolution presents superior compute density, low energy consumption, and distinctive flexibility for complicated workloads.
SiFive highlighted the robust curiosity from prospects in these mixed options, with a number of reaching success in silicon growth and commercialization utilizing high-performance SiFive merchandise.
The corporate continues to collaborate with companions throughout the ecosystem to make sure that software program, safety, and suppleness advantages of the open customary RISC-V ecosystem are available for SiFive processors, enabling corporations to commercialize their SiFive-powered merchandise.
SiFive is among the pioneers of RISC-V and it mentioned it’s dedicated to reworking the way forward for computing by harnessing the limitless potential of RISC-V for high-performance and data-intensive functions. Little mentioned in a press briefing that the agency has raised a whole bunch of hundreds of thousands of {dollars} and has over 500 workers in a number of places.
He mentioned the corporate has 350 design wins and prospects embrace Intel, Amazon, Qualcomm, Samsung, Google, NASA and extra. SiFive began within the embedded market and moved up the meals chain to high-performance cores.
“These prospects helped us earn our means,” Little mentioned. “Even after we promote high-performance cores, plenty of corporations will want embedded cores too.”
He mentioned the prior P600-Sequence is the highest-performing RISC-V processor which you can license.
“We transport these in quantity,” he mentioned.
As for competitors, Little mentioned it’s a two-horse race between SiFive and Arm. He famous each one among Arm’s prime prospects are additionally SiFive’s prime prospects. Particularly in cellular. It might take a couple of years to bear full fruit, however he famous the SiFive chips have 30% to 40% higher energy effectivity and compute density than Arm.
Arm has been round for many years and it just lately went public at a valuation of $54.5 billion. Little mentioned RISC-V is the fastest-growing software program ecosystem.
Nvidia guidelines information facilities with its AI processors and graphics processing items. Krste Asanovic, founding father of SiFive, mentioned these processors are sometimes duct-taped collectively.
For one thing like AI-based safety elements, you want a whole resolution, he mentioned. He mentioned AI and general-purpose computing are shifting to RISC-V, as AI wants agile {hardware}. SiFive can present full options across the core chips in a system.
“We don’t have these totally different techniques that we duct-tape collectively,” Asanovic mentioned. “We’re a really broad portfolio and inside that broad portfolio we offer all of the items.”
Little mentioned prospects doing new classes of computing usually tend to choose RISC-V over Arm, so long as they aren’t locked in. Arm is making an attempt to get into the core of processors in information facilities to compete with Intel and Superior Micro Gadgets, however Asanovic mentioned RISC-V shouldn’t be but there due to a “large software program carry” that must occur to allow it.
Little mentioned the corporate can be making good headway in penetrating numerous U.S. authorities procurements. SiFive most likely benefited straight from the Arm IPO, Little mentioned, however the development was already evident that the corporate was doing properly in locations like China, automotive, and extra.