Alibaba releases its first RISC-V CPU as open source solution for 5G and AI

Alibaba's investment in RISC-V could free them, and other Chinese companies, from licensing fees associated with the Arm ISA.

RISC-V: The open-source ISA aiming for Arm's domination of IoT and embedded devices The open source RISC-V ISA allows companies to build their own CPUs for IoT and embedded devices, without needing to pay royalties associated with Arm processors.

Pingtouge, the chip design subsidiary of e-commerce giant Alibaba, released their first core IP design based on the open-source RISC-V specification this week. The Xuantie 910 design can be integrated in CPU and system-on-chip (SoC) design, and is targeted toward 5G, artificial intelligence, and Internet of Things (IoT) usage. It "will be fully open-sourced so global developers can freely download its FPGA code," according to a report from Synced.

Alibaba claims that the Xuantie 910, built in a 16-core configuration, running at 2.5 GHz, is 40% more performant than reference RISC-V designs. According to Synced, this is enabled through the use of a "12-stage pipelined out-of-order triple issue processor with two memory accesses per cycle," and the addition of 50 instructions not in the reference RISC-V design materials.

These improvements will be open-sourced, allowing developers worldwide the ability to download their FPGA code, likewise, "Alibaba has also created a chip platform for domain specific SoC, providing hardware and software resources including CPU IP, SoC platform and algorithms, and various chip services for enterprises and developers for different AIoT scenarios," Synced reported.

SEE: Special feature: Managing the multicloud (free PDF) (TechRepublic)

While most RISC-V designs are targeted toward low-power IoT use cases, such as the Grove AI HAT for Raspberry Pi, or designed for image recognition, few high-performance RISC-V designs have been publicly announced, with the Hi-Five Unleashed one of the few commercially available boards for developers.

Alibaba's investment in RISC-V is likely an indicator of an industry trend, as the open source and effectively royalty-free RISC-V design (while the design is freely available, the name requires a license) is an effective means for hardware designers, and manufacturers to avoid the high prices of finished CPUs from Intel or AMD, or the licensing fees required for the Arm ISA. 

Economic uncertainty is also likely to play a role, with Arm suspending cooperation with Chinese smartphone and telecom equipment manufacturer Huawei following their inclusion on the "entity list," effectively banning companies that trade in US-origin technology from cooperating with blacklisted companies. While Huawei is receiving a partial and temporary reprieve, their circumstances, and the case against ZTE in 2018, created uncertainty in the market about reliance on vendors with ties to the US. Anecdotally, though Arm Holdings is a UK-based subsidiary of Japanese telecom firm SoftBank, the wide-ranging effect and strict penalties for noncompliance with US trade sanctions effectively forced their decision to end cooperation with Huawei.

Bringing the reference RISC-V architecture to enterprise applications requires a substantial amount of investment. "A lot of the things [needed] to make a RISC-V processor look like a high performance server processor are left as an exercise for the reader. Although it's a 64-bit ISA, it's not just instruction set that matters. It's all the goodies that go around the outside of it. All the memory management, virtualization, cache control, all the other stuff that delivers the processor performance," Alan Priestley, vice president analyst at Gartner, told TechRepublic.

For more on RISC-V, check out "CHIPS Alliance aims to ease RISC-V design and deployment," "Start developing for RISC-V with the $49 HiFive1 Revision B," and "RISC-V and Linux Foundations will partner to promote open source CPU" on TechRepublic.

Also see

aihardware.jpg

agsandrew, Getty Images/iStockphoto

By James Sanders

James Sanders is a staff technology writer for TechRepublic. He covers future technology, including quantum computing, AI/ML, and 5G, as well as cloud, security, open source, mobility, and the impact of globalization on the industry, with a focus on ...