Intel Will Offer SiFive RISC-V CPUs on 7nm, Plans Own Dev Platform

Intel Will Offer SiFive RISC-V CPUs on 7nm, Plans Own Dev Platform

Intel and SiFive made a pair of announcements yesterday that underscore how serious the chip giant is about ramping its own foundry services. First, Intel has announced it has reorganized its manufacturing division and licensed SiFive’s IP portfolio. This is significant, as it gives Intel the flexibility it needs to offer a diverse set of solutions to potential customers. It’s not enough to just license a RISC-V CPU core. That core needs to hook to memory controllers, expansion buses, and possibly other compute devices depending on the role the RISC-V core plays in the system.

Second, Intel has announced that it will build its own RISC-V development platform, code-named Horse Creek. Silicon is expected in 2022, built on a 7nm process. The new platform will tap the SiFive P550, a newly-announced CPU core that represents the highest-performance RISC-V CPU yet announced.

Intel Will Offer SiFive RISC-V CPUs on 7nm, Plans Own Dev Platform

The P550 is a triple-issue core with out-of-order execution support, but it doesn’t seem to have the vector extension capabilities that the weaker P270 does. The P550 appears to be intended for general purpose applications where a stronger CPU core is useful, while the P270 is a much weaker, in-order core with a specialized 256-bit vector unit.

The P550 will be available with up to four cores. These are said to take up the same space as a single Cortex-A75, though it’s not clear if SiFive is comparing against a theoretical Cortex-A75 built on Intel’s 7nm process, or if it is claiming that a quad-core P550 on 7nm will be smaller than a Cortex-A75 built on process nodes roughly the same size as Intel’s 14nm. If the former, SiFive has pulled off a significant improvement in performance per area. If the latter, not so much.

SiFive did not explain why the P550 lacks the vector unit; it’s possible that some variants of the core will offer one at a later date. There was a period of time when an FPU was optional in both x86 and ARM CPUs, so a vector processing unit that shows up on some chips but not others doesn’t seem crazy. The company writes: “With full support for the RISC-V Vector Extension v 1.0RC, and combined with SiFive Recode, which translates existing SIMD software from popular legacy architectures to RISC-V Vector assembly code, the SiFive Performance P270 is an ideal replacement for dated SIMD architectures.”

The P550’s SPECInt 2006 score of 8.65/GHz makes it the highest-performing RISC-V core announced to date, though we wouldn’t have minded seeing benchmark data that was a little easier to compare with other chips.

“We are pleased to be a lead development partner with SiFive to showcase to mutual customers the impressive performance of their P550 on our 7nm Horse Creek platform,” said Amber Huffman, Intel Fellow and CTO of IP engineering group at Intel. “By combining Intel’s leading-edge interface IP such as DDR and PCIe with SiFive’s highest performance processor, Horse Creek will provide a valuable and expandable development vehicle for cutting-edge RISC-V applications.”

It’s interesting to see Intel exploring non-x86 architectures, but it’s also expected given the company’s stated intent to create a foundry business. Running simpler P550 CPUs through its own 7nm process might also give Intel a pipecleaner for bringing its own chips up. Intel is expected to introduce EUV at 7nm for the first time. It’ll also be the company’s first major node transition since the troubled shift to 10nm.

Feature image by Gareth Halfacree, Flickr, CC BY-SA 2.0

Continue reading

Intel’s Desktop TDPs No Longer Useful to Predict CPU Power Consumption
Intel’s Desktop TDPs No Longer Useful to Predict CPU Power Consumption

Intel's higher-end desktop CPU TDPs no longer communicate anything useful about the CPUs power consumption under load.

VIA Technologies, Zhaoxin Strengthen x86 CPU Development Ties
VIA Technologies, Zhaoxin Strengthen x86 CPU Development Ties

VIA and Zhaoxin are deepening their strategic partnership with additional IP transfers, intended to accelerate long-term product development.

Nvidia Unveils ‘Grace’ Deep-Learning CPU for Supercomputing Applications
Nvidia Unveils ‘Grace’ Deep-Learning CPU for Supercomputing Applications

Nvidia is already capitalizing on its ARM acquisition with a massively powerful new CPU-plus-GPU combination that it claims will speed up the training of large machine-learning models by a factor of 10.

How L1 and L2 CPU Caches Work, and Why They’re an Essential Part of Modern Chips
How L1 and L2 CPU Caches Work, and Why They’re an Essential Part of Modern Chips

Ever been curious how L1 and L2 cache work? We're glad you asked. Here, we deep dive into the structure and nature of one of computing's most fundamental designs and innovations.