Cut Off From ARM, x86, What CPU Architectures Can Huawei Use?
This week, the trade war between China and the United States took a major turn. Multiple US companies, as well as ARM, have pledged to cut off Huawei’s access to critical semiconductor components, including SoCs and CPUs. My colleague, Ryan Whitwam, has written more on this topic, and I’ll refer you over to him for additional details.
The question we’re going to examine is a bit different: Without access to ARM or x86 processors, what architectures can Huawei practically tap for future products? Today, this question is largely academic, because experts have generally assumed that any disruption to Huawei’s product chains will be short-lived. This is still the most likely outcome. The US ban on ZTE products last year should have killed that company dead, but President Trump decided to reverse his own Department of Commerce on the issue. Huawei could be similarly spared as part of a broad agreement between China and the United States.
But let’s assume, just for fun, that Huawei remains cut off from ARM and x86 processors, and that the firm cannot acquire SoCs from MediaTek or Samsung due to the fact that these designs are still ARM-based. What could it do instead? There are two potential avenues the company could explore: RISC-V and MIPS.
MIPS is in the process of being open-sourced by the owner of the ISA, Wave Computing. The plan is to open-source every component of the MIPS ecosystem, including:
- An open use version of the baseline 32 and 64-bit MIPS Instruction Set Architecture (ISA), Release 6
- MIPS SIMD Extensions v1.0
- MIPS DSP Extensions
- MIPS Multi-Threading (MT)
- MIPS MCU
- microMIPS Architecture
- MIPS Virtualization (VZ)
Any attempt to build a MIPS-powered alternative to Android is going to run into serious problems. First, the only open-source cores released for MIPS to-date are 32-bit microcontrollers, not high-performance SoCs. Huawei would have to implement its own custom processor on the MIPS architecture, essentially starting from scratch.
One point potentially in favor of MIPS is that there have been Android devices ported to this architecture. Huawei has reportedly been designing its own OS version for some time before the ban came down, and it’s said to be based on an Android fork. Of course, Google Android ports for MIPS devices appear to have stopped after 4.0. We’re now on Android Pie (9.0), implying that an awful lot of work would have to be done to bring MIPS up on an Android device.
Huawei would likely find itself in the unenviable position of bringing up its own Android fork on the ARM-based SoCs it can still sell under license while simultaneously porting its OS to MIPS, fleshing out the capabilities of its own Android fork to compete with Google (something only Amazon has really tried to date), and building a brand-new SoC on a new instruction set with an all-new toolchain. Then it needs to optimize said OS to run on MIPS, to ensure it can offer equal or superior performance to its older ARM hardware. In our theoretical example, this would be made easier by the fact that Huawei is stuck on whatever ARM IP it licensed in 2019, while its MIPS processor would reflect the performance characteristics of 2023-2025 silicon.
It would take years to pull all the pieces of this puzzle together. It reliably takes 4-5 years to design a new CPU architecture, and while mobile designs can be simpler than an AMD or Intel chip, Huawei would be truly starting from scratch.
Meanwhile, there’s another open-source ISA you may have heard of, since it’s generated no small amount of attention in recent years. RISC-V has been used by a number of companies to design small, embedded chips, but it’s not ready for a high-performance microprocessor that would compete with the likes of ARM’s Cortex family. Nor has the architecture been popular in China to-date, at least not with developers.
RISC-V has failed to ignite much interest on the far side of the Pacific, according to EETimes, due to the popularity of ARM, the lack of free CPU implementations (being an open-source ISA doesn’t mean there’s a high-performance CPU core lying around for free), a lack of experience with this kind of design work, and a cultural mismatch between the community approach of RISC-V and Chinese engineers.
None of this suggests that Huawei would be likely to leap for RISC-V. And that puts the company behind a significant eight-ball. It can continue selling ARM SoCs it has already designed. It can work on its own fork of Android. It can ask the Chinese government to pressure the US to restore full access to US companies and services. What it can’t do — at least not quickly — is build a microprocessor for itself to replace ARM designs. While we’ve been focused strictly on the mobile market here, since Huawei is primarily a mobile company, the company will find itself unable to compete with x86 products in any market, or to leverage those same products for itself in its own server installations (apart from purchases it might make on the grey or black market, obviously).
Open source ISA’s are no replacement for toolchains, actual CPU and SoC implementations, and compatible software that can actually run on the devices you want to build. It would take Huawei 4-5 years to put all the pieces of this puzzle into place, best-case. The company probably doesn’t have that kind of time.
- Report: Huawei Is Losing Access to Critical ARM Chip Designs
- Huawei Announces Honor 20 Pro, Pretends Everything Is Fine
- Intel, Qualcomm, Broadcom, and Xilinx All Move to Cut Off Huawei