For years now, there has been what I consider a healthy, competitive tension between Arm CPU instruction set licensees and Arm’s pe-packaged and pre-validated IP licensees. (I am sure some licensees would challenge me on “healthy” given Apple’s performance.) I think it made sense that Arm would enable time to market and capability “rabbits” per market (geographically/platform) like Apple which would invest around $500M into designing custom CPUs based on an Arm ISA license. This way, Arm didn’t have to invest the R&D to have an Arm design that would yield the best metric (ie Performance- Power- Area) and then everybody else could license Arm’s pre-packaged IPs.
The Arm licensing hierarchy worked well… until it didn’t
This model of Arm licensing hierarchy works great for Arm and the market if the competitive gap isn’t too wide between custom IPs and pre-packaged IPs and that the unit volumes of the custom don’t outweigh the pre-packaged. First off, Arm doesn’t make a whole lot of revenue per unit off an Arm ISA license, and second, it makes it very hard for the pre-packaged IP customers to be market competitive. With Apple’s custom Bionic and Pro-branded Arm-based SoCs and the upcoming Nuvia-based Qualcomm custom core, we are very much entering this reality territory. In the datacenter compute market, Arm showed what it could do performance-wise with that pre-packaged IP, particularly with AWS Graviton, and it was impressive.
What about smartphones?
Enter “Blackhawk”
“Blackhawk” is Arm’s next-generation Cortex-X processor, which Arm plans to enable in smartphones shipping at the end of 2024. I think phones could be on the shelf a year from now at CES or maybe MWC. I am told that this is part of Arm CEO Rene Haas’s strategy to “eliminate the performance gap between Arm-designed processors and custom Arm implementations.” This is a big and bold claim because it is so difficult, and Apple has run the table for so long.
Arm is citing Geekbench 6 as the measure of performance merit and reflects the “largest year-over-year IPC performance increase in 5 years.” I am hopeful these performance goals translate to app performance as well. Arm also believes that Blackhawk will provide “great” LLM performance. I will assume that this has to do with big CPU IPC performance improvements as Arm says that its Cortex CPU is the #1 AI target for developers. I didn’t run the developer or app survey myself, but this makes sense to me as most AI inference in the datacenter is run on the CPU. The NPU and GPU can be an efficient way to run AI, but a CPU is the easiest and most pervasive way, which is why developers target it. A higher-performing CPU obviously helps here, but as the world moves increasingly to smaller language models, Arm’s platform with higher-performing CPU and GPU combined with its tightly integrated ML libraries and frameworks will likely result in a more efficient experience on devices.
I will be able to share more architectural details when Arm launches Blackhawk later this year, but its support for the Arm V9 architecture, with Scalable Vector Extensions 2 (SVE2) should also improve AI performance on the CPU. It could likely be the only CPU for Android phones with this capability at launch.
What If Blackhawk achieves design goals?
So, what “if” Arm delivers the highest-performance CPU core for smartphones later this year?
I want to throw out that I don’t know Blackhawk’s power or cost, but I don’t know Apple or Qualcomm’s next-generation processor architectural specs or costs either. “If” Blackhawk achieves its design goals, it means we will likely see even more competition between Arm, Qualcomm, and Apple-based CPU designs. And that is good. MediaTek is the next closest beneficiary as it doesn’t have the resources to invest in a custom CPU core. Samsung has had performance challenges with its custom CPU cores as well, so it’ll have some very big decisions to make on its next designs. I don’t see Qualcomm or Apple wavering from its custom CPU core journey until it becomes a drag on the business. Could this stir up the RISC-V talk? Maybe.
I am excited about this audacious goal for Blackhawk performance as well as other ambitious Arm strategies such as the work it’s doing on compute subsystems to help its customers further differentiate and get to market faster. I am a big believer that more competition is always better as it increases innovation and lowers prices. While Arm holds the ISA cards, winning the CPU wars falls on the backs of architects and its developers, who, to this point, haven’t been able to achieve CPU performance supremacy versus Apple. I remember people saying things that Arm would never be able to do, like being competitive in datacenter compute, and now it is. What will be the Blackhawk outcome? I would never count out Arm CEO Rene Haas.
