Arm held its annual developer conference team this week and many of the Moor Insights & Strategy team tuned in virtually. I got the chance to spend time with Rene Haas, Arm President in charge of its IP business, which was nice as the conference was virtual. I wanted to spend some time on what was announced and where relevant, shed some light on its impact.
Arm renamed the company’s developer conference, formerly referred to as Arm “TechCon,” to Arm DevSummit for 2020. Last year, we covered the last Arm TechCon event and talked about the ever-expanding Arm universe and the developers that support it. While this year’s conference was 100% virtual, Arm still had many announcements prepared for DevSummit and doubled the attendance (10K) of last year’s event. Last year’s theme centered around automotive, total computing, Project Cassini, and new licensing models. This year’s DevSummit, of course, happened so close to the official announcement of NVIDIA’s intent to purchase Arm from Softbank and in parallel with NVIDIA’s GTC conference.
SystemReady and NVIDIA
Arm Kicked off the show with CEO Simon Segars announcing SystemReady, which establishes a formal set of computing platform definitions to cover the range of systems that the Arm ecosystem supports. SystemReady builds on what Arm already made with ServerReady, covering an even broader set of markets and operating systems. There will be four different bands of SystemReady with SystemReady SR, ES, IR, and LS supporting server, embedded servers, infrastructure IoT and IoT edge, and Linux servers, respectively. SystemReady is the standardization component of the broader Project Cassini, which Arm announced at last year’s Arm TechCon and would announce updates to later in the conference. Recommended For You
- An Interview With Qualcomm’s Alex Rogers, EVP And President Of Its Technologies Licensing Business, On 5G IP
- Apple Claims The IPhone 12’s A14 Bionic ‘Challenges Laptops’ But Gives No Details
- NVIDIA Wants To Change The Game Yet Again With The ‘DPU’
As Arm’s customer R&D costs move to 50-50 software-hardware, SystemReady helps them to leverage its investments across a wide array of Arm-based portfolios. Competitively this gives more reasons to stay all Arm top to bottom versus spread across X86 or even delve into RISC-V.
Following the announcement of SystemReady by Arm’s CEO Simon Segars, the conference moved immediately into a fireside chat between Simon Segars and NVIDIA’s CEO Jensen Huang moderated by Arm’s President Rene Haas. This fireside chat between the two CEOs was an effort by both companies’ CEOs to answer some difficult questions about the combination of the two companies and how they are complementary in ways that move them into the future better together. Futurum Research’s Daniel Newman and I had the chance to talk with Arm President Rene Haas, and I recommend you listen to the first five minutes of our podcast with him as he clears up some misunderstandings about the deal.
Arm announced Project Cassini last year at Arm TechCon in San Jose. Project Cassini is Arm’s effort to create a reference design for edge computing, including hardware, software, and security. Because of Arm’s low-power pedigree and the low-power nature of edge computing, I believe Arm is uniquely positioned to be a key enabler of edge computing. Arm designed Cassini to address the inherent diversity of its cloud-native edge ecosystem by establishing platform standards and reference systems. At Arm DevSummit, Arm updated the standard component of Project Cassini with the implementation of SystemReady. Arm also implemented new APIs and certification programs that can be trusted by developers across a multitude of ecosystems, including PSA-certified and PARSEC. Project Cassini now provides reference solutions, developed with ecosystem partners that support cloud-native stacks at the edge. I believe these new reference solutions can help developers of edge-cloud applications to develop new applications more quickly.
CPU Roadmap Updates
Arm did not only make cloud or edge computing announcements at DevSummit. Arm also talked about the future core architectures from the company’s client line of business, which is what the company is primarily known for delivering. Because of Arm’s key role in enabling the smartphone ecosystem with its processor designs, the news that all future Cortex-A big cores will only support 64-bit starting in 2022 is a pretty big deal. Arm traditionally announces its CPU architectures in advance of its partners announcing new products using those new designs, and this year is no different.
The Matterhorn generation of CPUs will be what Arm delivers in 2021 while the Makalu generation slated for 2022, which is when Arm expects to go 64-bit only. As a part of Arm’s Total Compute strategy, the company expects total CPU performance uplift to be around 30% going from Cortex A78/X1 to Makalu. While performance is a crucial metric for Arm, the company is also focusing on security. It will introduce a new MTE (memory tagging extension) to tighten potential security vulnerabilities in the memory subsystems.
Arm is making big moves in all market segments right now except in its PC segments. That’s not to say it’s not making progress- it is, with Qualcomm and Microsoft, but just not enough right now to substantially challenge AMD or Intel. Arm needs a design with a monster CPU core to be the Windows version of what I think Apple is trying to put together.
Microsoft and Arm AI IoT Edge Collaboration
Last, but certainly not least is Arm’s partnership with Microsoft to accelerate AI innovation for IoT Edge devices. This new collaboration is the beginning of a new effort to accelerate the deployment of AI across billions of IoT edge devices, focusing on optimizing and accelerating the development lifecycle of AI workloads. Because developing AI workloads is a very iterative process, Microsoft is lending a hand with Azure cloud to train and tune ML models on the cloud. Microsoft’s partnership with Arm should help speed up the optimization, deployment, and running of AI models across almost any Arm-based endpoint device by utilizing the Azure cloud. Considering how aggressively Arm and its partners are leaning on AI as an enablement technology, it makes total sense for Microsoft to partner with Arm to streamline the process and increase its involvement in Arm’s edge cloud efforts.
Arm continues to be the company that drives both the mobile and IoT ecosystems with many announcements surrounding its core IP and significant initiatives like Project Cassini. It is a player in the datacenter in storage and networking, and I believe it has solid momentum in general-purpose server compute in university HPC and AWS’s Graviton2.
I believe that Arm did a decent job of trying to alleviate concerns around the NVIDIA acquisition. I like to say that these things take three “inoculations” or impressions for it to sink in. Still, the deal will ultimately sink or swim depending on the world’s different regulatory bodies. The merger of the two companies still has a long road ahead, but I am confident that if the merger does complete, the two companies combined will be stronger than before. I had nearly 100,000 reads of my initial analysis here if you want more of our insights.
Arm is still showing continued improvements down the road with the expected performance improvements from the Matterhorn and Makalu generations of Cortex processors. The move to only support 64-bit in big Cortex-A series cores is an interesting one but makes sense you consider that most of the high-performance computing occurs. Arm also spent a lot of time and effort reinforcing what it built with Project Cassini last year and including partnerships like Microsoft’s and establishing standards with SystemReady. Arm showed that its edge computing efforts continue to mature and prepare for the expected exponential growth of edge computing as 5G and AI applications continue to gain momentum and demand more cloud edge computing capabilities.
While Arm DevSummit lacked one single thread that tied everything together, I was OK with that as Arm no longer lacks major issues to entry into any market at this point. I have attended nearly seven Arm TechCons prior and the company’s singular thread used to be a weakness it needed to shore up or new markets it intended to enter. Arm is now a force across mobility, IoT, the edge, automotive, and the datacenter (compute, networking, storage), and I believe this year’s DevSummit represented that.
Moor Insights & Strategy Senior Analyst Anshel Sag also contributed to this article.