Qualcomm is Bringing an Exciting New Model for Computing, and It’s Built on Arm

The center of gravity of compute is moving from a centralized model to being distributed all the way to the edge with trillions of intelligent Arm-powered devices and an Arm-based infrastructure that unlocks their combined potential. Qualcomm’s Centriq 2400, based on the Arm architecture, is a remarkably powerful platform, enabling compute hungry customers to run the most critical workloads needed to deliver the promise of the edge.

For years we’ve used a model where computing is centralized in datacenters with stacks of x86-based servers running monolithic applications. This model has been the basis for our early definition of the cloud, but this model is essentially dead. 

Today, we’re experiencing a shift as the center of gravity of computing moves to running applications specifically designed for the cloud and a more distributed model with compute happening where it’s needed most. A major factor driving this change is the required support for the computational needs of a rapidly growing number of Arm-powered autonomous machines and intelligent edge devices, predicted to be over a trillion in number once deployed. Autonomous machines, for example, will use local compute to make instantaneous and latency intolerant decisions based on real-time sensor data, but will then aggregate other data types to be shipped to cloud compute servers for further analysis and training.

As this model develops, we are seeing new solutions being offered from companies that have a uniquely fresh perspective. Today, we attended the Qualcomm launch event in San Jose, CA as they announced the commercial shipment of the Centriq 2400, their new cloud server CPU based on the Arm architecture. As a force in the IT industry, Qualcomm’s announcement is significant, representing a shared vision around a new cloud computing model enabled through an Arm-powered infrastructure. Qualcomm showed some impressive demos and performance numbers:

  • Up to 45% better performance per watt
  • Up to 8% better performance per thread
  • Active power well below 120W TDP

The Centriq 2400 is an impressive accomplishment by a talented design team with broad industry support offered by the likes of Microsoft Azure, RedHat, HPE, Xilinx, Canonical, and more. This quote from Milan Shetti, CTO, Hybrid IT, HPE really captured the day:

Today is a historic and iconic day in the computing industry with the announcement of the Centriq 2400. Congratulations to the Qualcomm team and the extended ecosystem in making this such an exciting day.

Not to mention, Matthew Prince, CEO of CloudFlare, who serves 10% of global internet requests, said that performance per watt really matters. Period. The Centriq 2400 doubles their requests served per watt. That is stunning.

I’d like to congratulate the Qualcomm team on a significant milestone in bringing their Arm-based servers to market. While there is still a lot of work to be done to build the right infrastructure for this next phase in computing, it’s fun to stop and admire the great work the team has accomplished.

  • It seems they're doing the same UEFI/ACPI stuff we're already suffering from in the x86 world, so no actual alternative. The traditional big advantage of Arm SoCs is that we do *not* require such proprietary firmware, but can easily build your own bootloader. So, I wouldn't buy this, for the same reasons I don't buy x86.