ADLINK brings Ampere Altra SoCs to embedded with COM-HPC modules - Embedded.com

ADLINK brings Ampere Altra SoCs to embedded with COM-HPC modules

ADLINK said its new COM-HPC Ampere Altra is world’s first 80-core Arm based COM-HPC module that eliminates power-performance limitations.

ADLINK Technology has launched an 80-core Arm based COM-HPC server type module which brings the Ampere Altra systems on chip (SoCs) into the embedded systems market.

The new COM-HPC Ampere Altra server type module targets edge platforms that reliably and predictably process the most compute intensive workloads, eliminating bottlenecks and restrictions typically caused by memory caches and system memory limits on edge devices. The COM-HPC Ampere Altra core is an Ampere Altra SoC using the Arm Neoverse N1 architecture, providing premium performance within a relatively modest thermal envelope, lower total cost of ownership (TCO) than x86 designs, and significantly lower power consumption.

COM-HPC_Altra_F
The new module delivers a performance-power ratio of up to 80 Arm v8.2 64-bit cores running at up to 2.8GHz in a 175 watts envelope. (Source: ADLINK Technology)

The new module delivers a performance-power ratio of up to 80 Arm v8.2 64-bit cores running at up to 2.8GHz in a 175 watts envelope. The COM-HPC Ampere Altra provides three PCIe Gen4 x16 lanes with a homogenous architecture and valuable compute power for demanding workloads, such as real-time/near real-time applications including autonomous driving, stationary and mobile robotics, medical imaging and robotic surgery, test and measurement, and video broadcasting. In addition, it is excellently suited as a native arm64 development and compilation system for lower power embedded arm64 designs.

In a briefing with embedded.com, Henk van Bremen, director of embedded boards and modules at ADLINK Technology, explained that Ampere’s SoC is already established in the cloud computing market, and now the new COM-HPC server module will help bring this performance into the embedded market. He said the COM-HPC Ampere Altra is compliant with Arm SystemReady SR, which ensures out-of-the-box support from many standard operating systems, hypervisors, and software.

“Essentially, the architecture of the Arm chip is now accessible and allows alliances with companies like Ubuntu, even Yocto downstream. Arm has always been very closed off. Every vendor maintains their own Arm kernel. Hence for an ecosystem, this doesn’t scale. Now it can connect to several ecosystems that exist in the market.” He continued, “Having only a CPU is not that important – it’s how it can hook into different ecosystems. Now it can leverage other ecosystems.”

Jeff Wittich, chief product officer of Ampere, said, “Ampere Altra delivers the scalable power and performance necessary to drive a variety of use cases in the embedded development community from autonomous vehicles to medical instrumentation and industrial robotics. By delivering this family of COM-HPC modules in collaboration with ADLINK, we give these industries new choices for power efficient, high performance SystemReady design. These can be applied whether in a vehicle or in the many edge devices out there that have had only x86 options to choose from in the past.”

ADLINK added that by teaming up with Ampere and Arm and using their Arm Neoverse N1-based Ampere Altra SoC, its high performance-per-watt COM-HPC Ampere Altra architecture allows its strategic partners and customers to process data intensive workloads at the edge without worrying about big upfront investments, hardware overheating, or ongoing maintenance costs.

Ampere Altra is one of the first with Arm SystemReady SR certification; ADLINK is also working closely with Ampere and Arm to certify the COM-HPC Altra prototype system as a SystemReady SR device. COM-HPC Ampere Altra supports the open source edk2 as a bootloader with UEFI. Existing customers can just download a stock aarch64 (arm64) ISO such as Ubuntu and install it through booting a live ISO directly on the target. This is the same convenience developers have become used to by using x86 / amd64 target systems.

Supporting the software-defined car with Arm and partners

The new module is scalable from 32 to 80 Arm v8.2 64-bit cores (60 to 175 watts). These are the basis of the new Scalable Open Architecture for Embedded Edge (SOAFEE) announced by Arm for accelerating software-defined automotive architectures.

COM-HPC_Altra-prototyping-system
The COM-HPC Ampere Altra AVA developer platform for prototyping. (Source: ADLINK Technology)

SOAFEE’s reference hardware platforms are ADLINK’s 32-core-default COM-HPC Ampere Altra-powered AVA Developer Platform, which will be used by automakers to develop and test on Arm-based silicon, and the 80-core-default COM-HPC Ampere Altra-powered AVA-AP1 for in-vehicle prototyping.

The SOAFEE platform addresses a trend seeing automakers move their vehicles from hundreds of simultaneous discrete ECUs, collapsing their in-vehicle compute architecture into a smaller number of powerful domain controllers with a computer per major class of function (such as ADAS, infotainment) at first, to eventually one computer running all functions, with “mixed criticality” on a single powerful Arm SoC.

“ADLINK has found Ampere Altra to be the ideal solution for such a development platform for automakers to develop and test while waiting for their next generation IPs from silicon partners,” said Joe Speed, field CTO, ADLINK. “The combination of SOAFEE, led by Arm, and ADLINK’s COM-HPC Ampere Altra-powered reference platforms brings cloud technologies and best practices for CI/CD, virtualization, and security to the developer’s desk and vehicles.”

Shipping of samples of the prototype system to key ecosystem partners is already in progress. ADLINK is taking pre-orders now. ADLINK’s van Bremen said, “There is hardly anything based on the Arm 64 for software development, so this addresses that need. We have already shipped 30 of these systems to Arm and selected customers.


Related Content:

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.