As I reflect on 2021, I can’t help but be proud of our achievements. A year ago, I wrote about the game-changing performance of Ampere^® Altra^® as well as the cloud, hardware, and software ecosystems being built around it. Just 12 months later, we have added our 128-core Ampere Altra Max processor to our line-up, have delivered a rich hardware and software ecosystem with tremendous momentum and, as a result, customers around the globe are deploying Ampere platforms.

We are seeing the shift from legacy x86 processors built for client-server computing to modern cloud native computing built with Ampere CPUs. With data center compute expected to double in the next four years, the industry needs a new architecture to replace 40+ year old legacy platforms. If we continued building the cloud with legacy architectures, it would consume twice as much power and 60 percent more space to satisfy the expected increase in compute demand. Ampere is that cloud native platform and it is here today.

Here are just a few of the major milestones from the last year.

A Growing Ecosystem of Customers

Oracle, Microsoft, Tencent, Equinix and others announced they are using the Ampere platform.

Oracle launched the world’s first publicly available Ampere Altra platform. Since then, Oracle Cloud Infrastructure (OCI) users have been benefitting from Ampere’s core strengths of predictable high-performance, scalability, and power efficiency.

Tencent Cloud’s end customers are now able to access Ampere Altra based SR1 instances and benefit from high performance compute for their diverse workloads.

Equinix launched their Equinix Metal™ product offering with Ampere Altra processors. Through this product line, their global customers now have access to Ampere Altra based bare metal servers.

UCloud, a leading cloud computing platform, achieved top-end marks on compute-intensive workloads via their new Ampere Altra services.

Cloudflare showed that Ampere Altra delivers 16% more single core performance and 31% more multicore performance vs other alternatives.

Curiosio, a cloud-based trip planner service, demonstrated that Ampere Altra outclassed the competition on performance/server and performance/watt metrics for a wide set of applications.

Performance Unleashed

Earlier in the year, we demonstrated the performance of Ampere Altra on three key modern cloud benchmarks. The 80-core Ampere Altra Q80-33 outperformed the AMD EPYC 7742 processor on each workload, delivering higher raw performance above while also consuming less power. As an example, Ampere Altra consumed 16% lower power than the AMD EPYC 7742, while performing 9% better on SPECrate2017_int.

With the arrival of our 128-core Ampere Altra Max processor, more than twice that of any other data center processor, we showed how well the performance scales:

• Encryption – Ampere Altra Max can encrypt 57% more blocks than Ampere Altra using AES-256-gcm
• Media Encoding – Ampere Altra Max can encode 55% more aggregate frames per second than Ampere Altra, which already led current x86 processors.
• Web Serving – The performance of web servers is usually measured as throughput, or requests per second, under a specified Service Level Agreement (SLA) such as 99^th percentile (or p.99) latency. With a p.99 latency SLA of 10ms, Ampere Altra Max throughput is 51% higher than Ampere Altra using NGXINX.

Third party reviews of Ampere Altra Max also highlighted our strong performance and power efficiency:

• Ampere Altra Max review pushing it to 128-cores per socket
• Ampere Altra Max 128-Core CPU Is Priced Lower Than Flagship Xeon, EPYC CPUs

Robust Platform Ecosystem

Foxconn, Gigabyte, Inspur, Supermicro, and Wiwynn have released over 26 unique server designs across a range of processor, storage, memory, networking, and accelerator configurations. Ampere has also worked closely with ecosystem partners to develop an extensive AVL – including memory and storage drives from Kioxia, Micron, Samsung, and SK hynix; NICs from Broadcom, Intel, and Mellanox; and other accelerator cards from AMD and Nvidia.

Open-Source Firmware Updates

Ampere is committed to supporting open-source firmware on our platforms. During the last year, we established a monthly release cycle, which enables a rich set of features for Ampere Altra and Ampere Altra Max on the Mt. Jade reference platform. Because open source firmware is so important for industry innovation, our releases also include support for TianoCore/EDK, LinuxBoot, OpenBMC and OpenOCD.

The Launch of Ampere AI

AI continues to be a critical application for the cloud, driving more intelligent analytics, smarter recommendations, and better business insights. We launched Ampere AI and the AIO Optimization Layer. This solution consistently delivers inference results at the lowest latency and highest throughput for diverse set of use cases. Our customers can deploy AI services on Ampere quickly at higher performance and lower cost.

Complete Ampere Solutions

We launched a brand new solutions portal dedicated to Ampere platforms and the software solutions that they power. It provides comprehensive information for customers and partners about Ampere servers, partners and solutions in areas ranging from Cloud Native applications to Big Data. We also use the solutions portal to publish regular regression testing to our users and developers so that they have the confidence to run a full range of applications on Ampere platforms.

Bring on 2022!

To transform to a true cloud native architecture across both software and the underlying hardware, we cannot rely on legacy technology. We need a new direction that consumes less power, takes up less space and at the same time provides the predictable performance users expect and demand. This is why Ampere was founded and we are continuing to hit all key milestones on the path to this new computing paradigm. In 2022 you can expect more innovation from us – and from our partners in the growing ecosystem: new products, new platforms, and new capabilities to enhance cloud workloads.

This is an exciting time in our industry! Ampere is looking forward to leading the cloud phase of compute into the new year and many years to come.