Behind the green datacenter door

Sponsored Post As part of Energy Efficient Datacenters Week at The Register, The Next Platform's Timothy Prickett Morgan recently sat down with Lynn Comp, CVP of Server Product and Technology Marketing at AMD.

Lynn has a long career in the IT industry, having spent the first ten to twelve years working with RISC based system on chip (SoC) and communications infrastructure as well as mobile handsets. She also spent time aligning multi purpose processors with general compute and working with network packet processing, Linux and video processing.

Watch the video below to see Lynn and Tim tackle the problem of datacenter owner operators having to trade thermals for increased CPU, and therefore application/service, performance as they work out what to do with their ageing server infrastructure. This is a particular issue for companies who run hosting facilities in locations where the physical expansion options are limited, either due to local regulation or constraints on available power supply.

The challenge is how to get more processing muscle and storage capacity into existing datacenter footprints, then efficiently power and cool that infrastructure. Which is pretty much the only way hosting companies can expand their application and service provision cost effectively in locations where energy prices have tripled or even quadrupled, such as in Europe where energy prices have recently skyrocketed.

The pandemic also played its part by preventing engineers being able to physically visit datacenters in order to perform critical hardware upgrades. Which leaves many providers with a degree of technical debt as they now have large chunks of legacy, power hungry equipment which they are paying significant sums of money to operate and maintain.

Watch the video to hear Tim and Lynn ponder the best way to address these issues, which could include upgrading to modern, energy efficient servers based on the latest AMD EPYC™ processors.

AMD is also engaged in eating its own dogfood by striving to achieve a for its accelerated datacenter compute nodes (think EPYC CPUs and Instinct™ accelerators) between 2020 and 2025 – a metric which it thinks will be particularly useful in meeting the requirements of AI training and HPC applications being harnessed to solve some of the planet's thorniest problems in climate change, genomics and drug discovery.

And you can't get more important than that.

Sponsored by AMD.