AMD at Tech Field Day
I’ll get this out of the way right out the gate, I’m no expert in CPU technology, but the presentation from AMD was a fascinating glimpse into how these chip manufacturers are innovating to drive more performance out of less energy consumption – a classic pursuit of “doing more with less”. If you haven’t had a chance to review their full Tech Field Day Extra presentation a few weeks ago at VMware Explore 2023, I recommend you head over and take a look as it was informative. You don’t need to be an expert to understand what AMD is doing today and see how their goals for tomorrow align well with the market.
AMD split their presentation into 2 core sections. The first was focused on how their latest generation of EPYC processors are aiming to provide the performance demanded by the latest AI-driven applications, while also lowering the resource consumption needed to support that horsepower. The second half was a bit of a surprise as AMD sprung on us all a new collaboration with VMware on VM migration and disaster recovery (DR) called VAMT (VMware® Architecture Migration Tool). Totally something I was not expecting from a processor company! If you want to read about this very interesting OPEN-SOURCE toolkit, pop over to Part2 of my blog on AMD.
Now, let’s open that bag of chips!
AMD Part 1: The Impact of AI on Compute and Sustainability Initiatives
AMD has the future in view as they continue to strive and meet the ever-increasing demands of the modern data center. AMD’s chipsets have long been a staple in the datacenter and are now working to deliver on achieving a balance between the growing compute needs of AI and other emerging technologies with the need for sustainability and efficiency. Regarding AI, Robert Hormuth (Corporate VP for Architecture and Strategy) set the tone early in his section of the AMD presentation commenting that today AI is in the stage of “Big Brother” where it is being used to influence humans to, buy, click, watch, and/or recommend this, that, and the other thing. Where the future of AI is going, beginning with recent GenAI models & toolkits, is a shift to what he calls the “Big Mother” phase, the human assistance era – where AI is used to help, nurture, and enhance our overall life and user experiences.
AI is driving the rapid growth of demand for compute, performance, and memory with those needs accelerating well beyond Moore’s Law over the past decade. With so much demand for performance, how do you balance that with the current goals around sustainability, efficiency, and critical among those: Power!
This is where AMD shifted gears in their presentation to focus on their diverse, heterogeneous, and custom chipsets – focusing primarily on the improvements to “more with less” from their latest generation of EPYC processors. In my view, the mantra is fundamentally built around 2 guiding principles:
1) Efficiency – Driving reductions in servers, racks, and power vs. competing chipsets. In the examples shown, the latest generation of EPYC CPUs were seeing an average of 67% reduction across all these key datacenter resources when compared to competing processors.
2) Performance – Maximizing the performance from that smaller datacenter footprint. A combination of 3rd party and AMD’s own tests highlighted an average of ~1.8x greater performance vs. Intel’s latest Xeon chips.
The numbers are impressive and AMDs portfolio of processor offering and IP is expansive. If they can solve the power concerns often present in the datacenter then the potential is obvious for the AMD CPU solutions to enable AI compute to easily scale as the inevitable demand for AI services continues to grow.
For more on the AMD CPU discussion, check out the full presentation via Tech Field Day:
Or take in the expert blogs from the rest of the delegates:
– by W. Curtis Preston (Mr. Backup himself!)
Written by Matt Tyrer. These posts reflect my own opinion and are not necessarily the opinion of my employer.