Friday, June 14, 2024

AMD Ups Its Datacenter Game with Visualization Cards

Datamation content and product recommendations are editorially independent. We may make money when you click on links to our partners. Learn More.

One of the huge changes in data centers is how these massive computing resources are being used. The need to both create and interpret images has increased substantially.

Today, people working with images often need to work remotely. But creating and interpreting images is very compute-intensive, and it’s something most mobile devices just can’t do. As a result, the workloads get placed in datacenters where they run on specialized servers created by both the major vendors and smaller, and often more focused, specialized companies.

As we shift these loads from the edge back to connected, centralized, optimized servers, both AMD and NVIDIA accurately see this as a major opportunity.

This trend is interesting in the context of the AMD Radeon Pro V340 launch.

The Ping Pong Between Centralized and Edge Computing

Much of the PC market’s rise was predicated on a lie. That lie was that edge computing was inherently better than sharing a central resource.

With the emergence of the cloud and the increased capability of services like IBM Cloud, Dell Virtustream, Microsoft Azure, Google Cloud and, of course, AWS, not to mention the emergence of private clouds, things are flipping back very quickly. Companies are seeing the benefits of having centralized compute resources that can be shared among employees. Centralized systems are more highly utilized, easier to secure and far easier to upgrade and support than the large numbers of high-performance PCs they supplement or replace.

In addition, you really can’t get the performance you need for efforts like photorealistic real-time rendering in an affordable desktop or workstation product, let alone a laptop. Because of the extreme performance, power requirements and expense of the technology, placing it centrally makes sense. Power and cooling are available, and it allows the resource to be shared so that the cost can be spread among users. It also makes it easier to security the hardware physically so it doesn’t grow legs.

The market, as a result, is making a huge pivot back to centralized computing, and one of the most interesting parts of this pivot is visual data. This takes us to GPUs and AMD’s announcement.

AMD Radeon Pro V340

One of the big examples of this move if the creation of GPU cards designed specifically to fit into high-performance servers. It wasn’t that long ago that GPUs were only in PCs and workstations. That is no longer true, and while initially desktop cards were repurposed to work in servers, now the GPU vendors are building cards that are specifically designed to work in servers and thus provide more performance per dollar than their PC-focused and unoptimized (for servers) counterparts.

The AMD Radeon Pro V340 is a dual-GPU card based on AMD’s cutting-edge Vega architecture. One of the interesting differentiators is how the card is priced for the segment. There are no directly required end user licenses; you just pay for the hardware. Each card can support up to 32 virtual machines. AMD includes remote management tools. It uses fast HBM2 memory, and there are security features specifically designed for virtual machine environments.

As with all products in this class, validation must be assured by software solutions vendors that support it, and benchmarking should be done with real-world loads and applications consistent with final implementation.

The World Is Changing Back

While the AMD GPU cards are impressive, the bigger news is the huge shift to the cloud for what are typically workstation and, sometimes, PC workloads. I think we are now going to find these cloud workloads migrate downstream. In a few years, we’ll all be largely living off cloud-run apps and services.

We kind of already do that for much of our entertainment today. It’s just fascinating how quickly this future is coming and how much it looks like the not-too-recent mainframe past.

Photo courtesy of Shutterstock.

Subscribe to Data Insider

Learn the latest news and best practices about data science, big data analytics, artificial intelligence, data security, and more.

Similar articles

Get the Free Newsletter!

Subscribe to Data Insider for top news, trends & analysis

Latest Articles