AMD Launches Radeon Accelerators for AI and HPC

Radeon Accelerators

Using an advanced graphics processing unit (GPU) can considerably reduce power consumption and enhance performance and user experience. There is also portability, flexibility and the benefit of programmability in a straightforward fashion. Hence, they are being used to compliment technologies like Artificial Intelligence and high-performance computing (HPC). If we look into most of the AI algorithms, we can see a distributed pattern. And, these parallel problems can be resolved with the help of GPUs that can speed up 90 percent of the algorithms in AI.

And now AMD has launched the first 7nm graphics processing unit (GPU), which is also being termed as the first GPUs that are going to assist PCIe 4.0 features thoroughly. It is pertinent to mention that for more than four decades AMD has driven excogitation in graphics and visualization technologies along with high-performance computing ― the building blocks for gaming, immersive platforms, and datacenter.

Termed as Radeon Instinct MI50 and MI60, the new GPUs, although being called graphics cards, have been created solely for machine learning and AI tasks. To add to the already available features, the GPUs including both the features assist for High Bandwidth Memory 2 (HBM2), which gives them a massive 1 TB/s bandwidth to play around. Improved yet, using AMD’s Infinity Fabric, which massively enhances data bandwidth between harmonious components, is supported between graphics cards. The cards feature assist different types of multiplex computing tasks, well ahead than the traditional GPU, making them more suited for High-Performance Computing (HPC) applications. Features like built-in support for virtualization are an added advantage for cloud service providers.

According to AMD intelligent applications that interact and react with human-like reflexes demand a massive amount of computer processing power. AMD’s main contributions to machine learning and deep learning systems come from delivering high-performance compute both CPUs and GPUs with an open ecosystem for software development. Machine learning and deep learning systems applications depend on computer hardware that can assist the highest processing capabilities including speed, capacity, and organization at the same time to manage complex data sets from multiple input streams.

AMD’s GPU offer optimized deep learning operations by giving adjustable mixed-precision FP16, FP32, and INT4/INT8 capabilities to meet increasing demand for robust and ever-changing workloads, from training complex neural networks to maintaining inference against those trained networks. AMD is also providing one of the world’s fastest double precision PCIe2 accelerator5.  The AMD Radeon Instinct MI60 is the world’s high-speed binary precision PCIe 4.0 capable accelerator, delivering up to 7.4 TFLOPS peak FP64 performance5. The AMD Radeon Instinct MI50 delivers up to 6.7 TFLOPS FP64 peak performance1 while providing an economical, cost-effective solution that can assist in various deep learning workloads, as well as enabling high reuse in Virtual Desktop Infrastructure (VDI), Desktop-as-a-Service (DaaS) and cloud environments.