This week we introduced Radeon Instinct, a series of accelerators designed to propel machine intelligence forward and offer organizations powerful GPU-based solutions for deep learning, inference, and training.
For us, entering this space is literally about changing and improving our world. The applications are profound! It’s not just about leveraging the compute capabilities in our GPUs to study the staggering amount of daily data we produce; it’s also about aiding scientists in curing major diseases like cancer, HIV, and Type 1 Diabetes. It’s about improving machine intelligence to analyze and understand our vast universe, or enabling Nano robots to perform critical surgeries inside our bodies. It’s about turning us into better artists, better researchers, and better humans.
But as we’ve said before, great hardware is only half of the story.
Right now the datacenter is built on open source software. The implications and applications of machine intelligence are too important and too vast to be dominated by a single company with a proprietary ecosystem. The entire community needs to be participating. That’s why we purposely created the Radeon Instinct ecosystem as an open solution from the metal forward. We’re investing in open source software, open standards, and open interconnects that go beyond today’s PCIe Gen3 performance barrier.
That began with our ROCm (Radeon Open Source) platform, which brings a rich, highly scalable foundation to advanced computing by better integrating the CPU and GPU to help solve complex, real world issues. It’s also programming-language independent, and is the foundation of the coming evolution of machine intelligence problem sets.
Consider a cosmologist like University of Toronto’s Keith Vanderline , who is already using AMD’s machine intelligence and deep learning technologies and is eager to see the benefits afforded by Radeon Instinct. Keith is measuring massive chunks of the universe, so he needs insane amounts of computational power. The systems he uses to analyze these huge volumes of existence generate 1 Terabyte of data per second. 24 hours a day, 7 days a week, 365 days per year. Petaflops of continuous processing power are needed to turn the massive amount of data collected into something useful, and our platform allows him to continue scaling up as needed.
The following is something that Denise Dumas, vice president, Platform Engineering at Red Hat told us during our ROCm launch earlier this year, but it’s worth repeating as head toward a 1H 2017 launch of Radeon Instinct: “An open source approach to heterogeneous computing can help bring the benefits of high-performance computing directly to developers, providing the flexibility for them to use available compute resources and extract the best possible performance from their available hardware. As a leader in helping to nurture and maintain open source communities, we are pleased to see that AMD has elected to take a community-focused approach to ROCm.”
We also unveiled MIOpen, a free, open-source library for GPU accelerators. Its purpose is to provide GPU-tuned implementations for standard routines such as convolution, normalization, tensor format, pooling, and activation functions. MIOpen is planned to be available in Q1 2017 alongside a 1H 2017 launch of our Radeon Instinct products.
The Immersive Era of consumer computing will have a profound impact on our personal lives, our professional lives, and our collective experiences as a species. We’re thrilled to be entering a space where we can help make that impact a positive one, and enable a global community of brilliant minds to change the world.
Jason Evangelho is Sr. Technical Marketing Specialist for the Radeon Technology Group at AMD. His postings are his own opinions and may not represent AMD’s positions, strategies, or opinions. Links to third party sites, and references to third party trademarks, are provided for convenience and illustrative purposes only. Unless explicitly stated, AMD is not responsible for the contents of such links, and no third party endorsement of AMD or any of its products is implied.