AI inference is pushing technology limits and have ushered in the era of domain specific architectures, moving on from CPUs and fixed hardware accelerators, in order to meet performance requirements.
However, as AI continues to evolve at pace, it is creating a number of challenges. Nick Ni, Director of Product Marketing, AI, Software, Ecosystem, at Xilinx highlights these challenges and how Xilinx are meeting them. See below.
Challenges with AI productisation
1. Neural networks are evolving fast
2. Vendor’s TOPS officially fake news
3. AI application is much more than AI
4. Functional safety and security are critical
5. Limited options for low cost applications with AI
How Xilinx addresses AI productisation challenges
1. Adapatable to the latest neural networks
2. Highest ‘useable’ TOPS efficiency
3. Xilinx is long proven to accelerate the whole application
4. Xilinx is proven as a safety critical chip supplier
5. Low cost to high end with robust AI inference
