Carding 4 Carders
Professional
- Messages
- 2,724
- Reaction score
- 1,588
- Points
- 113
The company's tests showed NVIDIA's true place in the performance race.
Numenta has made a breakthrough in artificial intelligence by demonstrating that Intel Xeon processors can significantly outperform the best existing processors in AI tasks, using an innovative approach. Based on a number of techniques combined under the Numenta Platform for Intelligent Computing (NuPIC) brand, the company was able to achieve new levels of performance on traditional AI processors, according to the Serve the Home portal.
The most striking thing is that the developed approach allows you to surpass the performance of CPUs and GPUs specifically designed for working with artificial intelligence. For example, Numenta compared the performance of its solution to the Nvidia A100 GPU, using Intel's advanced 4th-generation 48-core Sapphire Rapids processor. In all tests, the improved processor showed better performance compared to the Nvidia chip in terms of overall throughput. The results showed that the new approach is 64 times faster than the 3rd generation Intel Xeon processor and 10 times faster than the A100 GPU.
Numenta, a company known for its neurobiological approach to artificial intelligence workloads, relies heavily on the idea of sparse computing – this is how the brain forms connections between neurons.
Most modern central processing units and graphics processors are designed for dense computing, which is a more "rough" approach compared to contextual data processing used in the brain. The use of sparse computing can significantly improve performance, but traditional processors cannot operate efficiently in this mode without specialized optimizations.
Numenta's solution is based on the application of Advanced Vector Extensions (AVX)-512 and Advanced Matrix Extensions (AMX) technologies to Intel Xeon processors. Technologies are extensions of the x86 architecture that serve as additional instruction sets that allow processors to perform more demanding functions.
Numenta's NuPIC service is delivered as Docker containers and can be run on companies ' own servers. The technology can be an optimal solution for reusing already deployed processors in data centers for AI tasks, especially given the long waiting times for deliveries of the leading Nvidia A100 and H100 GPUs.
Numenta has made a breakthrough in artificial intelligence by demonstrating that Intel Xeon processors can significantly outperform the best existing processors in AI tasks, using an innovative approach. Based on a number of techniques combined under the Numenta Platform for Intelligent Computing (NuPIC) brand, the company was able to achieve new levels of performance on traditional AI processors, according to the Serve the Home portal.
The most striking thing is that the developed approach allows you to surpass the performance of CPUs and GPUs specifically designed for working with artificial intelligence. For example, Numenta compared the performance of its solution to the Nvidia A100 GPU, using Intel's advanced 4th-generation 48-core Sapphire Rapids processor. In all tests, the improved processor showed better performance compared to the Nvidia chip in terms of overall throughput. The results showed that the new approach is 64 times faster than the 3rd generation Intel Xeon processor and 10 times faster than the A100 GPU.
Numenta, a company known for its neurobiological approach to artificial intelligence workloads, relies heavily on the idea of sparse computing – this is how the brain forms connections between neurons.
Most modern central processing units and graphics processors are designed for dense computing, which is a more "rough" approach compared to contextual data processing used in the brain. The use of sparse computing can significantly improve performance, but traditional processors cannot operate efficiently in this mode without specialized optimizations.
Numenta's solution is based on the application of Advanced Vector Extensions (AVX)-512 and Advanced Matrix Extensions (AMX) technologies to Intel Xeon processors. Technologies are extensions of the x86 architecture that serve as additional instruction sets that allow processors to perform more demanding functions.
Numenta's NuPIC service is delivered as Docker containers and can be run on companies ' own servers. The technology can be an optimal solution for reusing already deployed processors in data centers for AI tasks, especially given the long waiting times for deliveries of the leading Nvidia A100 and H100 GPUs.
