Home HPC VidAi Video Technologies MyClass

Multi GPU Scalability

Data just grows, and so does the demand for the compute power to process it.
For a scalable system that requires multiple GPUs, it is possible to increasing the GPU units in a system, however, scaling-up a software solution to utilize the processing capability of a multi GPU system is challenging, as data processing models tends to be iterative across multiple data segments, meaning that data needs to be exchanged and transferred between GPUs.



Data Access and transfer challenge has been addressed by Nvidia and Intel GPU interconnect technologies.


As an Nvidia technology partner, and Intel Gold Partner, Photon Computing adopts amazing innovations in production, providing customers with the most advanced GPU to GPU communication technology enabled solutions.


NVIDIA GPU-to-GPU interconnect models