- Joined
- Aug 19, 2017
- Messages
- 3,233 (1.12/day)
In a blog post, AMD's Executive Vice President and Chief Technology Officer, Mark Papermaster announced that AMD is officially joining Compute Express Link (CXL) Consortium. Previously AMD has joined different working groups that are addressing the problem of cache coherency inside various systems and accelerators, like CCIX, OpenCAPI and Gen-Z. CXL is initiative started by Intel that promises better solution to cache coherency problem and many companies have joined, now including AMD.
As workloads are split into different segments like machine learning and graphics processing that require a mix of scalar and vector processing, accelerators for those workloads need better inter connection to other processors such as CPUs, where coherency and memory semantics are one of the biggest challenges. AMD's goal of joining the CXL working group is to help advance the standard and possibly integrate it into future products. making the concept of heterogeneous computing a bit closer to reality.
View at TechPowerUp Main Site
As workloads are split into different segments like machine learning and graphics processing that require a mix of scalar and vector processing, accelerators for those workloads need better inter connection to other processors such as CPUs, where coherency and memory semantics are one of the biggest challenges. AMD's goal of joining the CXL working group is to help advance the standard and possibly integrate it into future products. making the concept of heterogeneous computing a bit closer to reality.
View at TechPowerUp Main Site
