Intel and NVIDIA are reportedly collaborating on a custom Xeon CPU that incorporates NVLink technology. a high-bandwidth interconnect traditionally used between NVIDIA GPUs and CPUs.
The development was reaffirmed alongside new details about Intel’s server roadmap, which includes:
-
Diamond Rapids (next-gen Xeon server CPUs)
-
16-channel memory
-
Coral Rapids with high memory bandwidth
-
SMT Nova Lake in Q4 targets
This custom Xeon with NVLink suggests deeper CPU-GPU integration possible in future server platforms, potentially improving data transfer speeds for AI and HPC workloads compared to PCIe alone.