A Preliminary Study on Simultaneous Coscheduling for Discrete GPU vs. Fused GPU
2026-08-10 • Distributed, Parallel, and Cluster Computing
Distributed, Parallel, and Cluster Computing
AI summaryⓘ
The authors studied how running a program on both a CPU and GPU at the same time works on two different systems: one with separate CPU and GPU parts (H100 PCIe) and one where they are combined (NVIDIA GH200 Superchip). They tested how splitting work between CPU and GPU and managing memory affects speed and ease of programming using a specific math problem (sparse conjugate gradient). Their results show that the combined chip (GH200) allows for more flexible ways to share tasks and makes automatic memory management more practical. This suggests that systems with integrated CPU and GPU might help programs run better and be easier to write when using both processors together.
CPU-GPU coschedulingNVIDIA GH200 SuperchipH100 PCIe platformsparse conjugate gradientworkload partitioningmemory managementexplicit copymanaged memorymapped memoryhybrid CPU-GPU execution
Authors
Poorna Gunathilaka, Nabayan Chaudhury, Kirshanthan Sundararajah, Wu-chun Feng
Abstract
CPU-GPU coscheduling enables simultaneous execution of an application across both processing units, but its efficiency depends on workload partitioning and memory architecture. This preliminary study evaluates coscheduling on the NVIDIA GH200 Superchip compared to a discrete H100 PCIe platform. Using sparse conjugate gradient (CG) as a case study, we assess various work divisions across three memory-management paradigms: explicit copy, managed memory, and mapped memory. Our evaluation highlights the run time and programmability tradeoffs of reducing manual CPU-GPU data movement. The results show that compared with the H100 PCIe platform, GH200 makes several hybrid CPU-GPU work divisions competitive and makes managed memory practical for several matrices. These results suggest that integrated CPU-GPU platforms such as GH200 can improve both performance and programmability for coscheduled workloads.