What is a Graphics/GPU Processing Cluster?
Abbreviated GPC
A Graphics/GPU Processing Cluster (GPC) is one of the largest physical building blocks on an NVIDIA GPU die: a group of Texture Processing Clusters, themselves groups of Streaming Multiprocessors, plus a dedicated raster engine for turning triangles into pixels. That raster engine is a reminder these are still graphics chips underneath everything: GPC used to stand for "Graphics Processing Cluster," and NVIDIA's documentation now more often expands it as "GPU Processing Cluster" instead.
On compute capability 9.0 chips like the H100, a GPC also gains a role in the CUDA programming model's thread hierarchy: a "cluster" of thread blocks can be scheduled across the SMs inside one GPC, the way a single block's threads are scheduled onto one SM, with its own slice of the memory hierarchy in distributed shared memory. Worth knowing before comparing an H100 or newer GPU architecture to older cards on core count alone.
Building on GPUs? Aquanode runs the workload.
Deploy on H100, H200, B200, A100 and MI300X across a multi-provider marketplace, without racking your own hardware or committing to one cloud's spec sheet.
See also
Streaming Multiprocessor
A Streaming Multiprocessor is the closest thing an NVIDIA GPU has to a CPU core, but far simpler and far more numerous. How SMs trade per-thread sophistication for massive parallelism and near-free context switching.
GPU Architecture
How NVIDIA GPUs are actually built, from Graphics Processing Clusters and Streaming Multiprocessors down to the memory hierarchy, and how Ampere, Hopper, and Blackwell differ.
