Graphics Card (GPU)

Overview

A graphics card is a complete subsystem for drawing images, sold as a board that plugs into a PC's expansion slot. Its processor is built for parallelism: where a CPU has a handful of fast cores, a GPU has thousands of small ones that shade pixels and crunch matrices in lockstep. That makes the card the workhorse for games, video, and increasingly for AI and scientific compute, but it also makes it hot and power-hungry, which is why so much of the card is dedicated to feeding and cooling one chip.

Everything mounts to the Main PCB, which carries the GPU Package, the surrounding GDDR Memory Package memory, and a VRM Phase supplying clean core voltage. Heat is pulled away by the Cooler and its Heat Pipe network under a Cooler Shroud, with a Backplate stiffening the board. The card draws power through the PCIe Edge Connector and Auxiliary Power Connector inputs and drives monitors through the Display Output Cluster cluster.

How it works

The GPU die does the actual rendering, streaming geometry and textures from its GDDR Memory Package frame buffer at hundreds of gigabytes per second, the memory sits right beside the package precisely to keep those traces short and fast. Because the die can pull hundreds of watts in bursts, the VRM Phase runs a dozen phases in parallel, each converting the 12 V input down to roughly one volt and sharing the current so no single stage overheats.

All that power becomes heat at one small spot. The cooler's vapor chamber spreads it across the Heat Pipe roots, the pipes carry it into the Fin Stack, and the Cooler fans exhaust it; thermal pads couple the memory and VRM to the same sink. Finished frames leave through the Display Output Cluster ports to the monitor, while the Auxiliary Power Connector sideband pins tell the card how much current its cables can safely deliver before it ramps up.