CUDA ENGINEERING

SCENE 02 · INSIDE THE MACHINE

Inside the machine

Six engines called SMs share one die. Click one open and look inside.

GPU · DEVICESM 0 — two warp schedulers, shared memorySM0W0 W1 · LANESW0 W1 · LANESSM 1 — two warp schedulers, shared memorySM1W0 W1 · LANESW0 W1 · LANESSM 2 — two warp schedulers, shared memorySM2W0 W1 · LANESW0 W1 · LANESSM 3 — two warp schedulers, shared memorySM3W0 W1 · LANESW0 W1 · LANESSM 4 — two warp schedulers, shared memorySM4W0 W1 · LANESW0 W1 · LANESSM 5 — two warp schedulers, shared memorySM5W0 W1 · LANESW0 W1 · LANESGLOBAL MEMORY · SHARED BY ALL SMs
STREAMING MULTIPROCESSOR

SM

Two schedulers drive warps of 32 lanes in lockstep.

W0
W1

Every lane fires the same instruction each cycle. Click elsewhere to close.

SMWARP SCHEDULERSHARED MEMORY
← 01 · Why a GPU?
03 · What happens when you launch? →
Agent-native web · research mode

Experimental WebMCP · evolving community-group draft, not a W3C Standard or on the Standards Track.

Native status: Initializing. Registered: none

DOM → inferred capabilities · WebMCP → declared capabilities. Tools read real CUDA/Inference Atlas content. Calls may highlight temporary UI state. No backend or remote telemetry.

Local harness verifies application logic; it does not demonstrate native discovery or a browser agent.

Input schema
{
  "type": "object",
  "properties": {
    "query": {
      "type": "string",
      "minLength": 1,
      "maxLength": 300
    },
    "limit": {
      "type": "integer",
      "minimum": 1,
      "maximum": 10
    }
  },
  "required": [
    "query"
  ],
  "additionalProperties": false
}
Local log: last 100 calls while this panel is open. Inputs may contain personal text; review before sharing.
 
[]