Graphics workloads need memory for textures, render targets, commands, and other data. Dedicated and shared memory labels describe different resource arrangements, not two universally interchangeable quantities that can simply be added into one performance figure.
How the pieces connect
Some processors use memory attached specifically to the graphics device. Others rely more directly on system memory, and operating systems can manage additional sharing or movement between resources. Addressability, bandwidth, latency, and scheduling affect how effectively a workload uses the available space. A capacity label says little about those other constraints by itself.
An everyday example
An integrated graphics processor can use part of the computer's ordinary memory for image data. A discrete device may have its own local memory but still interact with system allocations. An information panel can therefore show several values representing budgets, committed resources, or supported limits rather than a single physical pool.
Reading the information in context
More reported memory does not ensure that a workload runs faster, particularly when compute capability or bandwidth is the limiting factor. Some values are dynamic and depend on current use. Compare like-for-like definitions when reading specifications. Do not infer that changing a software preference can turn shared system memory into physically attached high-bandwidth graphics memory.
Display & Graphics
Note these before you change anything
Useful facts to have to hand while reading about dedicated and shared graphics memory.
- Current driver version and date from the adapter's Driver tab
- Resolution, refresh rate and scaling for each monitor
- Which port and cable each display uses
- Any per-application graphics settings you changed
Display & Graphics: Drawing a frame (output)
- Applicationrequests drawing
- Graphics APIfor example DirectX or Vulkan
- Display driveruser and kernel parts
- GPUrenders the frame
- Monitorover the display cable
