How to Track GPU Performance in Windows 10
Microsoft added a bunch of smaller features and improvements to the Windows 10 Fall Creators Update.
One of the new features displays the GPU performance in the Task Manager provided that a compatible driver is installed. On Windows 10, the GPU information is available in Task Manager using the Windows Display Driver Model (WDDM) version 2.0 or later, which is a special driver architecture that video cards must support to render the desktop and apps on the screen.
You can launch the Windows Task Manager with the shortcut Ctrl + Shift + Esc. If you're in the compact mode, click the More details button.
GPU performance is displayed as a column under Processes tab.
If you don't see these columns, right-click a column, and check the GPU and GPU Engine options.
If your machine supports WDDM version 2.0 or later, the Performance tab will list your GPU in the left pane. In the case that you have multiple GPUs, each one will be named using a number that corresponds to its physical location. For example, GPU 0, GPU 1, GPU 2, etc.
The GPU section includes current information regarding the GPU engines, not individual GPU cores. (It's worth to define that a GPU engine is made up of many GPU cores.)
Task Manager, by default, will display the four most interesting GPU engines, which typically can include 3D, Copy, Video Decode, and Video Processing, but you can change these views by clicking the name and picking another engine. Handy to find out if the GPU is a bottleneck on the device; this can be the case if the available memory is maxed out whenever you play a game, or if GPU usage hits the 100% load mark regularly.
You can even change the graph view to a single engine by right-clicking anywhere within the section and selecting the Single engine option from the menu.
Immediately after the engines graphs, you'll find the video memory utilization and summary.
Task Manager shows two types of video memory, including dedicated and shared memory.
The dedicated memory is the memory that will only be used by the graphics card. Usually, this is your VRAM on discrete cards or the amount of memory a computer is configured to explicitly reserved for the integrated graphics card, but the CPU can still use it.
At the bottom right, you'll also notice the "Hardware reserved memory," which represents the amount of memory reserved for the video driver.
The amount of dedicated memory in this section represents the total amount of memory actively being utilized across the processes, while the amount of shared memory in this section represents the amount of system memory consumed for graphics.
Also, in the left pane, under the GPU name, you'll see the current total aggregated performance utilization.
By default, the "Details" tab doesn't display any GPU information, but you can always right-click a column, click the Select columns option, and check enable these options:
- GPU Engine.
- Dedicated GPU memory.
- Shared GPU memory.
When using the "Details" tab, you just need to be aware that adding the used memory by each process can end up being larger than the total available memory as the shared memory will be counted multiple times. This information is useful to understand the memory usage per-process, but you should be using the "Performance" tab to see a more accurate overall video utilization.
Did you find this tutorial helpful? Don’t forget to share your views with us.