Understanding the Limits of the GPU

Brief overview to understand the limits of the GPU or video card

The number of concurrent users that a given GPU can support is dependent on both the available VRAM on the card and the use case being considered. For example, an average CT series may contain 250 slices and the memory on the graphics card being used is a 4GB Quadro K5000. How many concurrent users are possible with this configuration?

  • Each image contains 512*512(pixels) and each pixel is 2 Bytes.

  • Each image is 512*512*2 = 524'288 Bytes = 0.524MB.

  • Each series is 0.524MB*250(images) = 131 MB.

Theoretically. this card could support 4000MB(graphic card memory) / 131 MB (series) = 30 concurrent users.

In a different scenario, consider the same situation as before, but each series now contains 1000 images.

  • Each image contains 512*512(pixels) and each pixel is 2 Bytes.

  • Each image is 512*512*2 = 524'288 Bytes = 0.524MB.

  • Each series is 0.524MB*1000(images) = 524MB.

Theoretically, this card could support 4000MB(4GB Graphics card) / 524MB (series) = 7 concurrent users. In reality, 7 concurrent users may fail. This could be a result of memory allocated to the functionality of the OS, as well as any error checking the graphics card may have incurred.

In conclusion, the number of concurrent users is dependent on the graphics card’s available memory, as well as how many slices are in each series.