New video memory monitoring is available for NVIDIA and AMD video cards and GPU AI-enabled

The gputaskman utility, designed for detailed monitoring of video cards and processes using the GPU, has been released into the public domain. Unlike conventional monitoring tools, GPU, the program helps you understand not only the overall load and video memory consumption, but also which applications are using the graphics card’s resources. It displays the activity of individual processes, the use of dedicated video memory, and the system RAM available GPU, as well as supported indicators of temperature, power consumption, frequencies, and graphics unit operation. The interface runs directly in the terminal and allows viewing graphs of indicator changes in real time. The project is open-source under the MIT license, and version 0.3.3 is published in the repository.

gputaskman places a special emphasis on local neural networks. The utility can detect software environments in use, including CUDA, HIP, Vulkan and other technologies, as well as recognizing some models launched through llama.cpp, Ollama, and LM Studio. For compatible models, the program evaluates the memory distribution between neural network weights, the KV cache (which stores data from previous tokens when…