How Microsoft’s Open-Source GPU Task Manager Reveals What Uses Your VRAM
AIThis post was created with the assistance of artificial intelligence (AI).

TL;DR

Microsoft and GitHub executive Scott Hanselman developed GPU Taskman, an open-source terminal utility that brings GPU telemetry and per-process resource information together. It can also identify supported local AI models and estimate parts of their memory use, though the project is experimental and some figures are not direct measurements.

Scott Hanselman, a Microsoft and GitHub vice president and member of technical staff, has developed GPU Taskman, an open-source terminal utility that shows GPU memory use alongside per-process activity and hardware telemetry. The tool aims to make it easier to see which applications are using GPU resources, including supported local AI runtimes, while its developers continue to validate capabilities across different hardware.

For playersOffer from Amazon

Play games on Amazon Luna for your game nights, included with Prime

  • A rotating selection of games, no download needed
  • Play on TV, laptop or phone
  • Fast, free delivery for your gear, too
Start playing with Prime Free trial · Cancel anytime
As an affiliate, we earn on qualifying purchases.

Hanselman said in an November 3 post on X that he wanted better visibility into what was consuming his GPU memory and invited developers to contribute. GPU Taskman presents VRAM consumption, GPU-accessible system memory and individual process activity in one terminal interface. It also reports metrics such as GPU utilization, temperature, power draw, clock speeds, memory activity and PCIe statistics.

The utility can help distinguish workloads such as desktop compositing, browsers, games and AI applications. Its project documentation says it recognizes llama.cpp, Ollama and LM Studio, with the amount of detail dependent on the runtime and information available. For supported models, it may show architecture, quantization, parameter counts and information about model weights and estimated KV-cache requirements.

GPU Taskman combines operating-system GPU counters with process metadata, model files, runtime logs and local APIs. That can help separate estimated memory for model weights from KV-cache and compute buffers, but some model-specific figures are estimates, not directly measured GPU allocations. The software is available under the MIT license through GitHub Releases and PyPI; portable Windows builds do not require a separate Python installation.

At a glance
updateWhen: Introduced in an X post on November 3;…
The developmentScott Hanselman introduced GPU Taskman, an open-source utility for examining GPU memory, process activity, hardware telemetry and supported local AI workloads.
Top Steam deals right now
Booster Pack Heroes-30%$6.99
Mortal Shell II-25%$37.49
Casualties: Unknown-15%$19.54
Find The Needle-15%$12.74
Horses of Hoofprint Bay-15%$22.94
Warhammer Survivors-10%$7.19
DREADMOOR-10%$17.99
Probably Stolen – Cyberpunk Shopkeeper Sim-10%$17.99
Live · Steam store (current discounts)
GPU monitoring · Briefing

See what uses your VRAM with GPU Taskman

Microsoft and GitHub executive Scott Hanselman developed this open-source terminal utility to bring GPU telemetry, process activity and supported local AI model details together.

0.3.5
Version added Intel GPU support on Linux
3
Named AI runtimes: llama.cpp, Ollama and LM Studio
Nov 10
Version 0.3.5 published

What it brings together

A single terminal view can help users investigate which workloads are using GPU resources.

Processes

Per-process activity

See GPU memory use alongside activity from applications such as browsers, games and AI tools.

Telemetry

Hardware readings

View utilization, temperature, power draw, clock speeds, memory activity and PCIe statistics.

AI models

Runtime details

For supported models, it may show architecture, quantization, parameter counts and estimated memory needs.

Supported environments

Capabilities vary by operating system, GPU and runtime.

EnvironmentProject support
Windows 10 and 11NVIDIA and AMD GPUs
LinuxNVIDIA and AMD; Intel GPU support added in version 0.3.5
WSLListed, with features varying
Apple Silicon macOSListed, with features varying

How to read its memory clues

1
Check process use

Compare GPU memory and activity across the applications running on your system.

2
Inspect supported AI workloads

Model weights, KV-cache needs and compute buffers may be shown, but some figures are estimates.

3
Validate on your setup

Check which metrics work with your GPU, operating system and workload.

CaveatGPU-accessible system RAM does not prove VRAM has overflowed. The tool cannot identify individual game textures or specify exactly which resources were evicted.
The takeaway
GPU Taskman offers useful clues about GPU and AI memory use, but its experimental readings—especially estimates—need checking against your own setup.

A Closer View of GPU Memory

GPU Taskman addresses a practical monitoring gap: seeing a total VRAM figure does not always explain which processes account for it or what type of workload is involved. By bringing process-level use and hardware readings together, the tool could help developers and PC users investigate high memory use, compare workloads and identify when a local AI application is a major contributor.

That visibility may also help when a game or other demanding program stutters. GPU memory pressure can contribute to performance problems, but a reading showing GPU-accessible system RAM is not proof that VRAM has overflowed. Hanselman cautions that the utility cannot identify individual game textures or establish exactly which resources have been evicted. Its readings are useful diagnostic clues, not a complete account of every allocation or its performance effect.

Breville Barista Express Espresso Machine with Built-In Grinder, Brushed Stainless Steel

Breville Barista Express Espresso Machine with Built-In Grinder, Brushed Stainless Steel

Hisense 43-Inch E6 Cinema Series Hi-QLED 4K UHD Smart Fire TV (43E6QF)

Hisense 43-Inch E6 Cinema Series Hi-QLED 4K UHD Smart Fire TV (43E6QF)

KitchenAid Artisan Series 5-Quart Tilt-Head Stand Mixer, Juniper (KSM150PS)

KitchenAid Artisan Series 5-Quart Tilt-Head Stand Mixer, Juniper (KSM150PS)

As an affiliate, we earn on qualifying purchases.

From Windows Tool to Multi-Platform Project

Windows Task Manager already reports per-process GPU memory statistics. GPU Taskman’s stated distinction is its combination of those kinds of process details with broader telemetry and, for supported AI software, model-related information. The project is aimed at people who want more detail than a single memory total, rather than establishing that Windows has no GPU monitoring.

The project lists support for NVIDIA and AMD GPUs on Windows 10 and 11, as well as Linux, Windows Subsystem for Linux and Apple Silicon macOS, with features varying by platform and hardware. Version 0.3.5, published November 10, added Intel GPU support on Linux, CPU and system RAM monitoring, and more options for recording GPU statistics. The rapid release cadence and experimental status mean users should check platform-specific support rather than assume every metric works everywhere.

“Hanselman said he wanted better visibility into what was consuming his GPU memory and invited other developers to contribute.”

— Scott Hanselman, in a post on X on November 3

Limits of the Current Readings

GPU Taskman remains experimental, and the project’s capabilities need further validation across hardware configurations. Support and detail vary by operating system, GPU and AI runtime. The available information does not establish how consistently every listed metric performs across those combinations.

It is also not yet clear how accurate model-specific estimates are for every runtime or configuration. The tool can help identify resource use, but it does not expose every allocation: system RAM accessible to the GPU does not by itself confirm VRAM overflow, and the program cannot trace a game’s individual textures or specify precisely which resources were evicted.

Development and Hardware Validation

The project is available for users and developers to try through its GitHub Releases page or PyPI, and Hanselman has invited contributions. Version 0.3.5’s additions show development continuing, including expanded Linux GPU support and recording options.

No schedule for a stable release or specific validation milestone is given. The next practical test for users is whether the metrics they need are supported and reliable on their own GPU, operating system and workload. Further releases may add capabilities, but the project’s current status calls for treating its readings—especially estimated AI memory breakdowns—as diagnostic information rather than definitive measurements.

Source: Wccftech

Key Questions

What does GPU Taskman show?

It combines GPU memory use, process activity and hardware telemetry, including utilization, temperature, power, clock speeds and PCIe statistics. The precise information available varies by platform and hardware.

Can it show which AI models are using VRAM?

It recognizes supported runtimes including llama.cpp, Ollama and LM Studio. For some workloads it can report model details and estimate memory categories, but support and detail vary, and some figures are not direct measurements.

Does GPU-accessible system RAM mean VRAM has run out?

No. Hanselman cautions that GPU-accessible system RAM use does not necessarily indicate VRAM overflow. GPU Taskman also cannot identify individual game textures or determine precisely which resources have been evicted.

Which platforms and GPUs does it support?

The project lists NVIDIA and AMD support on Windows 10 and 11, plus Linux, WSL and Apple Silicon macOS, with capabilities differing by setup. Version 0.3.5 added Intel GPU support on Linux.

Is GPU Taskman stable software?

No. It is described as experimental software, and its monitoring capabilities still need validation across hardware configurations. Users can get it from GitHub Releases or install it through PyPI.

Source: Wccftech

You May Also Like

Is The Alienware Aurora ACT1250 With Core Ultra 9 285 And RTX 5070 Ti Worth $2,300?

Slickdeals lists an Alienware Aurora with Core Ultra 9 285, RTX 5070 Ti, 32GB RAM and 1TB SSD for $2,299.99, with free pickup.

14 Best Mechanical Keyboards For Gaming, Work, And Everyday Typing In 2027

A 2027 roundup compares 14 mechanical keyboards, led by the AULA F75 Pro, with picks for office typing, wired gaming and number-pad workflows.

PCIe 5.0 Doubles PCIe 4.0’S Bandwidth, But Your Graphics Card Barely Cares

PCIe 5.0 doubles the theoretical bandwidth of PCIe 4.0, but tests show little gaming difference for most current graphics cards.

Is A Used RTX 3090 Still The Best-value GPU For Local AI?

The RTX 3090’s 24GB of VRAM suits local AI models, but used-card risks and shifting prices complicate its value against newer GPUs.