The news
Microsoft engineer Scott Hanselman released a utility that reports GPU usage broken down by individual process. The tool answers a direct question: which running programs are consuming graphics hardware and by how much. The announcement provides no further details on distribution method, licensing terms, or supported hardware.
Context
Developers and power users have operated without a focused view of per-process GPU activity for years. Standard system monitors aggregate activity or require multiple steps to isolate contributors. Hanselman created the utility to meet an internal need at Microsoft before making it available more widely. The single public source describes the purpose but supplies no timeline for release, no code repository link, and no indication of ongoing support.
Details
The program produces a list of active processes paired with their measured GPU utilization. It narrows attention to per-process attribution rather than overall system metrics. No sampling interval, no list of compatible GPU vendors, and no performance characteristics are stated in the available description. The source consists of a brief statement confirming the tool’s existence and core function.
Because the announcement contains so little technical information, readers cannot yet determine whether the utility works with integrated graphics, discrete cards from multiple vendors, or only specific driver versions. Questions about installation method, output format, and command-line options remain open until the code itself appears.
Reactions / counterpoints
No reactions from other engineers or Microsoft product teams appear in the source. The absence of discussion leaves unclear whether the utility will be adopted inside Microsoft or maintained as a personal project. Users who encounter similar gaps in existing monitors will have to test the tool once it is published to judge its accuracy against other measurement methods.
Why it matters
Engineers who run mixed workloads on a single machine—local model training, video encoding, browser-based applications, and background services—need a fast way to assign responsibility for GPU load. Aggregate counters hide the offender and force repeated context switching between tools. A narrow utility that isolates the contribution of each process reduces that friction even if it offers nothing else.
The release also illustrates a common pattern: small internal scripts that solve repeated daily problems sometimes surface publicly when an engineer decides the effort to clean them up is justified. Such tools rarely receive formal product support, so longevity depends on individual interest. For anyone who has spent time guessing which application is driving fan noise or throttling other work, the existence of a direct answer outweighs the lack of polish or documentation.
In practice, the value will be measured by whether the output proves reliable across common driver stacks and whether the code is easy to inspect and modify. Until the utility is available, the announcement functions mainly as a signal that the underlying visibility gap remains noticeable enough for an experienced Windows developer to address it.
---
Sources:
No comments yet