Features

One optimization layer for GPUs and CPUs

Slow AI workloads can be constrained by GPU kernels, CPU waits, hardware faults or delayed collective operations. Zymtrace brings execution and health evidence together with recommended fixes, so your team can address the causes rather than just the symptoms.

Cluster-wide
Self-hosted
Multi-silicon

Start optimizing Talk to an engineer

Profiling

Where the time goes, on GPUs and CPUs

NVIDIA GPU profiling CUDA kernels, SASS and stall reasons AMD GPU profiling HIP kernels, instructions and stall reasons Inference profiling Engine metrics linked to kernels Model shapes Module trees, tensor shapes, layer attribution CPU profiling Whole-system eBPF, all major languages

Diagnostics

Problems Zymtrace finds for you

Recommendations Your profiles, analyzed for you GPU Straggler Detector The GPU that slows down training GPU topology & health NVLink, XID errors, NUMA and GPU metrics

Get started now

Zymtrace is fully self-hosted. 5 minutes is all you need to get it up and running.

Start optimizing Talk to an engineer