GPU compute software platform
AMD ROCm
Provide GPU runtimes, compilers, math and communication libraries for supported AMD workloads.
When it is relevant
- Plan a serving engine or training framework on supported AMD hardware
- Check a coordinated driver, firmware, operating-system and framework combination
Hardware and software assumptions
Documented ecosystem: AMD. Platform support is not a guarantee that every checkpoint, optimization or feature works on every device.
- Confirm the exact GPU, OS, firmware, driver and user-space versions in AMD's matrix
- The serving framework must support the model, precision and required kernels on ROCm
- This platform is not itself a hosted inference endpoint or an Aguasto capacity claim
Operating trade-offs
- CUDA-specific code and optimizations may need porting or a different implementation
- Support differs across devices, operating systems and component versions
Sources & verification
Reference revision 1 · Reviewed . Manufacturer claims and documentation are attributed; Aguasto has not run the cited products in production.
- AMD ROCm official documentation ↗
runtimes, compilers and libraries; AMD hardware scope; vLLM and SGLang integration references. Publication: date not stated in the reviewed source. Verified: 2026-10-12.
- AMD ROCm compatibility matrix ↗
coordinated firmware / driver / user-space version requirements; hardware and operating-system support; framework version matrix. Publication: 2026-08-26. Verified: 2026-10-12.