A new version of the ROCm Radeon compute platform has been released with new features and enhancements, such as a GPU cluster networking guide, revised HIP runtime documentation, additional sections on HIP graphs, SOMA, Porting CUDA driver API, and post-installation instructions. The documentation also features updated validation of LLM inference performance on AMD Instinct MI300X, along with new models and the float8 data type. The purpose of these updates is to deliver more precise guidance tailored to diverse user needs and scenarios.

ROCm 6.2.4 Release

The release notes provide a summary of notable changes since the previous ROCm release.

• Release highlights

• Operating system and hardware support changes

• ROCm components versioning

• Detailed component changes

• ROCm known issues

• ROCm upcoming changes

If you’re using Radeon PRO or Radeon GPUs in a workstation setting with a
display connected, continue to use ROCm 6.2.3. See the [Use ROCm on Radeon
GPUs](https://rocm.docs.amd.com/projects/radeon/en/latest/index.html)
documentation to verify compatibility and system requirements.

Release highlights

The following are notable new features and improvements in ROCm 6.2.4. For changes to individual components, see
Detailed component changes.

ROCm documentation updates

ROCm documentation continues to be updated to provide clearer and more comprehensive guidance for
a wider variety of user needs and use cases.

• Added a new GPU cluster networking guide. See
  Cluster network performance validation for AMD Instinct accelerators.
  This documentation provides guidelines on validating network configurations
  in single-node and multi-node environments to attain optimal speed and bandwidth
  in AMD Instinct-powered clusters.

• Updated the HIP runtime documentation.

  • Added a new section on how to use  HIP graphs.

  • Added a new section about the  Stream ordered memory allocator (SOMA).

  • Updated the  Porting CUDA driver API section.

• Updated the  Post-installation instructionswith guidance on using the update-alternatives utility and environment modules to help you manage multiple ROCm
  versions and streamline PATH configuration.

• Updated the  LLM inference performance validation on AMD Instinct
  MI300X
  documentation with more detailed guidance, new models, and the float8 data type.

Operating system and hardware support changes

ROCm 6.2.4 adds support for the  AMD Radeon PRO V710 GPU for compute workloads. See
Supported GPUsfor more information.

This release maintains the same operating system support as 6.2.2.

ROCm components

The following table lists the versions of ROCm components for ROCm 6.2.4, including any version changes from 6.2.2 to 6.2.4.

Click the component's updated version to go to a detailed list of its changes. Click to go to the component's source code on GitHub.

Category	Group	Name	Version
Libraries	Machine learning and computer vision	Composable Kernel	1.1.0
		MIGraphX	2.10
		MIOpen	3.2.0
		MIVisionX	3.0.0
		rocAL	2.0.0
		rocDecode	0.6.0
		rocPyDecode	0.1.0
		RPP	1.8.0
	Communication	RCCL	2.20.5
	Math	hipBLAS	2.2.0
		hipBLASLt	0.8.0
		hipFFT	1.0.15 ⇒  1.0.16
		hipfort	0.4.0
		hipRAND	2.11.0 ⇒  2.11.1
		hipSOLVER	2.2.0
		hipSPARSE	3.1.1
		hipSPARSELt	0.2.1
		rocALUTION	3.2.0 ⇒  3.2.1
		rocBLAS	4.2.1 ⇒  4.2.4
		rocFFT	1.0.29 ⇒  1.0.30
		rocRAND	3.1.0 ⇒  3.1.1
		rocSOLVER	3.26.0 ⇒  3.26.2
		rocSPARSE	3.2.0 ⇒  3.2.1
		rocWMMA	1.5.0
		Tensile	4.41.0
	Primitives	hipCUB	3.2.0 ⇒  3.2.1
		hipTensor	1.3.0
		rocPRIM	3.2.1 ⇒  3.2.2
		rocThrust	3.1.0 ⇒  3.1.1
Tools	System management	AMD SMI	24.6.3 ⇒  24.6.3
		rocminfo	1.0.0
		ROCm Data Center Tool	0.3.0
		ROCm SMI	7.3.0
		ROCm Validation Suite	1.0.0
			Performance	Omniperf	2.0.1
Omnitrace	1.11.2
ROCm Bandwidth Test	1.4.0
ROCProfiler	2.0.0
ROCprofiler-SDK	0.4.0
ROCTracer	4.1.0
	Development	HIPIFY	18.0.0
		ROCdbgapi	0.76.0
		ROCm CMake	0.13.0
		ROCm Debugger (ROCgdb)	14.2
		ROCr Debug Agent	2.0.3
Compilers		HIPCC	1.1.1
		llvm-project	18.0.0
Runtimes		HIP	6.2.4
		ROCr Runtime	1.14.0

Detailed component changes

The following sections describe key changes to ROCm components.

AMD SMI (24.6.3)

Resolved issues

• Fixed support for the API calls amdsmi_get_gpu_process_isolation and
  amdsmi_clean_gpu_local_data, along with the amd-smi set --process-isolation <0 or 1> command. See issue
  #3500 on GitHub.

rocFFT (1.0.30)

Optimized

• Implemented 1D kernels for factorizable sizes greater than 1024 and less than 2048.

Resolved issues

• Fixed plan creation failure on some even-length real-complex transforms that use Bluestein's algorithm.

rocSOLVER (3.26.2)

Resolved issues

• Fixed synchronization issue in STEIN.

ROCm known issues

ROCm known issues are tracked on  GitHub.
Known issues related to individual components are listed in the  Detailed component changessection.

ROCm upcoming changes

The following changes to the ROCm software stack are anticipated for future releases.

rocm-llvm-alt

The rocm-llvm-alt package will be removed in an upcoming release. Users relying on the functionality provided by the closed-source compiler should transition to the open-source compiler. Once the rocm-llvm-alt package is removed, any compilation requesting functionality provided by the closed-source compiler will result in a Clang warning: "[AMD] proprietary optimization compiler has been removed".

rccl-rdma-sharp-plugins

The RCCL plugin package, rccl-rdma-sharp-plugins, will be removed in an upcoming ROCm release.

Release ROCm 6.2.4 Release · ROCm/ROCm