AMD Finance 26.01 release notes#
1 min read time
This release introduces the popular XGBoost library to the AMD Finance toolkit, enhancing machine learning capabilites for financial workloads.
Release highlights#
XGBoost is fully accelerated on AMD Instinct MI300X and MI325X GPUs (gfx942) through ROCm 7.1.1, 7.0.2, or 6.4.4. This integration optimizes kernels, improves memory management, and enables seamless multi-GPU scaling on supported AMD GPUs.
Key benefits of this integration include:
Hardware acceleration: Parallelizes split-finding on the GPU to deliver significant performance gains over CPU-only baselines.
Financial risk optimization: Accelerates large-scale tasks such as credit risk scoring, fraud detection, and loan default modeling.
Compliance and precision: Increases throughput while maintaining the native robustness (L1/L2 regularization) and missing value handling of XGBoost.
Note
AMD Finance libraries introduced in the 25.11 release, LightGBM and ThunderGBM, are currently only supported on MI300X GPUs, and ROCm 7.0.2 and 6.4.4.
Supported platforms#
To learn about hardware and software environment requirements, see the AMD Finance compatibility matrix and the AMD Finance installation prerequisites.
AMD Finance components#
Component |
Version |
Source |
|---|---|---|
3.1.1 |
||
4.6.0.99 |
||
0.3.16 |