what’s new in metal
Generated on 7/31/2024
1 search
What's New in Metal
Improvements in Machine Learning
-
JAX Metal Backend:
- Mixed Precision and Pfloat 16: The JAX Metal backend now supports the Pfloat 16 data type, which is beneficial for mixed precision training. This data type allows for a wide dynamic range of floating point values (Train your machine learning and AI models on Apple GPUs).
- Advanced Array Indexing: Enhanced support for advanced array indexing and update operations, allowing for more efficient manipulation of arrays with a numpy-like syntax (Train your machine learning and AI models on Apple GPUs).
-
TensorFlow and PyTorch:
- Distributed Training and Custom Operations: The Metal backend for TensorFlow and PyTorch supports distributed training, custom operations, and profiling, making it easier to benchmark and improve network performance (Train your machine learning and AI models on Apple GPUs).
-
MPS Graph:
- Transformer Model Efficiency: Improvements to MPS and MPS Graph are focused on enhancing the efficiency of transformer models (Accelerate machine learning with Metal).
Enhancements for Gaming
-
Metal Shader Converter:
- Ray Tracing and Mesh Shaders: The Metal Shader Converter now supports all shader stages, including ray tracing and mesh shaders, which accelerates the porting timeline for games (Port advanced games to Apple platforms).
- Resource Layouts: Helps in porting resource layouts to Metal and includes a runtime header-only library for binding resources to pipelines (Port advanced games to Apple platforms).
-
Resource Residency:
- Simplified Ray Tracing: Residency sets simplify adopting Metal ray tracing by allowing all ray-traced scene resources to be marked as resident as part of a set (Port advanced games to Apple platforms).
-
MetalfX:
- Performance Improvement: MetalfX scales lower resolution images up to improve game performance significantly (Port advanced games to Apple platforms).
General Advancements
- Unified Gaming Platform:
- Console-Class Gaming: With the advancements in Metal and Apple silicon, Apple devices like the latest iPhone 15 Pro and M series Macs can now play console-class games, creating a unified gaming platform (Platforms State of the Union).
Relevant Sessions
Port advanced games to Apple platforms
Discover how simple it can be to reach players on Apple platforms worldwide. We’ll show you how to evaluate your Windows executable on Apple silicon, start your game port with code samples, convert your shader code to Metal, and bring your game to Mac, iPhone, and iPad. Explore enhanced Metal tools that understand HLSL shaders to validate, debug, and profile your ported shaders on Metal.
Platforms State of the Union
Discover the newest advancements on Apple platforms.
Train your machine learning and AI models on Apple GPUs
Learn how to train your models on Apple Silicon with Metal for PyTorch, JAX and TensorFlow. Take advantage of new attention operations and quantization support for improved transformer model performance on your devices.
Accelerate machine learning with Metal
Learn how to accelerate your machine learning transformer models with new features in Metal Performance Shaders Graph. We’ll also cover how to improve your model’s compute bandwidth and quality, and visualize it in the all new MPSGraph viewer.