Intel launched LLM Scaler version 1.0 software for Arc Pro graphics processing units as part of Project Battlematrix infrastructure. The company announced this inference workstation platform during Computex 2025 events. Project Battlematrix serves as comprehensive solution for running multiple Arc Pro GPUs simultaneously. The software delivers substantial performance gains across various model sizes and configurations. Intel designed the container system specifically for Linux operating environments.
Performance metrics demonstrate significant improvements over previous versions. The software achieves 1.8 times faster processing for 40,000 sequence lengths on 32-billion parameter models. Larger 70-billion parameter models show 4.2 times performance increases under similar conditions. Output throughput improves by approximately 10 percent for models ranging from 8 billion to 32 billion parameters. Multi-GPU scaling configurations can achieve performance uplifts reaching 80 percent through optimized data transfers.
The software package incorporates enterprise features such as error correction, virtualization support, and remote firmware management capabilities. Intel plans additional releases throughout the current quarter with enhanced performance optimizations. The company expects to deliver complete feature sets by the fourth quarter of this year.
Performance metrics demonstrate significant improvements over previous versions. The software achieves 1.8 times faster processing for 40,000 sequence lengths on 32-billion parameter models. Larger 70-billion parameter models show 4.2 times performance increases under similar conditions. Output throughput improves by approximately 10 percent for models ranging from 8 billion to 32 billion parameters. Multi-GPU scaling configurations can achieve performance uplifts reaching 80 percent through optimized data transfers.
The software package incorporates enterprise features such as error correction, virtualization support, and remote firmware management capabilities. Intel plans additional releases throughout the current quarter with enhanced performance optimizations. The company expects to deliver complete feature sets by the fourth quarter of this year.
