BIRENTECH BR100 series general-purpose GPU chip

Ultra-high performance and energy efficiency ratio: core performance reaches the world's leading level, achieving more than 3 times performance improvement compared to mainstream products on the market

Advanced process: Using 7nm process and innovatively applying Chiplet and 5D CoWoS packaging technology, taking into account high yield and high performance

Advanced interface system: supports PCIe 5.0 interface technology and CXL communication protocol, with bidirectional bandwidth up to 128 GB/s

Advanced interconnection system: Original BLink™ high-speed GPU interconnection technology, single-card interconnection bandwidth up to 448 GB/s, and supports single-node 8-card full interconnection

Multi-data precision support: In addition to natively supporting mainstream data precisions such as FP32, BF16, FP16, and INT8, the original definition of TF32+ data precision provides higher data precision and throughput performance than TF32.

Secure Virtual Instance (SVI): supports up to 8 independent instances, each instance is physically isolated and equipped with independent hardware resources, and can run independently