Hotchip Dojo System
Dojo Super-Compute System Scaling for ML Training
Flexible Building Block
V1 Dojo Interface Processor
- 32GB High-Bandwidth Memory 800 GB/s Total Memory Bandwidth
- 900 GB/s TTP Interface Provideds full DRAM bandwidth to Training Tile
- 32 GB/s Gen4 PCIe Interface
- 160GB total DRAM per Tile edge Shared memory for training tiles
- 5 DIP Cards provide Max Bandwidth 4.5 TB/s aggregate bandwidth to DRAM over TTP
- 80 Lanes PCIe Gen4 Interface Provide standard connectivity to hosts
Remote DMA Topology
V1 Dojo Training Matrix
Dissaggregated Scalable System
Dojo Supercomputer for ML Training
