Lambda GH200
A GPU cloud server instance offered by Lambda, equipped with the NVIDIA GH200 Grace Hopper Superchip, which integrates an ARM CPU and an H100 GPU. It is intended for large-scale AI training.
Timeline (last 90 days)
The training scale is 4096 games in parallel, about 330,000 agent steps per second end-to-end, completing about 19 billion agent steps in about 30 hours on a single Lambda GH200.
Compute used a single Lambda GH200 (64 ARM cores plus an H100 96GB), running 15,360 games in parallel, with throughput of 1.2 to 1.9 million samples per second.
All Facts (2)
The training scale is 4096 games in parallel, about 330,000 agent steps per second end-to-end, completing about 19 billion agent steps in about 30 hours on a single Lambda GH200.
50%UnverifiedCompute used a single Lambda GH200 (64 ARM cores plus an H100 96GB), running 15,360 games in parallel, with throughput of 1.2 to 1.9 million samples per second.
50%