Skip to content
Rota Nacional

Radar ·

PufferLib 4.0 reports multi-GPU scaling results

A June 21, 2026 publication reports 6.6× scaling on eight GPUs in a worst-case test and throughput above 100 million steps per second on eight RTX 5090 GPUs.

A publication dated June 21, 2026 presents multi-GPU scalability results for PufferLib 4.0. In the described worst-case test, a policy with 32 times as many parameters achieved 6.6× scaling on eight GPUs. The publication also reports peak throughput above 100 million steps per second on eight RTX 5090 GPUs.

These figures offer a performance reference for GPU workloads, not a guarantee for other configurations. To assess them, consult the original publication and verify the test conditions, what “worst case” means, the metrics, and the reported hardware; compare only with measurements made under equivalent conditions.

Get new articles

Privacy, AI engineering and security in your inbox.

Rota Nacional

Bring privacy into your workflow.

30 days, no card, with a starting quota. After that, Pix credit from R$ 5,00.

Try free