NVIDIA STX is a rack-scale reference architecture for a storage node containing NVMe drives and BlueField-4 DPUs. It is designed for the Vera Rubin platform, and one of its main use cases is CMX.
It formally debuted at GTC26.
STX servers
I think the main component within an STX rack is a server which I’ll just call an STX server:
- a pair of self-hosted, low-bin Bluefield-4 DPUs (64 cores, 128G LPDDR5X)
- 24 NVMe drives
And that’s about it. They are meant to be the physical servers underpinning ICMS.
Here are some photos of Quanta and Supermicro’s implementations from GTC26:
The intended use case is to attach these SSD boxes to the non-blocking parts of a SuperPOD’s front-end network and configure them to act as dumb JBOFs. On the individual GPU servers, there are complementary Bluefield DPUs which actually implement the storage logic that turn a bunch of network-attached NVMe drives into some kind of distributed KV cache using Dynamo’s KVBM.