Parts/Accelerators/SambaNova RDU
Accelerators · RDU · 2021
SambaNova RDU
Reconfigurable Dataflow Unit, sold as a complete rack-level system rather than as an accelerator card.
Why this part
A reconfigurable dataflow architecture: rather than executing a kernel at a time, the compiler maps the whole model graph onto the fabric so intermediate results never round-trip to memory. Large addressable memory per node is the pitch for very large models. It is also the category's recurring difficulty, since every departure from the GPU programming model narrows the set of software that runs without effort.
Systems using this part
- Systems
- 1
- Rmax underneath
- 0.0 GFlop/s
- Units recorded
- -
0 of 1 edges disclose a quantity
| System | Role | Country | Units | System Rmax | Supplied by | Rank |
|---|---|---|---|---|---|---|
| Killellan AI Growth Zone | Accelerator | United Kingdom | - | - | SambaNova | - |
“Units recorded” sums only the edges where a public source states a quantity. It is a floor, not a total, and should never be read as installed base.