Parts/Accelerators/Groq LPU
Accelerators · LPU · 2023
Groq LPU
Language Processing Unit. Deterministic execution with no caches, which trades flexibility for predictable latency.
Why this part
Deterministic execution with no caches, no branch prediction and no external memory: the compiler schedules every cycle in advance, so latency is exactly predictable. That makes it excellent at inference and useless for training, and it means a model has to be split across enough chips for its weights to fit in on-die SRAM. The bet is that inference, not training, is where the sustained volume ends up.
Systems using this part
- Systems
- 3
- Rmax underneath
- 0.0 GFlop/s
- Units recorded
- 19,000
1 of 3 edges disclose a quantity
| System | Role | Country | Units | System Rmax | Supplied by | Rank |
|---|---|---|---|---|---|---|
| GroqCloud Dammam Inference Cluster 2025-02 | Accelerator | Saudi Arabia | 19,000 | - | Groq | - |
| GroqCloud Sydney Inference Cluster | Accelerator | Australia | - | - | Groq | - |
| GroqCloud Kamloops Inference Cluster | Accelerator | Canada | - | - | Groq | - |
“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.