Prefill and decode are two different workloads

disaggregated-prefill-decode

helios-rackscale

wafer-scale-decode-offload

vendor-modelled-benchmark

split panel, one pool for both stages vs prefill on Helios / decode on the Wafer-Scale Engine, with both headline numbers and their actual baselines

Prefill and decode are two different workloads. On July 23 the hardware stopped pretending otherwise.

AMD launched Helios and, the same day, announced a Cerebras pairing that splits inference across two machines: Helios processes the prompts and large context windows, the Wafer-Scale Engine handles token generation. AMD's own term for it is disaggregated inference.

If you have sized an on-prem cluster you already knew the split. Prefill is compute bound and batches well. Decode is memory bandwidth bound and does not, which is why a single pool sized for both means you overpay on prefill and still miss your latency SLO on decode.

Two numbers came out that day. Keep them apart.

30% more tokens per dollar: Helios against NVIDIA Vera Rubin NVL72. AMD Performance Labs estimate, Kimi K2 Thinking, 32K in and 8K out.

5x tokens per second per watt: AMD plus Cerebras against a Cerebras WSE-only configuration. Different model, Kimi 2.6 1T. Vendor modelling, not measured silicon.

The trap: that 5x is not a win over NVIDIA. It is Cerebras with Helios against Cerebras without it, and the headline says per watt while the footnote says per kilowatt.

The topology is right. The benchmark is still someone else's marketing.