// RUN โ€” FITS

Can the NVIDIA GeForce RTX 4080 run RNA-FrameFlow?

Yes โ€” here's how fast and up to what context, with real measured numbers.

RNA-FrameFlow on the NVIDIA GeForce RTX 4080

QuantFits?Max contextSpeedBeyond context
FP16 yes 201k 15990.5-19057.1 t/s ~869-1035.7 (slow)
Q8 yes 201k 31981-38114.3 t/s ~1738.1-2071.4 (slow)
Q4 yes 202k 57108.8-68061.2 t/s ~3103.7-3699 (slow)

Estimate (memory-bound). Beyond "max context" the KV cache spills to system RAM offload (depends on your system RAM) and speed drops to the "beyond" figure.

FAQ

Can the NVIDIA GeForce RTX 4080 run RNA-FrameFlow?โ–ถ

Yes. At Q4 it generates ~57108.8-68061.2 tok/s and fits up to 202k context; beyond that it spills to system RAM offload (depends on your system RAM) and slows to ~3103.7-3699 tok/s.

How much context fits?โ–ถ

At Q4, up to about 202k tokens stay in fast memory; longer context spills to system RAM offload (depends on your system RAM) and throughput collapses.