// RUN — FITS

Can the NVIDIA GeForce RTX 5090 run AWD-LSTM-DRILL + dynamic evaluation† (PTB)?

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

AWD-LSTM-DRILL + dynamic evaluation† (PTB) on the NVIDIA GeForce RTX 5090

QuantFits?Max contextSpeedBeyond context
FP16 yes 426k 27253.3-32480 t/s ~608.3-725 (slow)
Q8 yes 426k 54506.7-64960 t/s ~1216.7-1450 (slow)
Q4 yes 426k 97333.3-116000 t/s ~2172.6-2589.3 (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 5090 run AWD-LSTM-DRILL + dynamic evaluation† (PTB)?

Yes. At Q4 it generates ~97333.3-116000 tok/s and fits up to 426k context; beyond that it spills to system RAM offload (depends on your system RAM) and slows to ~2172.6-2589.3 tok/s.

How much context fits?

At Q4, up to about 426k tokens stay in fast memory; longer context spills to system RAM offload (depends on your system RAM) and throughput collapses.