NVIDIA · 48 GB GDDR6 · 768 GB/s bandwidth ·
Google · 27.4B · model ctx up to 128K
Yes — the RTX A6000 48GB (48 GB GDDR6) can run gemma-3-27b-it. At Q4 with the default 8K context it needs ≈ 21.51 GB VRAM (weights 15.07 GB + KV cache 5.44 GB + 1 GB runtime overhead), which fits within 48 GB. Decode at Q4/8K: ≈ 18.0 tok/s (estimated, batch 1). FP16 needs ≈ 61.24 GB — does not fit on this card. Estimates come from memory-bandwidth math; rows tagged measured override estimates. Relative ranking is more reliable than absolute tok/s. Methodology.
Fit & speed by quant and context
Quant
Context
VRAM needed
Fits?
tok/s (decode)
Q4
4K
18.79 GB
Yes
18.0est.
Q4
8K default
21.51 GB
Yes
18.0est.
Q4
32K
37.83 GB
Yes
18.0est.
FP16
4K
58.52 GB
No
9.5est.
FP16
8K default
61.24 GB
No
8.7est.
FP16
32K
77.56 GB
No
5.4est.
Fits = weights + KV(ctx) + 1 GB overhead ≤ 95% of VRAM. Measured
anchors are context-agnostic; the fit verdict is recomputed per context. A missing tok/s
means the model is far beyond this card (offload-only territory).