← AI Hashrate

L40S 48GB vs RTX 3080 10GB for Local LLMs

The L40S 48GB has about 14% more memory bandwidth (864 vs 760 GB/s). Batch-1 decode tracks bandwidth closely, so a model that fits both cards should decode roughly 14% faster on the L40S 48GB. The L40S 48GB fits more of the catalog at Q4/8K — 39 vs 17 of 61 models (48 vs 10 GB VRAM). Largest model that fits the L40S 48GB but not the RTX 3080 10GB: Qwen3-Coder-Next (79.7B). MSRP is $8,000 for the L40S 48GB vs $699 for the RTX 3080 10GB. MSRP is a launch list price, not live retail — Amazon prices move; click through for the current price. In the Q4/8K head-to-head below (38 shared models), the L40S 48GB posts the higher tok/s in 29 rows vs 9. Methodology.

Spec comparison

L40S 48GBRTX 3080 10GB
VendorNVIDIANVIDIA
VRAM48 GB10 GB
Memory typeGDDR6GDDR6X
Memory bandwidth864 GB/s760 GB/s
TDP350 W320 W
MSRP (list)$8,000$699
BuySearch Amazon (affiliate)Search Amazon (affiliate)

MSRP is the launch list price, not live retail. Buy links are Amazon affiliate links — the price after you click is the live price.

Q4 / 8K head-to-head — top 15 of 38 shared models

ModelParams (B)L40S 48GB tok/sL40S 48GB fitsRTX 3080 10GB tok/sRTX 3080 10GB fitsWinner
Qwen3-0.6B0.8687.3 est.Yes604.5 est.YesL40S 48GB
gemma-3-1b-it1.0549.8 est.Yes483.6 est.YesL40S 48GB
Llama-3.2-1B-Instruct1.2458.2 est.Yes403.0 est.YesL40S 48GB
Qwen3-1.7B2.0274.9 est.Yes241.8 est.YesL40S 48GB
gpt-oss-20b21.5152.7 est.Yes209.1 est.NoRTX 3080 10GB
Qwen3-30B-A3B30.5183.3 est.Yes136.3 est.NoL40S 48GB
Qwen3-Coder-30B-A3B-Instruct30.5183.3 est.Yes136.3 est.NoL40S 48GB
GLM-4.7-Flash31.2183.3 est.Yes130.9 est.NoL40S 48GB
Qwen3-30B-A3B-Instruct-250730.5183.3 est.Yes136.3 est.NoL40S 48GB
Llama-3.2-3B-Instruct3.2171.8 est.Yes151.1 est.YesL40S 48GB
Phi-4-mini-instruct3.8144.7 est.Yes127.3 est.YesL40S 48GB
Phi-4-mini-reasoning3.8144.7 est.Yes127.3 est.YesL40S 48GB
gemma-3-4b-it4.3141.0 est.Yes124.0 est.YesL40S 48GB
Qwen3-4B-Instruct-25074.0137.5 est.Yes120.9 est.YesL40S 48GB
Qwen3-4B-Thinking-25074.0137.5 est.Yes120.9 est.YesL40S 48GB

est. = bandwidth-based estimate; measured = curated real benchmark that overrides the estimate. Fits = weights + KV(8K) + 1 GB overhead ≤ 95% of VRAM.

Fit coverage at Q4 / 8K

L40S 48GBRTX 3080 10GB
Catalog models that fit39 of 6117 of 61
Largest exclusive fitQwen3-Coder-Next (79.7B)—

L40S 48GB details → · RTX 3080 10GB details →