← AI Hashrate 中文

DGX Spark (GH200 480GB) vs RTX 4080 Super 16GB for Local LLMs

The RTX 4080 Super 16GB has about 47% more memory bandwidth (736 vs 500 GB/s). Batch-1 decode tracks bandwidth closely, so a model that fits both cards should decode roughly 47% faster on the RTX 4080 Super 16GB. The DGX Spark (GH200 480GB) fits more of the catalog at Q4/8K — 66 vs 24 of 69 models (480 vs 16 GB VRAM). Largest model that fits the DGX Spark (GH200 480GB) but not the RTX 4080 Super 16GB: DeepSeek-V4-Pro (861.6B). MSRP is $3,000 for the DGX Spark (GH200 480GB) vs $999 for the RTX 4080 Super 16GB. 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 (44 shared models), the RTX 4080 Super 16GB posts the higher tok/s in 39 rows vs 5. Methodology.

Spec comparison

DGX Spark (GH200 480GB)RTX 4080 Super 16GB
VendorNVIDIANVIDIA
VRAM480 GB16 GB
Memory typeLPDDR5XGDDR6X
Memory bandwidth500 GB/s736 GB/s
TDP170 W320 W
MSRP (list)$3,000$999
BuySearch Amazon (affiliate)ASUS ProArt GeForce RTX 4080 Super OC (affiliate) · More listings

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 44 shared models

ModelParams (B)DGX Spark (GH200 480GB) tok/sDGX Spark (GH200 480GB) fitsRTX 4080 Super 16GB tok/sRTX 4080 Super 16GB fitsWinner
MiniCPM5-1B1.08294.6 est.Yes433.7 est.YesRTX 4080 Super 16GB
Nemotron-3-Nano-Omni-30B-A3B-Reasoning-BF1630.0106.1 est.Yes370.4 est.NoRTX 4080 Super 16GB
Qwen3-30B-A3B-Instruct-250730.5106.1 est.Yes337.0 est.NoRTX 4080 Super 16GB
GLM-4.7-Flash31.2106.1 est.Yes323.5 est.NoRTX 4080 Super 16GB
Qwen3.5-35B-A3B34.7106.1 est.Yes280.1 est.NoRTX 4080 Super 16GB
qwen35b-a3b-fable-sft-abliterated34.7106.1 est.Yes265.9 est.NoRTX 4080 Super 16GB
Ornith-1.0-35B35.0106.1 est.Yes261.9 est.NoRTX 4080 Super 16GB
Qwen3.6-35B-A3B36.0106.1 est.Yes261.1 est.NoRTX 4080 Super 16GB
Qwen3.5-2B2.0159.1 est.Yes234.2 est.YesRTX 4080 Super 16GB
DeepSeek-R1-0528-Qwen3-8B-layer-mix-bpw-3.8-mlx2.3138.3 est.Yes203.6 est.YesRTX 4080 Super 16GB
gpt-oss-20b21.588.4 est.Yes186.5 measuredYesRTX 4080 Super 16GB
Llama-3.2-3B-Instruct3.299.4 est.Yes146.4 est.YesRTX 4080 Super 16GB
gemma-4-26B-A4B-it25.283.7 est.Yes123.3 est.YesRTX 4080 Super 16GB
Qwen3-4B-Instruct-25074.079.5 est.Yes117.1 est.YesRTX 4080 Super 16GB
NVIDIA-Nemotron-3-Nano-4B-BF164.079.5 est.Yes117.1 est.YesRTX 4080 Super 16GB

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

DGX Spark (GH200 480GB)RTX 4080 Super 16GB
Catalog models that fit66 of 6924 of 69
Largest exclusive fitDeepSeek-V4-Pro (861.6B)—

DGX Spark (GH200 480GB) details → · RTX 4080 Super 16GB details →