DiggTop ⌕

RTX 3090 x2

Nvidia · 24 GB VRAM

10recorded runs
240.8best tok/s
7sourced

How it compares — 24 GB per card · multi-GPU

RTX 3090 x2240.8RTX 3090 x4145.8RTX 3090 x891.2

Best verified tok/s per GPU in this VRAM tier — compare any two like-for-like.

Evidence: 7 sourced runs · 3 awaiting review · 2 independent sources · first recorded May 10, 2026 · last recorded Oct 13, 2026. What these labels mean.

What fits on 24 GB — measured

ModelQuantPeak VRAMHeadroom
Qwen3.8-27BGPTQ-4bit45.3 GB-21.3 GB
Qwen3.8-Flash-NextUnsloth-Dynamic-IQ3_XXS42.9 GB-18.9 GB
Qwen3.8-Flash-Next-REAP-256-duoUnsloth-Dynamic-Q3_K_XL33.9 GB-9.9 GB
NVIDIA-Nemotron-3.5-Lightning-30B-A3BNVFP419.5 GB4.5 GB

Peak process memory recorded by the submitter on this card — evidence of what fit, not a guarantee. Plan with the VRAM calculator.

All benchmarks Compare RTX 3090 x2 →

GPUModelQuantBackendCtx tok/sprompt tok/sVRAM GBSource
RTX 3090 x2 · 24 GBQwen3.6-27BAutoRound INT4vllm 0.19.1200000240.8 ✓——src
RTX 3090 x2 · 24 GBQwen3.8-27BAutoRound INT4 + fp8 KVsglang 0.5.19262000219.8 ✓——src
RTX 3090 x2 · 24 GBNVIDIA-Nemotron-3.5-Lightning-30B-A3BNVFP4vllm 0.27.1131072197.0 ✓6719.019.5src
RTX 3090 x2 · 24 GBQwen3.8-27BW4A16-AutoRoundvllm 0.27.1+syv-ai.00210159262144182.4 ✓1338.2—src
RTX 3090 x2 · 24 GBMiniCPM-V 4.7-35B-A3Bint4vllm 0.31.1rc1.dev296 (nightly fe4df3d)262144172.2 ✓——src
RTX 3090 x2 · 24 GBQwen3.8-Flash-Next-REAP-256-duoUnsloth-Dynamic-Q3_K_XLllama.cpp 0.3.0-dev / commit 035e227 (llamacpp…3276853.5 ✓948.633.9src
RTX 3090 x2 · 24 GBQwen3.8-Flash-NextUnsloth-Dynamic-IQ3_XXSllama.cpp 0.3.0-dev / commit 035e227 (llamacpp…21708840.1 ✓340.142.9src
RTX 3090 x2 · 24 GBLing-3.0-tinyQ8_0llama.cpp —2048232.8 …9243.2—src
RTX 3090 x2 · 24 GBQwen3.8-27BGPTQ-4bitvllm —819296.6 …1311.145.3src
RTX 3090 x2 · 24 GBMuse-Glimmer-30BUnsloth-Dynamic-Q6_K_XLllama.cpp —204831.7 …1604.0—src

Related reading: GPU buying guide · used GPUs · Value index · Street prices

↑