readygpu

NVIDIA GeForce RTX 3090 vs NVIDIA RTX 6000 Ada Generation

For running models locally, priced in Germany. At 128k context the first runs 31 models fully in memory, the second 60.

In short

The NVIDIA RTX 6000 Ada Generation has more memory for models: 48 GB against 24 GB. At 128k context the NVIDIA RTX 6000 Ada Generation runs 60 of the models we track fully in memory and the NVIDIA GeForce RTX 3090 31; only the NVIDIA RTX 6000 Ada Generation holds Qwen 3.8 27B. On the models both run fully in memory, the NVIDIA GeForce RTX 3090 and the NVIDIA RTX 6000 Ada Generation are about as fast as each other (estimated).

Updated . Speeds are estimates; how they are worked out.

NVIDIA GeForce RTX 3090

2.290 €

Amazon

Add to build
NVIDIA RTX 6000 Ada Generation

Not priced here

Add to build

Specifications

SpecNVIDIA GeForce RTX 3090NVIDIA RTX 6000 Ada Generation
GPUGeForce RTX 3090RTX 6000 Ada Generation
Memory24 GB GDDR6X48 GB GDDR6
Memory bus384-bit384-bit
Memory bandwidth936 GB/s960 GB/s
FP16 compute71 TFLOPS91.1 TFLOPS
SoftwareCUDACUDA
Board power350 W300 W
PCIe generationGen 4Gen 4
PCIe lanesx16x16

What each runs

Estimated decode speed with the most precise quant that fits each one est.

ModelNVIDIA GeForce RTX 3090NVIDIA RTX 6000 Ada Generation
Qwen 3.8 27Boffload16tok/sQ8_0
Qwen 3.5 122B-A10Boffload53tok/sUD-Q2_K_XL
Qwen 3.6 27Boffload16tok/sQ8_0
Gemma 4 31Boffload14tok/sQ8_0
Mistral Small 4 119Boffload66tok/sUD-Q2_K_XL
Qwen 3.6 35B-A3Boffload61tok/sQ8_0
Qwen 3.5 27Boffload16tok/sQ8_0
Gemma 4 26B-A4B59tok/sQ5_K_M51tok/sQ8_0
Qwen 3.5 35B-A3Boffload61tok/sQ8_0
GLM 4.7 Flashoffload29tok/sQ8_0
Devstral Small 2 24Boffload13tok/sQ8_0
Nemotron 3.5 Lightning 30B-A3Boffload86tok/sUD-Q6_K_XL
gpt-oss 20B62tok/sMXFP463tok/sMXFP4
Kimi Linear 48B-A3Boffload100tok/sQ6_K
Qwen3 Coder 30B-A3Boffload22tok/sQ8_0
Gemma 4 12B38tok/sQ8_039tok/sQ8_0
Qwen 3.5 9B42tok/sQ8_043tok/sQ8_0
Gemma 3 27Boffload15tok/sQ8_0
Mistral Small 3.2 24Boffload13tok/sQ8_0
Ministral 3 14Boffload17tok/sQ8_0
Qwen3 14Boffload16tok/sQ8_0
DeepSeek R1 0528 Qwen3 8Boffload21tok/sQ8_0
Qwen3 8Boffload21tok/sQ8_0
Phi-4 14Boffload14tok/sQ8_0
Qwen 3.5 4B66tok/sQ8_067tok/sQ8_0
Ministral 3 8B25tok/sQ4_K_M22tok/sQ8_0
Granite 4.2 8Boffload19tok/sQ8_0
Gemma 4 E4B48tok/sQ8_050tok/sQ8_0
LFM2.5 8B-A1B46tok/sQ8_048tok/sQ8_0
Llama 3.1 8B25tok/sQ5_K_M25tok/sQ6_K
Qwen2.5 0.5B271tok/sQ8_0278tok/sQ8_0
Qwen3.5 0.8B238tok/sQ8_0244tok/sQ8_0
Qwen3.5 2B189tok/sQ5_K_M194tok/sQ5_K_M
DeepSeek R1 Distill Qwen 1.5B103tok/sQ8_0105tok/sQ8_0
Mellum2.1 12B A2.5B Thinking76tok/sQ8_078tok/sQ8_0
Spark X2.5 4B73tok/smlx-4bit75tok/smlx-4bit
MiniCPM5 2B70tok/sQ8_072tok/sQ8_0
FrogNano 4B 260965tok/sQ8_066tok/sQ8_0
LFM2.5 2.6B53tok/sQ8_055tok/sQ8_0
LFM2.5 VL 3B53tok/sQ8_055tok/sQ8_0
Kimi VL A3B Thinking 250649tok/sQ8_050tok/sQ8_0
Kimi VL A3B48tok/smlx-8bit49tok/smlx-8bit
GLM 4 9B 041443tok/smlx-5bit44tok/smlx-5bit
granite vision 4.1 4b40tok/sQ8_042tok/sQ8_0
granite 4.2 3b40tok/sQ8_041tok/sQ8_0
GLM 4.6V Flash38tok/sQ8_039tok/sQ8_0
DeepSeek R1 Distill Qwen 7B37tok/sQ8_038tok/sQ8_0
GLM 4.1V 9B Thinking34tok/smlx-8bit35tok/smlx-8bit
Ministral 3 3B Instruct 251233tok/sQ8_034tok/sQ8_0
Muse Glimmer 30B27tok/sQ5_K_M17tok/sQ8_0
DeepSeek R1 Distill Llama 8B25tok/sQ5_K_M23tok/sQ8_0
Ministral 3 8B Reasoning 251225tok/sQ4_K_M22tok/sQ8_0
LFM2.5 24B A2Boffload28tok/sQ8_0
Qwen3.8 27B OBLITERATEDoffload20tok/smlx-4bit
Ministral 3 14B Reasoning 2512offload17tok/sQ8_0
DeepSeek R1 Distill Qwen 14Boffload14tok/sQ8_0
Devstral Small 2507offload13tok/sQ8_0
Magistral Small 2506offload13tok/sQ8_0
Magistral Small 2507offload13tok/sQ8_0
Magistral Small 2509offload13tok/sQ8_0
Nemotron 3 Super 120B-A12Bwon't runoffload
Qwen3 Coder Nextwon't runoffload
gpt-oss 120Bwon't runoffload
GLM 4.5 Airwon't runoffload
Qwen3 Next 80B-A3Bwon't runoffload
Qwen3 32Bwon't runoffload
DeepSeek R1 Distill Qwen 32Bwon't runoffload
granite 4.2 30bwon't runoffload
DeepSeek V4 Flashwon't runwon't run
GLM 5.3 Flashwon't runwon't run
Qwen 3.8 Flash Nextwon't runwon't run
MiniMax M2.7won't runwon't run
Llama 3.3 70Bwon't runwon't run
DeepSeek R1won't runwon't run
DeepSeek R1 0528won't runwon't run
DeepSeek R1 Distill Llama 70Bwon't runwon't run
DeepSeek R1 Zerowon't runwon't run
DeepSeek V3won't runwon't run
DeepSeek V3 0324won't runwon't run
DeepSeek V3.1won't runwon't run
DeepSeek V3.1 Terminuswon't runwon't run
DeepSeek V3.2won't runwon't run
DeepSeek V4 Flash 0731won't runwon't run
DeepSeek V4 Flash Vision Expwon't runwon't run
Devstral 2 123B Instruct 2512won't runwon't run
GLM 4.5won't runwon't run
GLM 4.5Vwon't runwon't run
GLM 4.6won't runwon't run
GLM 4.6Vwon't runwon't run
GLM 4.7won't runwon't run
GLM 5won't runwon't run
GLM 5.1won't runwon't run
GLM 5.2won't runwon't run
GLM 5.3won't runwon't run
Kimi Dev 72Bwon't runwon't run
Kimi K2won't runwon't run
Kimi K2 Instruct 0905won't runwon't run
Kimi K2 Thinkingwon't runwon't run
Kimi K2.5won't runwon't run
MiniMax M2won't runwon't run
MiniMax M2.1won't runwon't run
MiniMax M2.5won't runwon't run
MiniMax M3won't runwon't run
Mistral Medium 3.5 128Bwon't runwon't run

Questions

Which is better for local AI, the NVIDIA GeForce RTX 3090 or the NVIDIA RTX 6000 Ada Generation?
Choose the NVIDIA RTX 6000 Ada Generation to run bigger models.
Which can run bigger models, the NVIDIA GeForce RTX 3090 or the NVIDIA RTX 6000 Ada Generation?
At 128k context the NVIDIA RTX 6000 Ada Generation runs 60 of the models we track fully in memory and the NVIDIA GeForce RTX 3090 31; only the NVIDIA RTX 6000 Ada Generation holds Qwen 3.8 27B.
Which is faster for local AI, the NVIDIA GeForce RTX 3090 or the NVIDIA RTX 6000 Ada Generation?
On the models both run fully in memory, the NVIDIA GeForce RTX 3090 and the NVIDIA RTX 6000 Ada Generation are about as fast as each other (estimated).

More comparisons

More hardware for local AI