System · desktop

Dual RTX 3090 (128 GB DDR5)

Runs models up to ~73B parameters at 4-bit entirely on the GPU. 48 GB across two used GPUs; the classic 70B-at-home build.

GPU memory: 48 GB
System RAM: 128 GB
Runs: 18 of 19model variants, 8K
Estimated cost: ~$3,500

Full compatibility

What it runs

Largest models that fit without spilling into system memory, at an 8K contextHow much text the model can consider at once, counted in tokens — roughly ¾ of a word each..

Everything →

Variant
Llama 3.1 70B Instruct
70.6B · Q4_K_M via llama.cpp
Tight fit
~13 tok/s est.
Variant
Llama 3.3 70B Instruct
70.6B · Q4_K_M via llama.cpp
Tight fit
17.6 tok/s measured
Variant
Mixtral 8x7B Instruct v0.1
46.7B · Q4_K_M via llama.cpp
Runs well
~69 tok/s est.
Variant
Qwen2.5 32B Instruct
32.8B · Q5_K_M via llama.cpp
Runs well
~24 tok/s est.
Variant
Qwen2.5-Coder 32B Instruct
32.8B · Q8_0 via llama.cpp
Runs well
~16 tok/s est.
Variant
DeepSeek-R1-Distill-Qwen-32B
32.8B · Q4_K_M via llama.cpp
Runs well
~28 tok/s est.
Variant
Qwen3 30B-A3B
30.5B · Q8_0 via llama.cpp
Runs well
~143 tok/s est.
Variant
Gemma 3 27B IT
27.4B · Q4_K_M via llama.cpp
Runs well
~33 tok/s est.
Variant
Gemma 2 27B IT
27.2B · Q4_K_M via llama.cpp
Runs well
~34 tok/s est.
Variant
Mistral Small 24B Instruct 2501
23.6B · Q6_K via llama.cpp
Runs well
~29 tok/s est.

Components

2× Hardware NVIDIA GeForce RTX 3090 gpu · 24 GB
1× Hardware AMD Ryzen 9 7950X cpu
System RAM bandwidth 83 GB/s

Measured performance

Model · quantization	Runtime	Context	Measurements	Source
Llama 3.3 70B Instruct Q4_K_M (LM Studio Community)	llama.cppcuda · b4600	4K	390 tok/s · Prompt processing (512) 17.6 tok/s · Generation (128)	Mutinai illustrative fixtures

Community results

Contributions are not open yet, so there is nothing here from members.

Reviews

Contributions are not open yet, so there is nothing here from members.