gpu//db
live · 9 cards indexed
$ gpudb query --workload=llm-inference --sort=ai_value --order=desc
// 9 GPUs indexed · ranked for local AI workloads · fp16 unless noted

GPUs ranked for AI inference

Sort by AI value, FP32 TFLOPS, or VRAM. Local-fit flags show which models run without CPU offload.

▸ Top AI Value
4.3/5
NVIDIA GeForce RTX 3090
▸ Local 70B Capable
0cards
VRAM ≥ 48 GB · 4-bit quant
▸ FP32 Leader
82.6TFL
NVIDIA GeForce RTX 4090
▸ GPUs Indexed
9cards
consumer · datacenter · pro
# GPU Vendor VRAM FP32 AI Value tok/s 8B Local Fit Price Tier
01 NVIDIA GeForce RTX 3090
enthusiast · 350W
 nvidia 24 GB 35.6
4.3
95
Check enthusiast
02 NVIDIA GeForce RTX 3060 12GB
mid-range · 170W
 nvidia 12 GB 12.7
4.2
45
Check mid-range
03 NVIDIA GeForce RTX 4070 Super
enthusiast · 220W
 nvidia 12 GB 36.0
4.1
110
Check enthusiast
04 NVIDIA GeForce RTX 4060 Ti 16GB
mid-range · 165W
 nvidia 16 GB 22.1
4.0
55
Check mid-range
05 NVIDIA GeForce RTX 4090
flagship · 450W
 nvidia 24 GB 82.6
4.0
140
Check flagship
06 NVIDIA GeForce RTX 4080 Super
flagship · 320W
 nvidia 16 GB 52.2
4.0
125
Check flagship
07 NVIDIA GeForce RTX 4070 Ti
enthusiast · 285W
 nvidia 12 GB 40.1
3.8
95
Check enthusiast
08 AMD Radeon RX 7900 XTX
enthusiast · 355W
 amd 24 GB 61.4
3.7
88
Check enthusiast
09 AMD Radeon RX 7800 XT
enthusiast · 263W
 amd 16 GB 37.0
3.6
75
Check enthusiast
// 9 GPUs · all metrics ≤ 90d old browse by tier ▸