~/gpu/rtx-3090-ti vs rtx-4090

RTX 3090 Ti manufacturerRTX 3090 TivsRTX 4090 manufacturerRTX 4090

Head-to-head for local LLM inference. The honest comparison: VRAM, bandwidth, compute, and which of the 30 catalog models actually fit on each.

The specs.

$ diff specs rtx-3090-ti rtx-4090
Stat
rtx-3090-ti
rtx-4090
Δ
VRAM
24 GB
24 GB
0%
Memory bandwidth
1,008 GB/s
1,008 GB/s
0%
FP16 compute
160 TFLOPS
330 TFLOPS
+106%
Weights budget at 8K ctx
18 GB
18 GB
0%

Model fit difference.

$ models that change with the card
Fits on both
21of 30
Only on rtx-3090-ti
0
Only on rtx-4090
0

// showing 12 of 30 models; differing fits first

Model
rtx-3090-ti
rtx-4090
Llama 3.2 1B1.23B
fitsFP16/BF16
fitsFP16/BF16
Llama 3.2 3B3.21B
fitsFP16/BF16
fitsFP16/BF16
Llama 3.1 8B8B
fitsFP16/BF16
fitsFP16/BF16
Llama 3.3 70B70.6B
overQ4_K_M
overQ4_K_M
Llama 3.1 405B405B
overQ4_K_M
overQ4_K_M
Qwen 2.5 7B7B
fitsFP16/BF16
fitsFP16/BF16
Qwen 2.5 32B32.5B
fitsAWQ 4-BIT
fitsAWQ 4-BIT
Qwen 2.5 Coder 32B32.5B
fitsAWQ 4-BIT
fitsAWQ 4-BIT
Qwen 2.5 72B72.7B
overQ4_K_M
overQ4_K_M
Qwen3 30B A3B30.5B
fitsQ4_K_M
fitsQ4_K_M
Qwen 3.5 9B9B
fitsFP16/BF16
fitsFP16/BF16
Qwen 3.6 27B27B
fitsQ4_K_M
fitsQ4_K_M

Which one wins for…

$ ./recommend --by-workload
More VRAM headroom

Tied at 24 GB. Choose on bandwidth or price.

Faster decode (bandwidth)

Tied at 1,008 GB/s.

Faster prefill (compute)

RTX 4090 by +106% TFLOPS.

Catalog models that fit

Tied: 21 of 30 fit on each.

Drill into either card.

$ ./vrambudget --gpu
$ openRTX 3090 Ti24GBampere$ openRTX 409024GBada · flagship

Discussion.

$ gh discussion list

// sign in with github to leave a comment. threads live in the repo's discussions tab.