~/gpu/arc-pro-b60 vs rtx-4070

Arc Pro B60 manufacturerArc Pro B60vsRTX 4070 manufacturerRTX 4070

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 arc-pro-b60 rtx-4070
Stat
arc-pro-b60
rtx-4070
Δ
VRAM
24 GB
12 GB
-50%
Memory bandwidth
456 GB/s
504 GB/s
+11%
FP16 compute
56 TFLOPS
230 TFLOPS
+311%
Weights budget at 8K ctx
18 GB
8.3 GB
-54%

Model fit difference.

$ models that change with the card
Fits on both
12of 30
Only on arc-pro-b60
9
Only on rtx-4070
0

// showing 12 of 30 models; differing fits first

Model
arc-pro-b60
rtx-4070
Qwen 2.5 32B32.5B
fitsAWQ 4-BIT
overQ4_K_M
Qwen 2.5 Coder 32B32.5B
fitsAWQ 4-BIT
overQ4_K_M
Qwen3 30B A3B30.5B
fitsQ4_K_M
overQ4_K_M
Qwen 3.6 27B27B
fitsQ4_K_M
overQ4_K_M
Qwen 3.6 35B A3B35B
fitsQ3_K_M
overQ4_K_M
gpt-oss 20B20.9B
fitsQ6_K
overQ4_K_M
Mistral Small 324B
fitsQ5_K_M
overQ4_K_M
Gemma 4 26B A4B26B
fitsQ5_K_M
overQ4_K_M
Yi 34B34B
fitsQ3_K_M
overQ4_K_M
Llama 3.2 1B1.23B
fitsFP16/BF16
fitsFP16/BF16
Llama 3.2 3B3.21B
fitsFP16/BF16
fitsFP16/BF16
Llama 3.1 8B8B
fitsFP16/BF16
fitsFP8/INT8

Which one wins for…

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

Arc Pro B60 has 12 GB more.

Faster decode (bandwidth)

RTX 4070 by +11%.

Faster prefill (compute)

RTX 4070 by +311% TFLOPS.

Catalog models that fit

Arc Pro B60: 21 fit · RTX 4070: 12.

Drill into either card.

$ ./vrambudget --gpu
$ openArc Pro B6024GBbattlemage · pro$ openRTX 407012GBada

Discussion.

$ gh discussion list

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