Arc B580 12GB vs Tesla P4 8GB for Local LLMs

The Arc B580 12GB has about 138% more memory bandwidth (456 vs 192 GB/s). Batch-1 decode tracks bandwidth closely, so a model that fits both cards should decode roughly 138% faster on the Arc B580 12GB. The Arc B580 12GB fits more of the catalog at Q4/8K — 18 vs 16 of 69 models (12 vs 8 GB VRAM). Largest model that fits the Arc B580 12GB but not the Tesla P4 8GB: OLMo-2-1124-13B-Instruct (13.7B). MSRP is $249 for the Arc B580 12GB vs $80 for the Tesla P4 8GB. MSRP is a launch list price, not live retail — Amazon prices move; click through for the current price. In the Q4/8K head-to-head below (39 shared models), the Arc B580 12GB posts the higher tok/s in 39 rows vs 0. Methodology.

Spec comparison

	Arc B580 12GB	Tesla P4 8GB
Vendor	Intel	NVIDIA
VRAM	12 GB	8 GB
Memory type	GDDR6	GDDR5
Memory bandwidth	456 GB/s	192 GB/s
TDP	190 W	75 W
MSRP (list)	$249	$80
Buy	Search Amazon (affiliate)	Search Amazon (affiliate)

MSRP is the launch list price, not live retail. Buy links are Amazon affiliate links — the price after you click is the live price.

Q4 / 8K head-to-head — top 15 of 39 shared models

Model	Params (B)	Arc B580 12GB tok/s	Arc B580 12GB fits	Tesla P4 8GB tok/s	Tesla P4 8GB fits	Winner
MiniCPM5-1B	1.08	268.7 est.	Yes	113.1 est.	Yes	Arc B580 12GB
gpt-oss-20b	21.5	179.9 est.	No	33.6 est.	No	Arc B580 12GB
Qwen3.5-2B	2.0	145.1 est.	Yes	61.1 est.	Yes	Arc B580 12GB
Nemotron-3-Nano-Omni-30B-A3B-Reasoning-BF16	30.0	129.0 est.	No	24.1 est.	No	Arc B580 12GB
gemma-4-26B-A4B-it	25.2	128.2 est.	No	23.9 est.	No	Arc B580 12GB
DeepSeek-R1-0528-Qwen3-8B-layer-mix-bpw-3.8-mlx	2.3	126.2 est.	Yes	53.1 est.	Yes	Arc B580 12GB
Qwen3-30B-A3B-Instruct-2507	30.5	117.3 est.	No	21.9 est.	No	Arc B580 12GB
GLM-4.7-Flash	31.2	112.6 est.	No	21.0 est.	No	Arc B580 12GB
Qwen3.5-35B-A3B	34.7	97.5 est.	No	18.2 est.	No	Arc B580 12GB
qwen35b-a3b-fable-sft-abliterated	34.7	92.6 est.	No	17.3 est.	No	Arc B580 12GB
Ornith-1.0-35B	35.0	91.2 est.	No	17.0 est.	No	Arc B580 12GB
Qwen3.6-35B-A3B	36.0	90.9 est.	No	17.0 est.	No	Arc B580 12GB
Qwen3-4B-Instruct-2507	4.0	72.5 est.	Yes	30.5 est.	Yes	Arc B580 12GB
NVIDIA-Nemotron-3-Nano-4B-BF16	4.0	72.5 est.	Yes	30.5 est.	Yes	Arc B580 12GB
Qwen3-4B-Thinking-2507	4.0	72.5 est.	Yes	30.5 est.	Yes	Arc B580 12GB

est. = bandwidth-based estimate; measured = curated real benchmark that overrides the estimate. Fits = weights + KV(8K) + 1 GB overhead ≤ 95% of VRAM.

Fit coverage at Q4 / 8K

	Arc B580 12GB	Tesla P4 8GB
Catalog models that fit	18 of 69	16 of 69
Largest exclusive fit	OLMo-2-1124-13B-Instruct (13.7B)	—

Arc B580 12GB details → · Tesla P4 8GB details →