Tesla P4 8GB vs RTX 5090 32GB for Local LLMs

The RTX 5090 32GB has about 833% more memory bandwidth (1,792 vs 192 GB/s). Batch-1 decode tracks bandwidth closely, so a model that fits both cards should decode roughly 833% faster on the RTX 5090 32GB. The RTX 5090 32GB fits more of the catalog at Q4/8K — 39 vs 16 of 69 models (32 vs 8 GB VRAM). Largest model that fits the RTX 5090 32GB but not the Tesla P4 8GB: Qwen3.6-35B-A3B (36.0B). MSRP is $80 for the Tesla P4 8GB vs $1,999 for the RTX 5090 32GB. 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 RTX 5090 32GB posts the higher tok/s in 39 rows vs 0. Methodology.

Spec comparison

	Tesla P4 8GB	RTX 5090 32GB
Vendor	NVIDIA	NVIDIA
VRAM	8 GB	32 GB
Memory type	GDDR5	GDDR7
Memory bandwidth	192 GB/s	1,792 GB/s
TDP	75 W	575 W
MSRP (list)	$80	$1,999
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)	Tesla P4 8GB tok/s	Tesla P4 8GB fits	RTX 5090 32GB tok/s	RTX 5090 32GB fits	Winner
MiniCPM5-1B	1.08	113.1 est.	Yes	1055.9 est.	Yes	RTX 5090 32GB
Qwen3.5-2B	2.0	61.1 est.	Yes	570.2 est.	Yes	RTX 5090 32GB
DeepSeek-R1-0528-Qwen3-8B-layer-mix-bpw-3.8-mlx	2.3	53.1 est.	Yes	495.8 est.	Yes	RTX 5090 32GB
Qwen3.6-35B-A3B	36.0	17.0 est.	No	380.1 est.	Yes	RTX 5090 32GB
Qwen3.5-35B-A3B	34.7	18.2 est.	No	380.1 est.	Yes	RTX 5090 32GB
GLM-4.7-Flash	31.2	21.0 est.	No	380.1 est.	Yes	RTX 5090 32GB
Qwen3-30B-A3B-Instruct-2507	30.5	21.9 est.	No	380.1 est.	Yes	RTX 5090 32GB
Nemotron-3-Nano-Omni-30B-A3B-Reasoning-BF16	30.0	24.1 est.	No	380.1 est.	Yes	RTX 5090 32GB
Ornith-1.0-35B	35.0	17.0 est.	No	380.1 est.	Yes	RTX 5090 32GB
qwen35b-a3b-fable-sft-abliterated	34.7	17.3 est.	No	380.1 est.	Yes	RTX 5090 32GB
Llama-3.2-3B-Instruct	3.2	38.2 est.	Yes	356.4 est.	Yes	RTX 5090 32GB
gemma-4-26B-A4B-it	25.2	23.9 est.	No	300.1 est.	Yes	RTX 5090 32GB
Qwen3-4B-Instruct-2507	4.0	30.5 est.	Yes	285.1 est.	Yes	RTX 5090 32GB
NVIDIA-Nemotron-3-Nano-4B-BF16	4.0	30.5 est.	Yes	285.1 est.	Yes	RTX 5090 32GB
Qwen3-4B-Thinking-2507	4.0	30.5 est.	Yes	285.1 est.	Yes	RTX 5090 32GB

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

	Tesla P4 8GB	RTX 5090 32GB
Catalog models that fit	16 of 69	39 of 69
Largest exclusive fit	—	Qwen3.6-35B-A3B (36.0B)

Tesla P4 8GB details → · RTX 5090 32GB details →