RTX 3090 Benchmarks for Deep Learning – NVIDIA RTX 3090 vs 2080 Ti vs TITAN RTX vs RTX 6000/8000

NVIDIA RTX 3090 Benchmarks for TensorFlow

For this blog article, we conducted deep learning performance benchmarks for TensorFlow on NVIDIA GeForce RTX 3090 GPUs.

Our Deep Learning workstation was fitted with two RTX 3090 GPUs and we ran the standard “tf_cnn_benchmarks.py” benchmark script found in the official TensorFlow github. We tested on the the following networks: ResNet50, ResNet152, Inception v3, Inception v4. Furthermore, we ran the same tests using 1, 2, and 4 GPU configurations (for the 2x RTX 3090 vs 4x 2080Ti section). Determined batch size was the largest that could fit into available GPU memory.

Key Points and Observations

• The NVIDIA RTX 3090 outperformed all GPUs (Images/sec) across all models.
• A system with 2x RTX 3090 > 4x RTX 2080 Ti.
• For deep learning, the RTX 3090 is the best value GPU on the market and substantially reduces the cost of an AI workstation.

Interested in getting faster results?
Learn more about Exxact deep learning workstations starting at $3,700

RTX 3090 ResNet 50 TensorFlow Benchmark

	1x GPU	2x GPU	batch size
RTX 2080 Ti	522.52	959.78	128
RTX 6000	637.56	1248.54	512
RTX 8000	604.76	1184.52	1024
TITAN RTX	646.13	1287.01	512
RTX 3090	1139.15	2153.53	512

RTX 3090 ResNet 152 TensorFlow Benchmark

	1x GPU	2x GPU	batch size
RTX 2080 Ti	209.27	348.8	64
RTX 6000	281.94	519.76	256
RTX 8000	285.85	529.05	512
TITAN RTX	284.87	530.86	256
RTX 3090	457.45	857.14	256

RTX 3090 Inception V3 TensorFlow Benchmark

	1x GPU	2x GPU	batch size
RTX 2080 Ti	310.32	569.24	128
RTX 6000	391.08	737.77	256
RTX 8000	391.3	754.94	512
TITAN RTX	397.09	784.24	256
RTX 3090	697.98	1296.86	256

RTX 3090 Inception V4 TensorFlow Benchmark

	1x GPU	2x GPU	batch size
RTX 2080 Ti	150.59	247.16	64
RTX 6000	203.9	392.14	256
RTX 8000	203.67	384.29	512
TITAN RTX	207.98	399.16	256
RTX 3090	360	679.61	256

2x NVIDIA RTX 3090 Vs 4x RTX 2080 Ti – What config is Better?

	1x GPU	2x GPU	4x GPU	batch size
RTX 2080 Ti	522.52	959.78	1836.61	128
RTX 3090	1139.15	2153.53	N/A	512

TF CNN Benchmark Parameters

Description	Type
Number of Batches	100
Number of Epochs	0.01
Data Format	NCHW
Optimizer	Momentum
Variables	parameter_server

More About NVIDIA RTX 3090

The NVIDIA RTX 3090 has 24GB GDDR6X memory and is built with enhanced RT Cores and Tensor Cores, new streaming multiprocessors, and super fast G6X memory for an amazing performance boost.

Compared with RTX 2080 Ti’s 4352 CUDA Cores, the RTX 3090 more than doubles it with 10496 CUDA Cores. CUDA Cores are the GPU equivalent of CPU cores, and are optimized for running a large number of calculations simultaneously (parallel processing). More CUDA Cores generally mean better performance and faster graphics-intensive processing.

Have any questions about NVIDIA GPUs or AI workstations and servers?
Contact Exxact Today