[GPU] GPU Performance 및 Titan V, RTX 2080 Ti Benchmark

 

TPU

V100

TiTan V

Titan RTX

RTX 2080 Ti

RTX 2080

Titan Xp

Titan X

GTX 1080 Ti

RTX 2070

RTX 2060

GTX 1080

GTX 1070 Ti

GTX 1070

GTX 1660 Ti

GTX 1060

 

 

 

 

 

 

 

 

Titan V Benchmark 

 

FP32 TensorFlow Training Performance

 

 

FP16 TensorFlow Training Performance

 

 

 

 

 

 

 

 

RTX 2080 Ti Benchmark

 

 

RTX 2080 Ti - FP32 TensorFlow Performance (1 GPU)

For FP32 training of neural networks, the RTX 2080 Ti is...

• 37% faster than RTX 2080
• 35% faster than GTX 1080 Ti
• 22% faster than Titan XP
• 96% as fast as Titan V
• 87% as fast as Titan RTX
• 73% as fast as Tesla V100 (32 GB)

as measured by the # images processed per second during training.

 

 

 

RTX 2080 Ti - FP16 TensorFlow Performance (1 GPU)

For FP16 training of neural networks, the RTX 2080 Ti is..

• 72% faster than GTX 1080 Ti
• 59% faster than Titan XP
• 32% faster than RTX 2080
• 81% as fast as Titan V
• 71% as fast as Titan RTX
• 55% as fast as Tesla V100 (32 GB)

as measured by the # images processed per second during training.

 

 

 

 

FP32 Multi-GPU Scaling Performance (1, 2, 4, 8 GPUs)

For each GPU type (RTX 2080 Ti, RTX 2080, etc.) we measured performance while training with 1, 2, 4, and 8 GPUs on each neural networks and then averaged the results. The chart below provides guidance as to how each GPU scales during multi-GPU training of neural networks in FP32. The RTX 2080 Ti scales as follows:

• 2x RTX 2080 Ti GPUs will train ~1.8x faster than 1x RTX 2080 Ti
• 4x RTX 2080 Ti GPUs will train ~3.3x faster than 1x RTX 2080 Ti
• 8x RTX 2080 Ti GPUs will train ~5.1x faster than 1x RTX 2080 Ti

 

 

 

RTX 2080 Ti - FP16 vs. FP32

Using FP16 can reduce training times and enable larger batch sizes/models without significantly impacting the accuracy of the trained model. Compared with FP32, FP16 training on the RTX 2080 Ti is...

• 59% faster on ResNet-50
• 52% faster on ResNet-152
• 47% faster on Inception v3
• 34% faster on Inception v4
• 50% faster on VGG-16
• 38% faster on AlexNet
• 31% faster on SSD300

as measured by the # of images processed per second during training. This gives an average speed-up of +44.6%.

Caveat emptor: If you're new to machine learning or simply testing code, we recommend using FP32. Lowering precision to FP16 may interfere with convergence.

GPU Prices

• RTX 2080 Ti: $1,199.00
• RTX 2080: $799.00
• Titan RTX: $2,499.00
• Titan V: $2,999.00
• Tesla V100 (32 GB): ~$8,200.00
• GTX 1080 Ti: $699.00
• Titan Xp: $1,200.00

 

 

 

 

참고자료  1 : https://timdettmers.com/2019/04/03/which-gpu-for-deep-learning/

Which GPU(s) to Get for Deep Learning

참고자료 2 : https://www.quora.com/Is-the-Titan-X-better-than-the-1080-ti-for-deep-learning-and-if-so-why

Is the Titan X better than the 1080 ti for deep learning, and if so, why?

참고자료 3 : https://lambdalabs.com/blog/2080-ti-deep-learning-benchmarks/

RTX 2080 Ti Deep Learning Benchmarks with TensorFlow - 2019