The new “6 x Tesla® T4” configuration is now available for rent at LeaderGPU

Tesla® T4 shows excellent results in machine learning, analytics and rendering. This card has been specifically designed for inferencing. The Tesla® T4 supports a full range of precisions for inference: FP32 (8,1 TFLOPS), FP16 (65 TFLOPS), INT8 (130 TFLOPS) and INT4 (260 TFLOPS).

The range of server offers from LeaderGPU® has been updated recently with another powerful option that will be appreciated by rendering specialists and experts in the field of artificial intelligence and neural networks.

New servers available for rent include:

• GPU: 6xTesla T4,
• CPU: 2xIntel® Xeon® Gold 5217 Processor, 3.00 GHz 8 cores in each CPU, 16 threads in each CPU.
• RAM: 192 GB DDR4 ECC REG 2666.
• SSD: 1x1TB NVMe P4510.

Among the advantages of the Tesla® T4 graphics card is the impressive single-precision HPC performance.

The Tesla® T4 is based on Turing™ architecture and includes 2560 CUDA® cores, as well as 320 Tensor cores. As a result, all requests are processed approximately 40 times faster than on traditional CPU.

If we consider INT8 precision, Tesla® T4 cards work up to 2.8x better than Tesla® P4 cards, as benchmarks show (with AlexNet, GoogLeNet, ResNet and VGG). Inference performance for multiple T4 GPUs will scale linearly.

Rent GPU servers with 6xTesla T4 on LeaderGPU® right now get the power you need!