NVIDIA TESLA T4

The T4 introduces Tensor Core technology with multi-precision computing, making it up to 40 times faster than a CPU and up to 3.5 times faster than its Pascal predecessor, the Tesla P4.

Get access to 8.1 TFLOPS of single precision performance from a single T4 GPU.

Transcode up to 38 full HD video streams simultaneously with a single Tesla T4 GPU paired with our HP BL460c blade server.

*Results may vary, based on server configuration.

• TURO TU104
• 320 TURING TENSOR CORES
• 2560 CUDA CORES
• 16GB GDDR6
• 8.1 TFLOPS SINGLE PRECISSION
• 65 FP16 TFLOPS
• 130 INT8 TOPS
• 260 INT4 TOPS
• 320 GB/s Max Bandwidth

Compatible: VMWare ESXi, Citrix Xenserver, KVM, Linux, Windows.

The Coral USB Accelerator

You can now add an Edge TPU coprocessor to any Linux-based system with the Coral USB Accelerator designed by Google. The small ASIC chip provides high-performance ML inferencing with low power cost. For example, it can execute 100fbps on MobileNet v2 models, while using very little power (500mA at 5V).

Specifications

• ARM 32 Bit Cortex 32 MHz
• Edge TPU ASIC (for Lite TensorFlow models)
• USB 3.1 5Gb/s transfer speed

Compatible with Linux machines, Debian 6.0 or higher, or any derivative (such as Ubuntu 10.0+), but also with Rasberry Pi (213 Mode B/B+).

NVIDIA GeForce RTX 2080 / RTX 2080 Ti

NVIDIA’s new Turing chip architecture delivers up to six times the performance of previous generation GPU’s, with breakthrough technologies and next generation, ultra-fast GDDR6 memory.

RTX 2080 Specifications

• 8GB GDDR6
• 2944 CUDA Cores
• 448 GB/s Max Bandwidth
• NVIDIA GPU Boost 4.0

RTX 2080 TI Specifications

• 11GB GDDR6
• 2944 CUDA Cores
• 616 GB/s Max Bandwidth
• NVIDIA GPU Boost 4.0

Compatible with Linux, CUDA/OpenCL, KVM.

NVIDIA GeForce GTX 1080/1070 TI

NVIDIA’s previous chip architecture, great for mining, graphics rendering and computing. The NVIDIA Pascal architecture delivers excellent performance at a budget friendly price.

• 8GB DDR5
• 2560 CUDA Cores
• 320 GB/s Max Bandwidth
• NVIDIA GPU Boost 3.0

Compatible with Linux, CUDA/OpenCL, KVM.

NVIDIA TITAN V

The first GPU to break the 100 teraflop barrier of deep learning performance. NVIDIA’s Volta chip, is up to 3x faster than it’s Pascal chip predecessor.

Your deep learning project design can now be a reality, with little investment. Get the maximum per machine deep learning performance, replacing up to 30 single CPU servers with just one Titan V configuration.

Use the Titan V for high performance computing, from predicting the weather to discovering or finding new energy sources. Get your results up to 1.5x faster than NVIDIA’s Pascal predecessor.

• NVIDIA Volta Chip
• 5120 CUDA cores
• 640 Tensor Cores
• 12 GB CoWoS Stacked HBM2
• 653 Gbps max bandwidth

Compatible: VMWare ESXi, Citrix Xenserver, KVM, Linux, Windows.