This site requires Javascript in order to view all its content. Please enable Javascript in order to access all the functionality of this web site. Here are the instructions how to enable JavaScript in your web browser.

Enterprise Servers for Accelerated Workloads

T4 brings GPU acceleration to the world’s leading enterprise servers.

Watch Now

NVIDIA T4 enterprise GPUs and CUDA-X acceleration libraries supercharge mainstream servers, designed for today’s modern data centers. With NGC-Ready validation, these servers excel across the full range of accelerated workloads. Optional support through NVIDIA NGC Support Services ensures peace of mind. With NVIDIA T4, IT teams can build out standard GPU-accelerated data center infrastructures with the world’s leading servers.

Drive Utilization and Productivity with a Single Accelerated Platform

Up to 33% more performance in virtual desktop infrastructures (VDI)*

Up to 35X faster on machine learning training

Up to 16X faster on deep learning training

Up to 68X faster time to insights on deep learning inference

CONFIGURATION INFORMATION

Enterprise Servers for Accelerating Machine Learning

Dramatic Acceleration of Machine Learning

While deep learning is a fairly recent arrival to the AI market, companies have been using machine learning to glean insights from data for decades. The NVIDIA T4 data center GPU can accelerate these machine learning techniques using RAPIDS, a set of open-source libraries for GPU-accelerating data preparation and machine learning. Using familiar development tools like Python, a T4 GPU can accelerate machine learning up to 35X compared to a CPU-only server, including algorithms like XGBoost, PCA, K-means, k-NN, DBScan, and tSVD.

LEARN MORE

Turing Tensor Cores Accelerate NVIDIA T4 in Enterprise Servers

Unmatched AI Acceleration with Turing Tensor Cores

NVIDIA T4 is based on the revolutionary NVIDIA Turing^™ Tensor Core technology with multi-precision computing for AI workloads. Powering breakthrough performance from FP32 to FP16 to INT8, as well as INT4, T4 delivers up to 68X higher inference performance than CPUs. And for training, a single server with two T4 GPUs replaces nine dual-socket CPU-only servers.

Developers can unleash the power of Turing Tensor Cores directly through NVIDIA CUDA-X AI software libraries and integrations with all AI frameworks. Built on top of CUDA, NVIDIA’s parallel programming model, CUDA-X libraries provide optimizations for the specific computing requirements of artificial intelligence, autonomous machines, high-performance computing, and graphics.

LEARN MORE

Exceptional VDI Experiences

NVIDIA T4-equipped mainstream servers are certified by our partners for virtual GPU software, ensuring users enjoy high-quality virtual desktop experiences from anywhere, on any device. From knowledge workers and mobile professionals to designers and engineers, T4, when combined with NVIDIA virtual GPU software—NVIDIA GRID^® Virtual PC (GRID vPC) and NVIDIA Quadro^® Virtual Data Center Workstation (Quadro vDWS)—can accelerate the VDI experience, delivering up to 33 percent better performance than CPU-only VDI environments.

LEARN MORE

NGC-Ready-Validated Servers

Optimizing software for world-class performance requires a deep understanding of many different software layers, from operating systems and drivers to libraries, frameworks, and applications. NGC containers are pre-built and highly optimized for GPU computing. They bring NVIDIA’s unmatched expertise and ecosystem support for GPU-accelerated deep learning frameworks, RAPIDS software, NVIDIA TensorRT^™, and 600+ high-performance computing applications.

NVIDIA’s NGC-Ready program enables all server makers to validate NGC containers on their systems. NGC-Ready servers give data center managers the confidence to deploy GPU-equipped servers, knowing that they can be quickly set up to excel at a wide variety of accelerated applications.

LEARN MORE

ENTERPRISE SERVER PARTNERS

VDI: GRID vPC tested on a server with 2x Intel Xeon Gold 6148 (20c, 2.4 GHz), GRID vPC (using 1B profiles) with 4x T4 GPUs supports 64 user VMs, VMware ESXi 6.7, NVIDIA vGPU Software (410.91/412.16), Windows 10 (1803), 2 vCPUs, 4GB RAM, Resolution 1920x1080 resolution, Single Monitor, VMware Horizon 7.6 User experience was measured using an NVIDIA internal benchmarking tool which measured remoted frames running office productivity applications such as Microsoft PowerPoint, Word, Excel, Chrome, PDF viewing and video playback.

Machine Learning: CPU nodes (61 GB of memory, 8 vCPUs, 64-bit platform), Apache Spark. 200 GB CSV dataset; Data preparation includes joins, variable transformations. GPU Server Config: Dual-Socket Xeon E5-2698 v4@3.6GHz, 20 T4 GPUs on 5 nodes, each with 4x T4 GPUs. All run on InfiniBand network, CPU data for XGBoost and Data Conversion steps are estimated based on measured data for 20 CPU nodes, and reducing execution time by 60% to normalize for training on smaller data set on T4.

Training: Workload: ResNet-50; CPU Server: Dual-Socket Xeon E5-2698 v4@3.6GHz, 512GB System Memory; GPU Server: Dual Xeon E5-2698 v4@3.6GHz with 4x T4 GPUs; Framework: MXNet v1.4; Mixed Precision, CUDA 10.1.105; NCCL 2.4.3, cuDNN 7.5.0.56; cuBLAS 10.1.105; DALI 0.7.0; NVIDIA Driver: 410.104; Batch size: T4: 208 | Dual-Socket CPU Server: 256

Inference: Workload: VGG-19; CPU Server: Dual-Socket Xeon Gold 6140@2.30GHz using OpenVINO Toolkit; GPU Servers: 4x T4 GPUs; TensorRT 5.1 GA; CUDA 10.1.105;
NCCL 2.4.3, cuDNN 7.5.0.56, cuBLAS 10.1.105; NVIDIA Driver 410.104; Batch-size: 128