GPU servers

HOSTKEY deploys a well-established environment for machine learning applications such as neural networks with high-performance GPUs and dedicated servers with NVIDIA GTX 1080/1080Ti and RTX 2080Ti graphics cards.

No technical expertise needed! Just start your TensorFlow experience in a straightforward and user-friendly environment making it easy to build, train and deploy machine learning models at scale.

Save time. TensorFlow runs up to 50% faster on our high-performance GPUs and scales easily. Now your machines learn in hours, not days.

Get started today with GPU-Ready TensorFlow.

Cloud GPU server is an excellent solution that allows you to benefit from graphical calculations without capital and time expenditures. By renting a server from HOSTKEY, you get:

• Access to most advanced NVIDIA technologies without significant costs. Thousands of CUDA cores designed to accelerate information processing will be used for parallel computing.
• Equipment that will never become obsolete. The default advantage of dedicated servers is also applicable to GPU servers — you do not need to maintain the fleet of outdated equipment. You can always get the latest solution and you do not have to concern yourself with getting rid of the old one.
• Possibility to choose the most suitable country depending on the location of your users or other characteristics of the project: Russia, USA or the Netherlands.
• Proven configurations — the experience of our customers allowed us to form configurations with the most successful combinations of CPU and GPU.
• Servers that can be used as stand-alone servers or as part of a cluster.

You should consider getting a GPU server if your project is associated with the following:

• Mining
• Neural network and deep learning
• Virtualization
• 3D virtualization
• Rendering
• Facial recognition
• Telemedicine
• Cloud gaming

GPU NVIDIA GeForce GTX 1080 and GTX 1080Ti are top of the range graphics cards that were initially designed for gaming but were also put to good use in other areas.

GeForce GTX 10 series are built using Pascal, world's most advanced GPU architecture.

These graphics cards provide an increase in performance up to 3 times compared with the previous generation. They accelerate 3D graphics processing or act as coprocessors, using CUDA cores for single-precision calculations.

The use of GPU allows to obtain and process a huge amount of source data in an extremely short space of time for efficient self-learning of neural networks or for creating a ready-made 3D model in a few minutes using photogrammetry. GTX 10 series also perform well in solving tasks related to rendering, mining, software development, streaming, transcoding, etc.

Main specifications	1080	1080Ti
GPU architecture	Pascal	Pascal
Frame buffer	8 GB GDDR5X	11 GB GDDR5X
Memory clock	10 Gbps	11 Gbps
Boost clock	1733 MHz	1582 MHz

NVIDIA TESLA M60 and NVIDIA GRID K2

Both cards are using NVIDIA GRID technology, which was developed specifically for desktop virtualization.
NVIDIA GRID K2 is a basic solution for creating virtual desktops. The cards have an optimized multi-GPU design that helps to maximize user density. GRID K2 includes two higher-end Kepler GPUs and can host 16 concurrent users. Virtual desktops are suitable for working with regular applications as well as more demanding projects — video and 3D graphics. Alternative scenarios for using servers with GRID K2 cards involve graphics virtualization and streaming.

NVIDIA TESLA M60 provides a superior experience for any user, from office workers to mobile professionals to designers and engineers.
Almost two-fold superiority as compared to GRID K series is not just a marketing indicator — extended capabilities allow to double either the productivity of each user's workstations or the number of users. Enhanced components and software have brought user density to a new level — 32 users per card or 128 users per server.
These cards can be used for cloud gaming, rendering, transcoding, and much more.

Main specifications	GRID K2	TESLA M60
NVIDIA CUDA cores	3072 (2x1536)	4096 (2x2048)
Graphics memory size	8 Gb GDDR5 (2х4 GB)	16 Gb GDDR5 (2х8 GB)
Concurrent users	16	32

Things to consider:

1. NVIDIA GRID is available in three editions:
   • GRID Virtual Applications (vApps) — for deployment of Citrix XenApp or other RDSH solutions, such as VMware Horizon Hosted Apps.
   • GRID Virtual PC (vPC ) — for leveraging Windows applications, browsers and high definition video.
   • GRID Virtual Workstation (vWS) — for accessing professional graphics applications with full features and performance.
2. Annual subscription or perpetual license.
3. Level of technical support — annual subscription to SUMs (Support, Upgrade and Maintenance) program.