Intel® Xeon® and AMD EPYC™ processors with base speeds up to 4.05 GHz are available, paired with fast NVMe disks.
For optimal video processing performance, select GPUs with strong compute capabilities.
For training large models, it is recommended to use video cards with a memory capacity of at least 80 GB. This will allow the video card to speed up data processing.
To run already trained models with a small amount of processed data, a video card with a relatively small amount of memory, for example, 24 GB, is sufficient.
For most tasks, a standard processor with a clock speed of 2.2–2.4 GHz is suitable.
If you need a GPU for HPC, a high-frequency processor (3.0+ GHz) may be required.
The main difference between local and network drives is data retrieval latency. Local drives offer higher performance.
When working with many small files or a single large file, using a local drive reduces the overall drive operation time by minimizing the network latency.
Network drives are superior in terms of easily and quickly adjusting capacity without the need to halt site operations.
Most tasks don’t have specific network requirements. Generally, the data transfer rate of 100 Mbps is sufficient.
However, additional networking requirements may arise in case of importing data for processing from external sources. For example, ML tasks require large amounts of data, and, therefore, require the local network’s data transfer rate of 1 Gbps or higher.
The main difference between standard dedicated servers and GPU-powered dedicated servers is the presence of GPUs or GPU accelerators.
GPUs are integrated into dedicated servers to provide enhanced computational power for various applications. They are particularly effective at handling tasks that can be broken down into many smaller, parallel operations, such as graphics rendering, machine learning, and scientific computing.
A tariff plan is established when you order a dedicated server in your user account in the control panel. This plan determines the duration of the paid period and the payment amount. By default, the billing day is set on the day the dedicated server is delivered, marking the start of the first paid period. The payment day can be adjusted in the control panel after placing the order. The server fee is automatically deducted from the balance in the panel according to the tariff plan.
Dedicated servers with GPU are suitable for variety of resource-intensive tasks, including:
— Virtual Desktop Infrastructure (VDI).
— Graphical design tasks (e.g., 3D modeling, 3D rendering).
— High Performance Computing (HPC).
— Data science rendering.
— Machine learning training.
— Machine learning inference.
We keep our clients’ data secure with the help of a comprehensive 5-tier solution security system. It protects your data at all levels of access to it — from a data center to specific application. You can learn more about security at Servercore by following the link to the solution security page.
To address any questions regarding the technical or organizational aspects, make requests in the ticket in the control panel or email our engineers at hello@servercore.com. We will respond to you within one working day.
In the price of GPU server renting included:
— The assistance in the selection of a server with the necessary resources (SSD, RAM, GPU, etc.) for your tasks.
— Free traffic: unlimited port up to 1 Gbps.
— Private network — up to 1 Gbps.
— Public IPv4 address.
— KVM console.
— Automated OS installation and uploading your own ISO images.
— DDoS mitigation.
— Server assembly.
— SLA — 100%.
— 24/7 technical support.
— Component replacement in three hours.
Join the waitlist for the latest news, best deals and platform availability infomation.