Skip to main content

Select components for a dedicated server of custom configuration

Last update:

You can assemble and order a server of custom configuration.

Depending on the components, server assembly takes from one to five working days. Read more about building servers of custom configuration in the Selectel blog article How we build custom servers: from configurator to testing.

When the server is ready for delivery, you will receive an email notification and will be able to connect to it.

If you need help with selecting components, create a ticket or contact us via other methods.

Select components

When ordering a server of custom configuration you select components in the server configurator. Not all components are compatible with each other. When adjusting the configuration, incompatible components are automatically excluded after selecting each new component.

Processor

For your information

After the server is ordered, the processor cannot be replaced. If a different processor is required or you need to change the number of processors, order a new server.

Intel and AMD processors are available for order.

In the configurator, processors are grouped by lines and families. The processor name indicates the model, core count, base clock frequency, and additional designations.

Processor examples:

  • Intel Silver 4510 (12x2.4 GHz HT);
  • AMD EPYC 7443P (24x2.85 GHz SMT).

Processors may have additional designations:

HTSupport for Intel® HyperThreading technology, which allows one physical processor core to execute multiple instruction threads simultaneously. Read more about the technology and operational principles in the Selectel blog article What is Hyper-Threading
SMTSupport for AMD Simultaneous Multithreading technology, which allows one physical processor core to execute multiple instruction threads simultaneously
EOL (End Of Life)A processor line or family that has been discontinued
CPU BenchmarkProcessor performance test result

Memory

DDR4 and DDR5 memory modules are available for order.

In the configurator, you can select:

  • total memory capacity;
  • or the size and quantity of modules. The configurator will automatically display the total ordered capacity. Use this option to select the number of modules based on the processor memory channel count.

The RAM name indicates the capacity, generation, and additional designations.

RAM example: 64 GB DDR4 ECC Reg.

RAM may have additional designations:

RegRegistered memory type. The presence of registers reduces the electrical load on the controller and allows for installing more memory modules in a single channel
ECCAn error detection and correction mechanism that ensures data stability and reduces the risk of failures
NON-ECCWithout an error correction mechanism

Volumes

Volumes are installed in the following order: NVMe ⟶ SSD ⟶ HDD. If you need a different installation order, specify it in the comments when placing your order.

In the configurator, volumes are grouped by type. The volume name indicates the capacity, volume type, and additional designations.

Volume example: 500 GB SSD NVMe M.2.

Read more about volume types in Selectel blog articles:

Volume types

HDDHard disk drives
SSD SATASolid-state drives that connect via the SATA interface
SSD NVMeSolid-state drives that connect via the high-speed PCI Express interface. The NVMe interface allows SSDs to achieve maximum read/write speeds and minimizes latency
OPTANEIntel® SSDs based on 3D XPoint technology. They can be used as cache memory, which speeds up OS loading and improves performance with HDDs and SSDs
Hi-PerformanceSSDs or HDDs for tasks requiring high data transfer speeds, minimal latency, and high reliability under intensive workloads
EnterpriseSSDs or HDDs designed for data centers and high-load computing systems. Optimized for intensive workloads

SSD form factors

Volumes may have additional designations:

HHHL (Half-Height, Half-Length)SSD form factor. Characterized by reduced dimensions. Connects via PCI-e interface
U.2SSD form factor. Supports hot-swapping
M.2SSD form factor. Does not support hot-swapping

HBA and RAID controllers

RAID controllers are devices for organizing and managing disk arrays using RAID technology. Controllers manage RAID operations at the hardware level, offloading the system processor.

Disk controllers are needed when the number of disks exceeds the number of SATA ports on the motherboard and for additional data protection using hardware RAID. Read more in the Selectel blog article Hardware RAID: features and use cases.

HBA and RAID controller types

HBA (no RAID) 16 ports
  • connects up to 16 drives via SATA and SAS interfaces;
  • hardware RAID array creation is not supported
RAID controller 8 ports + Cache Protection
  • connects up to 8 drives via SATA and SAS interfaces;
  • supports hardware RAID arrays;
  • protects cache data in case of sudden power failure
RAID controller 16 ports + Cache Protection
  • connects up to 16 drives via SATA and SAS interfaces;
  • supports hardware RAID arrays;
  • protects cache data in case of sudden power failure
RAID controller 24 ports + Cache Protection
  • connects up to 24 drives via SATA and SAS interfaces;
  • supports hardware RAID arrays;
  • protects cache data in case of sudden power failure
RAID controller 16 ports + SSD Cache
  • connects up to 16 drives via SATA and SAS interfaces;
  • uses an array of fast SSDs to buffer read/write requests to arrays;
  • stores cache in case of sudden power failure

Network cards

Every server of custom configuration is equipped by default with two RJ45 network interfaces with a data exchange speed of 1 Gbps each. The first interface is used for internet connection, and the second for the private network — learn more in the guide Dedicated Server Networks.

If you need more network interfaces or need to reserve a connection using MC-LAG, you can order additional network cards.

Network cards are available in the following slots:

  • PCI-e (Peripheral Component Interconnect Express) — a slot on the motherboard for connecting expansion boards, e.g., GPUs, RAID controllers, SSDs, network cards, etc.;
  • OCP (Open Compute Project slot) — a slot on the motherboard for connecting mezzanine network cards. OCP is an alternative to PCI-e slots, which may be occupied by other boards.

Types of network cards

2 x 1 GE + 1 Gbps local portAn additional network card with two 1 Gbps interfaces is installed. One port connects to the local network
2 x 1 GE + MC-LAG to 1 Gbps local networkAn additional network card with two 1 Gbps interfaces is installed. Both interfaces connect to the local network
2 x 1 GE + MC-LAG to 1 Gbps internet networkAn additional network card with two 1 Gbps interfaces is installed. Both interfaces connect to the internet
2 x 10 GE + 10 Gbps local portAn additional network card with two 10 Gbps optical interfaces is installed
2 x 10 GE + MC-LAG to 10 Gbps local networkAn additional network card with two 10 Gbps optical interfaces is installed
2 x 10 GE + MC-LAG to 10 Gbps internet networkAn additional network card with two 10 Gbps optical interfaces is installed
2 × 10 GE + 10 Gbps network volume SAN connectionAn additional network card with two 10 Gbps optical interfaces is installed with the ability to connect to a SAN network of storage system LUNs
2 × 10 GE + 10 Gbps network volume SAN connectionAn additional network card with two 10 Gbps optical interfaces is installed with the ability to connect to a SAN network of network volumes

GPU

A GPU (Graphics Processing Unit) is a graphics processor designed for accelerated data processing. It is used for tasks such as:

  • machine learning;
  • data analytics;
  • scientific computing;
  • Virtual Desktop Infrastructure (VDI) deployment;
  • video broadcasting;
  • graphics processing, video rendering, etc.

Read more about the terms and choosing a video card for specific tasks in the Selectel blog article Can desktop video cards compete with server-grade ones.

For servers with (GPU), video cards are available as PCI-e devices. Up to 8 video cards can be installed in a single server.

In the configurator, the video card name indicates the model, capacity, and video memory type. GPU example: RTX A5000 24 GB GDDR6.

PCI-e devices

The trusted boot module is a certified hardware/software module that:

  • helps prevent unauthorized access to the server during the operating system (OS) boot stage;
  • monitors the integrity of the server's hardware configuration and files before the OS boots;
  • blocks booting from removable media.

It is used for system certification to meet security requirements or to fulfill environment requirements for certified cryptographic information protection tools with a KS3 class.

PCI-e device example:

Motherboard

When ordering a server in the configurator, only motherboards compatible with the selected components are displayed. If only one motherboard is compatible, the field will be filled automatically.

Chassis

When ordering a server in the configurator, only available chassis compatible with the selected components are displayed. If only one option is compatible, the field will be filled automatically.

To ensure uninterrupted server operation, a chassis with two or more power supplies is recommended. If you might need to upgrade your server in the future, choose a chassis with a larger number of free drive bays.

Recommendations for component selection by task

Gaming servers

For servers that synchronize game processes, high processor frequency and large RAM capacity are important.

ComponentsRecommendations
Processor
  • core count:

    • 4-6 for servers supporting up to 30 players;
    • 8 cores and above for servers or hosting with multiple game sessions;
  • frequency — 3.5 GHz and above;

  • Processor cache — 8-16 MB.


Processor examples:

  • AMD Ryzen 9 7950X, 4.5 GHz, 16 cores;
  • Intel Core i9-13900K, 3 GHz, 24 cores;
  • AMD Ryzen 5 5600G, 3.9 GHz, 6 cores
RAM
  • type — ECC RAM;
  • capacity:

    • 16 GB for servers supporting up to 20 players;
    • 16-32 GB for servers supporting up to 50 players;
    • 64 GB and above for servers supporting 50+ players

Servers with large databases

Large databases require a high-performance processor with scalability. A high volume of active database operations requires more processor cores. Server performance is affected by RAM, as it is used for data caching, query processing, and performing operations on them.

ComponentsRecommendations
Processor
  • minimum core count — 8;
  • frequency — from 2.4 GHz. High frequency will ensure fast execution of operations on each core, which is important when processing transactions;
  • multi-threading support — will allow more threads to execute on each core, increasing efficiency for multi-threaded tasks.


Processor examples:

  • Intel Xeon W-2255, 3.7 GHz, 10 cores;
  • AMD EPYC 7513, 2.6 GHz, 32 cores;
  • Intel Core i9-13900K, 3 GHz, 24 cores
RAM
  • type — ECC RAM;
  • Minimum capacity — 64 GB. The memory capacity must be larger than the database size
VolumesType — SSD NVMe

1C servers

1C:Enterprise and its databases are demanding on processor performance and actively use RAM for calculations and data caching. Many 1C operations are single-threaded, so they require a processor with a high frequency. Data access speed is important in 1C operations, especially with a large number of transactions and active read/write operations.

ComponentsRecommendations
Processor
  • minimum core count — 8;
  • frequency — from 3 GHz;
  • multi-threading support — will allow better handling of parallel tasks from multiple users and server tasks simultaneously.


Processor examples:

  • Intel Xeon E-2456, 3.3 GHz, 6 cores;
  • AMD Ryzen 7 7700X, 4.5 GHz, 8 cores;
  • Intel Core i9-13900K, 3 GHz, 24 cores
RAM
  • type — ECC RAM;
  • minimum capacity — 16 GB. The more data in the system and concurrent users, the more RAM is required
VolumesType — SSD NVMe

Servers for virtualization

Virtualization requires significant computing resources, especially under conditions of high virtual machine (VM) density and heavy loads. The more cores and threads the processor has, the more VMs can operate on the server. If you need a larger number of cores, you can assemble a server with two or more processors.

Each VM requires allocating a portion of RAM for its operation. The more VMs on a server, the more RAM is required.

Storage system performance directly affects VM performance.

ComponentsRecommendations
Processor
  • minimum core count — 16. For high-load servers, 32+ cores are recommended;
  • frequency — 2.2-3.5 GHz;
  • multi-threading support — will increase system performance, allow assigning higher VM virtual core counts, and optimize resource allocation;
  • support for virtualization technologies such as Intel VT-x or AMD-V — they distribute resources between virtual machines and improve their performance.


Processor examples:

  • Ampere Altra Max M128-30, 3 GHz, 128 cores;
  • AMD EPYC 7763, 2.45 GHz, 64 cores;
  • Intel Xeon Silver 4214, 2.2 GHz, 12 cores
RAM
  • type — ECC RAM;
  • Minimum capacity — 64 GB
VolumesType — SSDs
GPU

GPUs with virtualization support are recommended, which will allow sharing capacities between several VMs.
GPU examples:

  • NVIDIA Tesla T4;
  • NVIDIA A16;
  • NVIDIA A2

Machine learning models

For servers intended for machine learning models, high-performance processors, GPUs with large video memory, large RAM capacity, and high-speed storage are important.

In machine learning, the main load falls on the GPU. The more parameters and the more complex the training model, the more GPU memory is required for efficient operation. Read more in the Selectel blog article How to choose a video card for training neural networks and Deep Learning.

ComponentsRecommendations
Processor
  • minimum core count — 16. The more cores, the better the processor handles data pre-processing and task distribution between GPUs;
  • frequency — from 2.4 GHz;
  • multi-threading support — will increase efficiency for multi-threaded tasks.


Processor examples:

  • AMD EPYC 7513, 2.6 GHz, 32 cores;
  • Intel Xeon Gold 6240R, 2.4 GHz, 24 cores
GPU
  • minimum video memory — larger than the training model size;
  • support for NVIDIA NVLink technology — will provide high bandwidth and allow transferring large data volumes between the GPU and other system components;
  • support for NVIDIA CUDA or AMD ROCm — will increase computation performance and speed through parallel data processing on the GPU.


GPU examples:

  • NVIDIA Tesla A100;
  • NVIDIA Tesla H100;
  • NVIDIA RTX6000
RAM
  • type — DDR5 ECC;
  • Minimum capacity — 64-128 GB
Volumes
  • type — SSD NVMe;
  • Minimum capacity — 1 TB
RAIDUsing RAID arrays will increase data access speed and improve system fault tolerance