New Server Hardware

This blog post provides an overview of my new server hardware and explains the reasoning behind each major hardware decision.

Use Case

This server is primarily used as a centralized data storage and file-sharing system, enabling data exchange between devices and users while improving everyday workflows. It is designed to run reliably 24/7 with room for future expansion.

Tip: Clearly define your use case before buying hardware.
Storage-heavy workloads, virtualization, and media servers all benefit from very different hardware choices.

General Specs

• AMD Ryzen 9 5950X @ 3.40 GHz (16 cores / 32 threads)
• 64 GB DDR4-3200 RAM
• 5 TB M.2 SSD storage
• 8 TB HDD storage

Platform

I already had a motherboard, 32 GB of DDR4 RAM, and a PSU, so I decided to stay on AM4.

Tip: Reusing existing components can significantly reduce costs.
A slightly older platform is often more than sufficient for homelab and self-hosted workloads.

CPU

My previous CPU on this motherboard was a Ryzen 5 3600X. While it was still perfectly usable, the increasing number of services and workloads made a higher core count necessary.

Based on the AM4 platform, these were my upgrade options:

Not all of these options made sense for my specific use case, so I narrowed them down based on generation, performance, and price.

I ruled out the Ryzen 9 3900X due to its older CPU generation. Zen 3 offers significantly better performance per watt and higher IPC compared to Zen 2, which is especially relevant for a server running 24/7.

I also decided against the Ryzen 9 5900XT because, at the time, it was more expensive while offering lower base clock speeds compared to the Ryzen 9 5950X, resulting in a worse price-to-performance ratio.

While the Ryzen 9 5900X would have been a solid option, I ultimately wanted to maximize core count on the AM4 platform.

I decided on the Ryzen 9 5950X because I already had an AM4 motherboard and needed a CPU with a high core count for parallel workloads. With 16 cores and 32 threads, it represents the most powerful CPU available for AM4 and was therefore the best fit for my use case.

Tip: For server workloads, core count and efficiency often matter more than maximum boost clocks.
Always consider power consumption for systems running 24/7.

RAM

I already had 32 GB of DDR4-3200 RAM installed, so I expanded the system with an additional 32 GB of DDR4 memory.

While the platform only supports dual-channel memory and does not offer true quad-channel operation, using four DIMMs provided the best balance between cost, capacity, and real-world performance for my use case.

Tip: Capacity is usually more important than memory speed for servers.
Running stable DDR4-3200 is often preferable to higher clocks with potential stability issues.

Storage

I use a 1 TB M.2 PCIe 3.0 x4 SSD for my Proxmox installation and VM storage. Low latency is important to me, as I want disk access to be as responsive as possible.

In my case, this is a 1 TB PCIe 3.0 x4 SSD with read and write speeds of around 2 GB/s. A faster SSD is not necessary, as the network is the main bottleneck in this setup.

For high-activity server data, I use two 2 TB PCIe 3.0 SSDs configured in a RAID 0 array, resulting in 4 TB of usable storage.
Note: RAID 0 improves performance but provides no redundancy.

For backups, I decided on two 4 TB HDDs configured in RAID 1 to ensure redundant local backups of my data.

Tip: Separate performance-critical data from backup storage.
SSDs are ideal for active workloads, while HDDs offer better cost efficiency for backups.

Tip: RAID is not a backup.
Important data should always be backed up to an external system or off-site location.

In a later article, I will write about my complete backup strategy.

Key Takeaways

• Clearly define your use case before selecting hardware
• Reuse existing components to reduce costs
• Prioritize reliability and efficiency over peak performance
• Separate active data from backup storage