Dedicated Server Benchmarks: Real CPU, NVMe, and Network Performance Numbers for Game Hosting

GHz numbers and core counts do not predict how a game server will actually run. Two CPUs with the same clock speed can differ by 60% on real workloads, and a disk that looks fast in sequential tests can choke on the random I/O that world saves produce. This article translates real benchmark numbers — CPU, NVMe, and network — into thresholds you can use when evaluating any dedicated server for game hosting.

CPU: Single-Thread Is the Bottleneck for Most Games

Most game server tick loops — Minecraft, Palworld, Valheim, ARK, and similar — run on one or two threads. That makes single-core performance the single best predictor of how many players you can host smoothly. Typical Geekbench 6 single-core ranges by CPU class in 2026:

CPU class	Typical single-core (Geekbench 6)	What it supports
Xeon E5 v3/v4 (2014–2016)	1,100–1,500	Small vanilla servers, 1–20 players
Xeon E-2300 / Ryzen 5000 (2021–2022)	1,700–2,000	50+ players on most games
Ryzen 7000 / Xeon E-2400 (2023–2024)	2,000–2,300	100+ players, modded servers, higher tick rates
EPYC 9004 / Xeon Gold (2022+)	2,000–2,300	Same single-thread ceiling, far more cores for multi-instance

Practical thresholds: a Minecraft server with view distance 10–12 and 50 players runs comfortably on a single-core score around 2,000; below 1,500 you will see tick lag during chunk generation. For Rust and ARK, which spread load across a few threads, look for at least 2,000 single-core plus 8+ physical cores. Multi-thread scores matter mainly when you run several instances on one box — there, roughly 4–8 GB of RAM and 2 cores per instance is a safer planning rule than core math alone.

Disk: Where NVMe Shows Up in Real Numbers

• Sequential read: PCIe 4.0 NVMe hits about 7,000 MB/s; SATA SSD tops out near 550 MB/s. Useful for modpack installs and map downloads, but not the number that predicts gameplay.
• Random 4K read: NVMe typically delivers 500,000+ IOPS (queued) and 50,000–100,000 IOPS at low queue depth, versus roughly 90,000 IOPS on a SATA SSD. World saves, chunk generation, and block lookups are random-I/O workloads — this is the number that separates a server that stutters on autosave from one that does not.
• Real effect: moving world files from a SATA SSD to NVMe commonly removes save-induced lag spikes, and cutting save frequency from every 5 minutes to every 30 seconds becomes painless.

When a host lists “SSD” without saying NVMe, ask. The difference is not marketing — it is a 5–10x gap on the exact I/O pattern game servers generate.

Network: Throughput Math That Predicts Bandwidth Needs

• Port math: 100 Mbps moves 12.5 MB/s; 1 Gbps moves 125 MB/s; 10 Gbps moves 1.25 GB/s. A 100 Mbps port becomes the ceiling for downloads, backups, and map transfers long before a 1 Gbps port does.
• Per-player usage: vanilla Minecraft typically uses 50–200 kbps per player with spikes far higher during chunk loads; Rust and ARK with many entities can exceed 500 kbps per player. A full 100-player Minecraft server rarely sustains more than 20–50 Mbps, but bursts and simultaneous downloads make a 1 Gbps port the safe baseline.
• Latency: benchmark ping from your region to the data center — under 20 ms is excellent for a regional server, 20–50 ms is good, and above 80 ms players will feel it. Each extra network hop (proxy, VPN, or NAT layer) adds roughly 0.5–2 ms when it is in the same data center.

Run Your Own Baseline Before You Commit

Any host worth using will let you benchmark during the refund window. A 30-minute baseline beats a week of forum reading:

• CPU: run Geekbench 6 (CLI) and compare single-core against the table above. For a quick proxy, use openssl speed or 7-Zip’s benchmark — both exercise single-thread performance.
• Disk: run fio with a 4K random-read profile on the data partition, plus a sequential test. Watch for numbers that are 3–10x below NVMe class — that is oversold or virtualized storage.
• Network: iperf3 to a second server in the same data center confirms you actually get the port speed you paid for, and ping tests from player locations confirm the route.

Thresholds at a Glance

Workload	Single-core score	Disk	Network
Vanilla Minecraft, 20 players	1,700+	SATA SSD acceptable	1 Gbps
Modded Minecraft, 50–100 players	2,000+	NVMe	1 Gbps
Rust / ARK, 100+ players	2,000+, 8+ cores	NVMe, RAID 1	1 Gbps
Multi-instance network	1,900+ per instance	NVMe, separate backup disk	1–10 Gbps

Benchmark numbers are the closest thing to a guarantee a spec sheet can offer. If a host will not let you test hardware during a refund window, that is an answer too. Once you know the numbers you need, you can compare dedicated server plans for gaming on our comparison page and filter by CPU generation, NVMe storage, and port speed before you ever open a ticket.