How to Get True 10Gbps Speed from Your Linux Dedicated Server

📅 August 21, 2026 - If you have recently upgraded your bare-metal dedicated server to a 10Gbps network port, you might be frustrated to see your actual transfer speeds stuck around 2 Gbps or 3 Gbps.

Having a 10Gbps port does not automatically guarantee 10Gbps of real-world throughput. At Servers99, we frequently help clients optimize their infrastructure. Here is a text-based checklist of what you need to tune on a Linux server to unlock your full bandwidth.

1. Verify Your Physical Link First
Before you change any Linux network settings, you must verify your hardware. Use network diagnostic tools like ethtool to check your negotiated link speed. If your server reports a 1000Mb/s speed, you have a physical layer issue—like a faulty cable, incorrect transceiver module, or a switch misconfiguration. Software tuning will not fix a hardware limit!

2. Stop Bottlenecking a Single CPU Core
At 10 Gigabits per second, your server is processing millions of packets. If all hardware network interrupts are sent to a single CPU core (usually CPU0), that core will hit 100 percent utilization. This throttles your network while your other CPU cores sit idle. Ensure your server is using Receive-Side Scaling (RSS) and that the irqbalance service is active to distribute the load evenly across all processor cores.

3. Choose the Right MTU
The standard Maximum Transmission Unit (MTU) is 1500. This is the safest default for internet-facing servers. However, if you are moving data between servers on a private backend network, switching to Jumbo Frames (MTU 9000) can heavily reduce CPU overhead. Just remember that every switch and router in the path must support Jumbo Frames for this to work.

4. Tune TCP Socket Buffers
For high-bandwidth connections, especially over geographic distances, standard Linux TCP buffers are often too small. You will need to increase the read and write socket buffer maximums in your sysctl configuration. Do not blindly set massive numbers; start with a conservative increase (like 16MB limits) and test your baseline.

5. Benchmark the Right Way
Never use a browser-based speed test to validate a 10Gbps server. The browser overhead will give you inaccurate results. Instead, use a tool like iperf3 between two 10Gbps-capable servers. Make sure to run the test using multiple parallel streams (for example, 8 streams) to see if a single TCP flow is the actual bottleneck.

Want the full step-by-step technical guide?
We have published a complete 17-step guide covering the exact commands, TCP buffer configurations, and troubleshooting steps for packet drops.