What Is Single Threads Performance And Why It Destroys Minecraft Server Lag

What Is Single Thread Performance and Why It Destroys Minecraft Server Lag

If you have ever asked your hosting provider why your Minecraft server lags, you have probably gotten an answer about RAM. Maybe it was about player slots. Maybe your modpack was too heavy.

Almost nobody mentions the one thing that actually matters most: single thread performance.

This article explains what single thread performance is, why it has more impact on Minecraft server lag than any other hardware spec, and how to evaluate whether the host you pay every month is running hardware actually capable of handling your server. We will go deep, with real benchmark numbers and a comparison table you can use to assess any host on the market.

The short version

Minecraft runs its main game loop, called the "tick", on a single CPU thread. The target is 20 ticks per second. If your CPU cannot complete one tick in 50 milliseconds or less, your server starts dropping ticks. Players see this as lag.

The single most important hardware question for a Minecraft server is therefore not "how many cores?" but "how fast is one core?"

A modern CPU with a high single thread benchmark will handle a heavy Minecraft server smoothly. A CPU from 2018 with 32 cores and 256 GB of RAM will still lag, because Minecraft cannot use any of those extra cores for the main tick loop.

Understanding this changes how you should buy hosting forever.

What single thread performance actually is

A modern CPU is built from multiple cores. Each core can execute instructions independently. Most CPUs also support hyperthreading or simultaneous multithreading, which means each core can handle two threads at once. A 16-core CPU with hyperthreading effectively presents 32 threads to the operating system.

Multi-thread performance refers to how much work a CPU can do when all of its cores are being used at once. This is what matters for things like video rendering, code compilation, and running 50 virtual machines.

Single thread performance refers to how fast one core can execute instructions. It is measured by benchmarks like PassMark CPU Mark Single Thread, Cinebench R23 Single Core, and Geekbench Single-Core.

Two factors drive single thread performance:

1. Clock speed (GHz). How many cycles per second the core can run.
2. Instructions per cycle (IPC). How much actual work the core gets done per cycle.

Modern CPUs have made huge gains in IPC over the past decade. A 2024 core running at 5.0 GHz is dramatically faster than a 2018 core running at 5.0 GHz, because the newer one does more work per tick of its clock.

This is why "GHz" alone is not a reliable measurement. A Ryzen 9 9950X at 5.7 GHz boost is roughly 60 percent faster per thread than an Intel i9-9900K at 5.0 GHz boost. They are both labeled around the 5 GHz range, but the gap in actual single thread output is enormous.

Why Minecraft is bottlenecked by one core

Minecraft was first released in 2009. Its codebase has evolved enormously since then, but the fundamental architecture of the server has stayed the same. There is one main thread that runs the game logic, and almost everything important happens inside it.

Inside the main thread, the server runs what is called the tick loop. Every 50 milliseconds, the server has to:

• Process movement for every player
• Run AI logic for every loaded mob
• Calculate redstone signals
• Save chunk state
• Handle plugin events
• Tick growing crops, dripstones, leaves, and other block entities
• Calculate lighting updates
• Spawn and despawn entities

If the server completes all of this in under 50 ms, it hits the target of 20 ticks per second. If it takes longer, the tick rate drops below 20.

Once you fall below 20 TPS, players experience the entire game world running in slow motion. Mobs walk slower. Crops grow slower. Day and night cycle slows. Combat hits register late. Redstone fires slowly.

This is what Minecraft players call TPS lag, and it is the single most frustrating experience on a hosted server.

The critical thing to understand is that every one of those tick activities runs sequentially, on the same thread. Adding more cores to the box does nothing to make the tick complete faster. Only making the one thread faster helps.

There are some exceptions. Modern fork projects like Folia attempt to multithread the tick loop by region, and Paper has offloaded chunk generation and some I/O to other threads. But for the vast majority of Vanilla, Spigot, and Paper servers running today, the rule still holds. One core does almost all the work.

What 20 TPS actually means

20 TPS is not just a number. It is the heartbeat of the entire game.

Mobs are programmed to move a certain distance per tick. Doors open over a certain number of ticks. Furnaces smelt over a certain number of ticks. The whole game world is measured in ticks, not seconds.

When TPS drops to 15, everything in the world runs at 75 percent speed. When TPS drops to 10, half speed. When it drops to 5, the game becomes unplayable.

The tricky thing about TPS is that the game often shows green icons or "good ping" indicators even when TPS is dropping. Ping measures network latency, which is unrelated to TPS. You can have perfect ping and a server that is essentially frozen.

The only reliable way to check TPS in Vanilla or Paper is to run the /tps command (Paper or Spigot) or look at server console output. On Vanilla, you can use /debug start and /debug stop to get a performance report.

A healthy Minecraft server should show 20.0 TPS consistently, with occasional dips during chunk loading or large redstone activations. A server that sits at 17 or 18 TPS during normal play is suffering from a hardware bottleneck or a plugin problem. It is not normal, and it is not okay.

The benchmark table that matters

Here is the comparison that should drive every hosting purchase decision. The numbers are PassMark Single Thread scores, which is the standard benchmark used to compare per-core performance across CPUs.

Look at the gap between the bottom of the table and the top. A 2014 Xeon scores about 1,800. A 2024 Ryzen 9 9950X scores about 4,650. That is a 2.6x difference in single thread output.

In Minecraft terms, this means the same exact server, the same exact world, the same exact players, will run at vastly different TPS depending on which CPU it lives on. A modpack that runs at 20 TPS on a 9950X might struggle at 12 TPS on a 9900K. The world itself does not change. Only the speed of one core does.

Why "more RAM" is the wrong question

When players ask hosts about lag, the standard sales response is to upgrade the RAM tier. Lag? Buy 8 GB instead of 4. Still lag? Buy 16 GB instead of 8.

This works for one specific kind of lag: out-of-memory crashes and chunk loading slowness. If your server is running modpacks with hundreds of mods and your RAM is too small to hold the loaded world, you will have problems. Adding RAM fixes that.

But if your TPS is dropping while you have plenty of free RAM, the bottleneck is your CPU. Adding RAM does nothing. You can put 256 GB of RAM into a 2018 CPU and your TPS will stay exactly where it was.

This is one of the most common ways players get sold up at a hosting company. The host knows that if they keep blaming RAM, the customer keeps buying bigger plans. The customer keeps lagging, because the actual problem is never addressed.

A simple test. Log into your server, run /tps on Paper, or /debug start on Vanilla followed by /debug stop after 10 minutes. If your TPS is dropping while RAM usage stays under 70 percent of allocated, your CPU is the bottleneck.

Why "more cores" is also the wrong question

Hosts love to advertise core counts. "12 cores!" "32 cores!" These numbers sound impressive, but they are mostly irrelevant for Minecraft.

A Minecraft server's main tick loop runs on one thread. It cannot be split across multiple cores. Adding cores to the box does not make your tick finish faster.

The only places where extra cores help a Minecraft server are:

• Chunk generation (offloaded to a separate thread in Paper)
• Network packet handling (a few additional threads)
• Logging and saving (background tasks)
• Plugin background work (depends on the plugin)

Even in best-case scenarios, having more than 4 cores per Minecraft server provides almost no benefit. What matters is whether one of those cores is fast.

A CPU with 32 cores at 3.5 GHz will lose to a CPU with 8 cores at 5.7 GHz, every single time, for Minecraft hosting purposes.

How to test the host you are looking at

Before you sign up with any host, ask them one specific question. What CPU model do you run? Provide the exact name.

A reputable host will tell you immediately. They will say something like "AMD Ryzen 9 7900X". You can then look up the PassMark Single Thread score on cpubenchmark.net and compare it to the table above.

If the host gives you a vague answer like "high-end enterprise grade" or "the latest Intel processors", that is a signal. Either they are running older hardware they are embarrassed about, or they oversell their boxes so heavily that they cannot guarantee what your server will actually run on. Both are bad.

You can also look at the Bukkit or Paper TPS report from any existing server you have access to. If you are currently lagging at 15 TPS in a modpack with light player load and reasonable RAM, your current host is running outdated CPUs. Time to switch.

What this means for you

The hosting industry has spent years convincing customers that the important specs are RAM, slot count, and bandwidth. These things matter, but they are second-order to single thread CPU performance.

If you want a Minecraft server that runs smoothly under load, your money is best spent on a host that runs current-generation high-clock CPUs. Anything else is a compromise.

For our part, Server Heron runs the AMD Ryzen 9 9950X across our entire fleet. Not just our top tier. Every customer, on every plan, is on the same generation of hardware. The 9950X currently sits at #27 on PassMark's all-time single thread leaderboard out of more than 5,800 CPUs ever benchmarked. Only the newest X3D variants score higher, and our roadmap includes upgrading to those when they become available at our scale.

We chose this hardware because it is the right answer to the actual problem. Minecraft is bottlenecked on single thread performance, so we put the best single thread performance available into every server we sell. There is no upsell tier. There is no premium plan hiding the good hardware.

If you are currently paying for hosting on hardware older than 2022, you are paying 2024 prices for 2018 performance. The math works against you every month you stay on it.

A short FAQ

Does this apply to modded servers too?
Yes. In fact, modded servers are usually more CPU-heavy than vanilla, because mods add more entities, more block entities, and more logic per tick. Single thread performance matters even more on modded servers.

Will Folia change this?
Folia, the experimental fork of Paper, can split the tick loop across regions of the world. It helps when you have many players spread across a large map. But it is still in early adoption, and most servers, plugins, and modpacks are not Folia-compatible. For the foreseeable future, single thread performance is the right thing to optimize for.

What about shaders, ray tracing, and graphics?
Server-side, none of these matter. The server does not render anything. All graphics processing happens on each player's own machine. The server only sends game state.

How often do CPU generations actually matter?
PassMark Single Thread improvements have averaged 10 to 15 percent per generation since 2018. After six or seven years, that compounds into a 60 percent or larger gap between an old and new CPU. That is the difference between a smooth 20 TPS and a server that sits at 12 TPS during peak hours.

Wrapping up

Single thread performance is the single most important hardware spec for a Minecraft server. If your host runs a CPU with a low PassMark Single Thread score, your server will lag, no matter how much RAM or how many cores it has.

Before paying for any Minecraft hosting plan, ask the provider for the exact CPU model. Look it up. Compare it to the table above. Make your decision based on what actually drives performance, not what the marketing copy emphasizes.

Hosting is one of those products where the marketing rarely matches the reality. The numbers tell the truth. Once you know how to read them, you stop overpaying for outdated hardware.