When to upgrade your plan, signals from TPS, RAM, and player count

Most "upgrade your hosting plan!" advice is from hosts. Self-interest aside, sometimes the upgrade is genuinely needed, and sometimes it's not. This article helps you tell the difference using actual metrics.

The wrong reasons to upgrade

Before the right ones, the bad reasons:

A player complained about lag. Once. Could be their Wi-Fi. Diagnose first.

You feel like the server is "slow." Vague. Check actual numbers.

Marketing convinced you "more RAM is better." Often false. See the RAM article.

A hosting sales rep recommended it. Maybe genuine, often biased.

Real signals vs noise

Signals that reward an upgrade	Signals that won't be fixed by more RAM
TPS sits below 18 even after the lag-diagnostic checklist	Bad client-side performance (one player's PC is the bottleneck)
Heap-after-GC sits above 80% of allocation	Network ping from your region is already great
Players consistently can't connect because the slot count is full	Specific plugins or mods are bugged (a fix is upstream, not RAM)
You're adding 5+ new players who'll be active during peak	The single player whose base lags everyone is the actual issue
The pack documentation says you're under spec for your player count	You haven't yet tried tuning JVM flags or simplifying mob farms

The right signals

Real reasons to upgrade, with the specific numbers to look for.

Signal 1: TPS regularly below 19

For Minecraft and similar, healthy TPS is 19.5 to 20. Brief dips during heavy events are normal.

If TPS sits at 17 to 19 during normal play, you have headroom problems. Sometimes this is a software issue (entity-heavy farms, bad plugins) and not a hardware one. Diagnose first (Spark profiler, see #15).

If you've optimized software and TPS is still consistently below 19: upgrade.

Signal 2: Heap consistently >75% used after GC

For Java-based servers (Minecraft especially), watch the post-GC heap.

• Heap allocated: 6 GB.
• Heap used after GC: 5 GB.
• Ratio: 83%.

This is starving. The JVM can't reclaim enough memory; you'll see longer GC pauses and increasing frequency. Either upgrade or aggressively reduce load (fewer chunks, fewer entities).

Signal 3: Player count regularly hits the comfortable limit

Each plan has a soft player limit beyond which performance degrades. If you're hitting that limit at peak hours:

• The soft limit isn't a hard cap; players will still connect. But each additional player past the optimum cost is more than the previous one.
• The signal: peak-hour TPS dips below idle TPS noticeably.

If your server is fine at 5 players, hits 18 TPS at 15 players, and is at 12 TPS at 25 players, you're at your scaling wall.

Signal 4: Save spikes are noticeable

Every save, players feel a half-second to one-second freeze. If saves are spiking your tick noticeably:

• Disk speed is your bottleneck.
• Common on shared / oversubscribed hosting.
• Upgrade to a plan with better disk (NVMe) or move to a host with better isolation.

Signal 5: Backups can't finish within their interval

Hourly backups taking 70 minutes is a problem. Daily backups taking 23 hours is a worse problem. If your backup is fighting the live server for resources:

• Network bandwidth or storage I/O is constrained.
• Upgrade the plan, the backup destination, or both.

Signal 6: Persistent crashes from OOM

The server logs OutOfMemoryError every few days. RAM is fundamentally insufficient.

Upgrade RAM. Or reduce load aggressively.

Signal 7: You've grown into a different category

You launched as a "5 friends" server. Now you have 20 active regulars. The plan from 5 friends doesn't fit 20 regulars even if it was great at launch.

Re-evaluate plan size against current scale.

A diagnostic ladder before upgrading

Each rung climbed without an upgrade saves you money and teaches you about your server.

How much to upgrade by

The trap: upgrade by a tiny amount, hit the same wall in 2 months, upgrade again.

When upgrading, jump enough to give yourself 6+ months of headroom. For RAM, that means roughly 50 to 100 percent more than your current use. For CPU, the next tier.

Small upgrades trap you in the cycle. Bigger ones break out of it.

The downgrade case

Less common but real. If you've upgraded over time and the server has stabilized, ask:

• Was the upgrade actually necessary? (Did the issue have a non-hardware cause that we later fixed?)
• Have we trimmed entities, plugins, players since then?
• Can we test on a smaller plan for a week?

Some hosts allow easy downgrades. Most charge for the upgrade and let you keep the resources. If you can downgrade, periodically do so to confirm you're not over-paying.

A typical lifecycle

A reasonable server lifecycle:

• Month 0: Launch on 4 GB, 5 players.
• Month 3: Community has grown to 10 players. TPS healthy. No change.
• Month 6: 15 players, occasional dips at peak. Run Spark, audit, fix.
• Month 9: 20 players, still dipping. Audit again, fix what you can. If still tight, upgrade to 8 GB.
• Month 18: Server at year-and-a-half. Major review. Possibly upgrade CPU tier if engagement is sustained.

This is the path of a healthy server. Most servers should upgrade at most once or twice per year, not constantly.

How to evaluate a "do I need this upgrade" decision

A simple framework:

1. What metric tripped my concern? (Specific TPS, RAM, etc.)
2. Have I optimized software first? (Spark, audit, etc.)
3. If I upgrade by X, what specific metric do I expect to improve?
4. Can I A/B test by tuning down config temporarily to simulate the upgrade?

If you can answer these clearly, you're making an informed decision.

Conclusion

Upgrading hosting is sometimes the right move and sometimes hiding a software problem with money. The signals are specific: TPS dips, heap pressure, save spikes, OOM errors. The temptations are also specific: vague complaints, marketing, hope.

Use real metrics. Optimize first. Upgrade meaningfully when you do. Don't upgrade often.