Cross-play stopped being a differentiator years ago. It is now closer to a baseline expectation, and players notice immediately when a cross-platform lobby feels worse than a same-platform one.

As far back as 2021, Unity estimated that over 50% of Americans played multiplayer games, and that 87% of those players had played cross-platform. The appetite was already there. What has not kept pace, in a lot of shipped games, is the infrastructure underneath it.

The complaint is consistent enough to be diagnostic: cross-play lobbies lag more than same-platform lobbies, even when the players live on the same street. That is rarely a netcode problem. It is almost always a routing problem.

Why Cross-Play Traffic Takes the Long Way Around

Plenty of studios start with peer-to-peer because it costs nothing to run. No servers, no relays, no per-CCU bill. For a same-platform lobby sitting on a single first-party network, that works reasonably well.

Cross-play complicates it in two ways.

The first is technical. NAT traversal succeeds at very different rates depending on which networks you are trying to punch through. Console networks, carrier-grade NAT on mobile, and a home PC behind a consumer router do not behave the same way. Direct connections that hold up fine inside one ecosystem start failing once you mix them, so you need a fallback path.

The second is that the fallback path is usually centralized. Once a studio adds relays or dedicated servers to cover the connections P2P cannot handle, those get deployed in a handful of regions to keep the bill down. Three or four locations, sometimes fewer.

So here is the scenario players actually describe. You are on PlayStation, your friend across the street is on Xbox. Physical distance between you: a few hundred metres. Distance your packets travel: to a datacenter several hundred kilometres away and back, for every single update, in both directions.

The netcode is fine. The route is not.

To be precise about something the industry gets loose with: cross-platform P2P is not forbidden. Epic Online Services and Steam Datagram Relay both carry cross-platform sessions across PlayStation, Xbox and PC today. The constraints are first-party certification requirements on specific features, and traversal reliability across mismatched network types. Those are real constraints. They are not the same as "a PlayStation cannot talk to an Xbox."

Relays, Dedicated Servers, and What Each One Actually Fixes

It is worth separating three things that get used interchangeably.

• Peer-to-peer sends packets directly between clients. Lowest cost, lowest latency when it connects, no authoritative state, and no fallback when traversal fails.

• Relays forward packets between clients without inspecting them. They fix traversal and they hide player IPs. They do not run game logic.

• Dedicated servers run the simulation and hold authority. Highest cost, best anti-cheat position, and the only option if you need the server to make decisions.

Most cross-play games end up using more than one of these. That is a sensible outcome, not a design failure. Our explainer on authoritative servers, relays and P2P covers which fits which genre in more depth.

What matters for lag is that a relay and a dedicated server are both a hop, and a hop you place badly costs you the same latency either way.

Relays Solve Distance. They Do Not Solve Incompatibility

There is a failure case worth naming because it looks like a networking problem and is not one.

A studio ships a console version in Unreal. A year later, a co-development partner ports it to mobile, and for schedule reasons that port ends up on a different networking stack. Now the two builds do not speak the same wire protocol.

No relay fixes this. A relay forwards bytes it never parses, so if the console client emits a packet the mobile client cannot decode, moving that packet through a relay changes nothing. The same is true of direct P2P.

Two things do fix it. Either both builds converge on one wire protocol, which is the cleaner answer and the one worth fighting for during production. Or you run a server-side component that parses one format and emits the other, which is a protocol gateway rather than a relay, even when it is deployed on the same infrastructure and looks like one from the outside.

Full engine swaps mid-port are uncommon. The narrower version of this is not. Different Unreal versions across builds, a replication rewrite to fit mobile bandwidth budgets, a swapped transport layer for a platform SDK. Each of those can drift the wire format enough to break interoperability, and each of them is easier to catch during production than after launch.

More Locations Is the Lever That Actually Moves Latency

Once traffic has to route through a hop, the only structural way to shorten it is to put that hop closer to the players.

This is unglamorous and it is also the whole game. Compression, tickrate tuning, and lag compensation all help at the margins, and none of them beat physics. A packet crossing an ocean pays that cost every tick, in both directions, for the whole match. We wrote about why more locations is the only real answer to latency if you want the longer argument.

A lot of relay and hosting providers advertise global coverage and deliver somewhere between five and thirty locations. That is enough to get a session running. It is not enough to keep two neighbours on different consoles inside a low-latency envelope, which is the exact case cross-play players complain about.

What This Looks Like in Practice

Edgegap deploys game servers and relays across 615+ locations worldwide on a regionless network, spun up on demand rather than reserved in advance. Servers boot from cold start in roughly 3 seconds, which is what makes on-demand placement viable instead of a nice idea.

The measured result across studios is a 58% average latency reduction, with sub-50ms latency reaching 78% of the player base. For competitive titles, closing the geographic gap between matched players also improves fairness by 28%, because the advantage that comes from living near a datacenter shrinks when the datacenter is near everyone.

None of that removes the possibility of an outage. What it removes is the single-region dependency that turns one datacenter having a bad day into your entire player base having a bad day.

Cross-play was never the thing that made games feel laggy. Centralizing the traffic it requires was.

If you are evaluating how to route your cross-play traffic, our breakdown of dedicated servers for cross-platform play covers the build-versus-buy side of the decision, and you can start for free.