Samy Duc, now Technical Director at Respawn Entertainment, provided insights into Apex Legends backend in 2021 highlighting how APEX Legends’ servers, netcode, and infrastructure worked behind every match. Providing an in-depth of Respawn’s blueprint to its online infrastructure.

Indirectly, this highlights best practices that any game studio, big or small, can add to their multiplayer to help improve its performance.

Player-Facing Performance Dashboard

Starting at Season 6 (November 2020), the APEX Legends team added a performance overlay. This in‑game dashboard displays upload and download rates in kilobytes per second, round‑trip latency in milliseconds, and packet loss and choke percentages, turning vague complaints into precise data points that both players and engineers can act upon. Which allows its players to monitor connection spikes.

By translating subjective experiences into quantifiable metrics, the development can also use this information to zero in on issues like bandwidth bottlenecks or jitter before they derail a match. Think of it as both immediate player feedback, and when looked at a macro level for the development team, insights to take action on in terms of performance. It has the added benefit of removing the guesswork, as user can report “15% packet loss and 200 ms latency,” which dramatically speeds up troubleshooting and patch deployment.

Note this is also a strategy adopted by Edgegap clients, namely Rivals of Aether 2. Which in addition to highlighting the latency during a match and potential latency prior to a match, allows players to favorite preferred server locations alongside reporting a server during a match to help investigate issues.  

Taking Action – Building Monitoring helps with Insights-Generation

Behind the scenes, APEX Legends’ data science crunches a week’s worth of matches. Every match’s performance data—from individual packet loss rates to server CPU frame times—is aggregated and analyzed to spot patterns that human teams might miss, such as a malfunctioning network switch in a particular datacenter. This approach scales. It lets the team detect systemic issues before they become player‑wide outages and collaborate with hardware partners to replace faulty machines. It’s proactive. And by sharing anonymized, opt‑in metrics, players help drive these insights, knowing that a single bad server can be pinpointed and fixed across hundreds of matches.

A real‑time monitoring pipeline triggers notifications the moment metrics breach thresholds, complete with a unique server identifier that lets engineers find and diagnose the exact machine without correlating timestamps. Which means no more manual log dives.

With faster fixes, slower servers and network hiccups should be relegated to edge cases through the simple process of iteration and improvement, making matches increasingly crisp and responsive over time.

Slow Motion Servers

Slow‑motion servers still (and will) continue happen.

When a server’s CPU can’t complete its 20 Hz simulation cycle—50 ms per tick—it starts falling behind, causing the entire match to stutter as the system waits for physics, animations, and state replication to catch up. That lag is painful.

Respawn actively removes underperforming hardware—machines suffering from overheating, underclocked processors, or faulty components—to maintain a fleet that consistently meets performance targets. Future updates aim to reduce these events. Real‑time alerts and predictive analytics will catch dips in performance before they impact gameplay, so slow‑mo matches become rare outliers instead of weekly frustrations.

WiFi – Latency’s Greatest Enemy

Jitter and latency spikes can be elusive. While Wi‑Fi users may see unpredictable ping swings, even wired players can suffer when server processes introduce delays, and until now engineers had to recreate conditions on test rigs to identify bottlenecks.

That changes with RPROF profiling. By capturing a detailed log of every server task—ballistic calculations, network I/O, physics simulations—RPROF files show exactly which operations are hogging CPU cycles so developers can optimize new features before they degrade performance. It’s like an X‑ray. And it helps the team trim latency from every corner of the code, ensuring that new legends, weapons, and abilities arrive without unexpected slowdowns.

Still, not all games have the opportunity to overcome WiFi as hardware like VR are designed to be used over WiFi. If you’re a VR developer, make sure to read Edgegap’s guide on how to how to minimize latency in multiplayer VR games and XR projects.

Packet Loss

Packet loss isn’t always the developer’s fault.

Internet traffic traverses multiple ISPs and peering points, and when packets vanish—whether at your home network, between backbone providers, or within a datacenter link—it disrupts the flow of inputs and updates that keep everyone in sync.

Tracing the problem takes a ton of work, as highlighted by the Respawn Team. The team collects network traces from players and probes from the datacenter, then negotiates with third‑party providers to repair weak peering links or congested routers.

Fortunately, there’s game server hosting orchestration services like Edgegap which handles the “plumbing” for you. By orchestrating the game server hosting of multiplayer games to all, on demand, of its 615+ locations worldwide (the world’s largest, and first region less network) across 17+ providers. Deliver 58%+ latency reduction on average, and increase fairness by 28%.

While Respawn can afford the DevOps, engineers and sales managers to negotiate with providers thanks to its ties with Electronic Arts, now all studio can.

Ping

Reconciling different pings is an art. APEX Legends’ lag compensation system rewinds the world state for each shot, merging what high‑ping and low‑ping players saw at the moment of firing to determine hits in a way that spreads “nonsense” equally rather than favoring low‑latency users. No solution is perfect. Players with 300 ms ping may get shot behind cover, just as low‑ping players sometimes feel unfairly vulnerable when lag skews their view, but the system ensures these oddities balance out over time. It’s a deliberate choice. And it means rural or international players aren’t left behind by a netcode that only prioritizes the fastest connections.

Rollback

Shots still sometimes don’t register. Whether due to prediction errors, physics‑engine quirks, or transient network desync, you might see blood and sound effects locally without any damage showing up on the server. That stings.

Respawn is building automated detection tools that flag discrepancies in hitboxes and trajectories, while also relying on player‑submitted clips to uncover bugs that only appear in live matches. Feedback loops matter. When you report a no‑reg with your performance metrics attached, it accelerates the process of isolating and fixing the root cause for everyone.

If you are considering adding rollback, make sure to read more on Rollback in our dedicated article.

Tickrate & Optimization

Tickrate is a balancing act as while increasing 20 Hz to 60 Hz could reduce input latency from roughly five frames to three, it would also triple bandwidth use (i.e., Network Egress) —from around 60 kB/s to 180 kB/s—and demand far more CPU resources to simulate and replicate full world states.

That extra cost isn’t trivial.

APEX Legends’ snapshot‑based model saves complete game states each tick, supporting global abilities and rich interactions that would suffer if bandwidth were constrained by higher tickrates. The marginal gains feel small. And the team argues those two frames of improvement don’t outweigh the benefits of full‑fidelity simulations and lower network requirements for a diverse, global player base.

---

This article is based on and cites the original article by Sammy Duc published on EA’s official blog. All rights in the original content are owned by their respective owners.