Tim Sweeney blames UE5 performance woes on dev workflow — here’s what that really means for players

Why this caught my attention

As someone who’s spent the last couple of years benchmarking Unreal Engine 5 games across a mid-tier PC and current consoles, Tim Sweeney’s new line – that UE5’s performance problems come from studios “building for top-tier hardware first” and saving low-spec testing for last – rings both true and conveniently self-serving. He reportedly made the point after Unreal Fest in South Korea, adding that the “main cause is the order of development.” Metal Gear Solid Delta’s rocky performance puts the topic back under the spotlight, but this isn’t a one-off. We’ve seen the same pattern with Remnant II, Lords of the Fallen (2023), Immortals of Aveum, Ark: Survival Ascended, Black Myth: Wukong, and more.

Key takeaways for gamers

• UE5’s defaults (Lumen, Nanite, Virtual Shadow Maps) assume headroom most rigs don’t have – scaling them down early matters.
• Studios often prove their vision on high-end PCs and next‑gen consoles first, then scramble to optimize late.
• Epic says it’ll ship automated optimization tools and expand training so “optimize early” becomes standard practice.
• Expect a divide: projects built on UE4 and ported forward tend to be steadier; first-time UE5 titles swing wider in performance.

Breaking down the claim: workflow vs. engine defaults

Sweeney isn’t wrong about the pipeline. Publishers want flashy vertical slices early, which pushes teams to lock in cinematic lighting and ultra geometry on dev kits and 4090-class PCs. The boring work — culling, LOD sanity, streaming budgets, PSO (shader) precompilation, CPU thread contention — gets triaged until it’s suddenly a certification risk. By then, ripping out expensive Lumen reflections or re-authoring texture packs is painful.

But the engine isn’t blameless. UE5’s “wow factor” features are heavy by default. Lumen can hammer both GPU and CPU. Nanite geometry is incredible, but careless use can crater 1% lows. Virtual Shadow Maps gobble memory and bandwidth. Yes, Unreal has Scalability Groups and Device Profiles, but the out-of-the-box path nudges teams toward the shiny preset rather than a conservative baseline. If Epic wants better outcomes, the templates, warnings, and profiling nudges need to push teams toward performance budgets from day one.

The evidence: UE5 hits and misses

Look at the recent slate. Remnant II shipped with aggressive upscaling expectations and noticeable CPU bottlenecks; still a fantastic game, but clearly designed around top-end targets. Lords of the Fallen delivered striking visuals with Lumen and Nanite yet struggled with traversal stutter at launch. Immortals of Aveum’s GPU demands shocked even high-end rigs. Black Myth: Wukong wowed visually but needed swift patches to stabilize frame pacing and VRAM usage across mid-tier cards. On the flip side, Tekken 8 (also UE5) demonstrated how tight scope and disciplined budgets can hit 60 FPS reliably. And Fortnite — because Epic owns the pipeline end to end — is the poster child for UE5 scalability, though it leans hard on TSR/DLSS/FSR to make it sing.

Notice the pattern: projects that migrated carefully or have narrow simulation scopes fare better; sprawling, first-wave UE5 games that chase visual showpieces across open worlds often pay the price at launch.

What Epic says is coming — and what it needs to be

Sweeney teases “automated optimization across devices” and “expanded developer education” so early optimization becomes standard. If that’s more than a slogan, here’s what would help players the most:

• Hard performance budgets in templates: project wizards that set platform-specific limits for draw calls, GPU memory, and CPU threads, with red flags when teams exceed them.
• Auto PSO (shader) baking and background compilation: fewer first-boot hitches and mid-play stutters, especially on DX12.
• Smarter default scalability: turn off the heaviest Lumen modes, cap Virtual Shadow Map resolution, and use sane Nanite fallback on low/medium by default.
• Visualizers that guilt-trip: real-time HUDs that show “You’re 40% over CPU budget on base consoles” during daily playtests.
• Education that’s practical: sample levels demonstrating shipping-quality settings for a 3060/4060/6700 XT class GPU, not just boutique 4090 captures.

If Epic lands even half of that, we’ll feel it in better 1% lows and fewer “just turn on DLSS Quality and pray” threads.

What players should watch for

Ignore raw “Ultra” presets. On UE5, medium/high with TSR/DLSS Quality often looks identical in motion and dramatically improves lows. Toggle Lumen’s global illumination and reflections separately; GI is usually the bigger hog. Prefer shadow settings that avoid Virtual Shadow Maps if your VRAM is tight. And pay attention to whether a game precompiles shaders on first launch — it’s the easiest tell of whether a team respected your time.

For console players, a 30 FPS quality mode isn’t inherently a fail if frame pacing is rock solid. The real red flag is inconsistent frame times and hitching when streaming — classic “optimization came late” symptoms.

Why this matters now

We’re heading into a dense 2025-2026 lineup of UE5 titles, from remakes to massive open worlds. The average PC on Steam hovers around 1080p/1440p with GPUs like the RTX 3060/4060 and RX 6700 XT, not 4090s. If studios adopt the “optimize early” discipline that Sweeney is preaching — and Epic backs it with tooling that enforces budgets instead of showcasing tech demos — UE5 can be the flexible, scalable engine it promised to be. If not, we’ll keep reliving the same launch-week patch cycles.

TL;DR

Sweeney’s right that building for high-end first and optimizing late hurts UE5 games — but UE5’s heavy defaults tempt teams down that road. Epic’s promised automation and training could help, if it bakes performance budgets into the engine’s starting point. Until then, set expectations: scale smart, watch for shader precompiles, and let real benchmarks guide your settings.