Performance woes in Unreal Engine 5 games are developers' fault, says Tim Sweeney

[Starfals]

Most Unreal Games do lag and stutter, but not all. Maybe hes right.. who knows. Some are smooth tho, thats what is puzzling. Is it the devs or the engine.. or both?

 

[godrilla]

Unreal did prove capable. You have tech demo running on PS5 looking best it can be.

You also have Fortnite showcasing in real time what UE is capable of.

Should a game engine be judged on outliered demos/handful of games or the summation of all the games that have a significant rate of dissatisfaction/rate of failure?
Also the revenue from Fornite alone could fund a better tooled engine for optimization especially in the era of ai hype. Not sure what the greed and scapegoating is about. This will only hurt their image further. Trolls defending this behavior might have a conflict of interest.

 

[dangh]

(there is a reason why most people play it in potato mode).

Yeah... Because they want 600fps, and grass or leaves hinders their aim. Those guys would love single colour flat textures, zero effects, just worst setting ever to get every little advantage. And it is not fortnite unique. War thunder "pro"streamers are playing potato as well. And most other competitive games where grass exists.

 

[Mr Majestyk]

Classic deflection is the the mainstay of liars.

 

[hwertz]

Yeah this is gaslighting, pure and simple. It might be true that a lot of devs are not doing things as well as they should be. But the fact of the matter, some devs that started out developing their game as a UE4 game then switched it over to UE5, performance tanked on those as well (whether these games were well-optimized or not, that made it pretty clear that it was UE5 slowing things down.)

That said, UE5.6 is supposed to have a 30-35% speedup compared to earlier UE5 releases, they quite simply found a bunch of stuff in UE5 that could use some optimizations and did so. I guess as some games ship using UE5.6 (it came out in June 2025) we'll see how much real-world speedup it has.

Luckily, I'm just not that picky, my monitor does 60hz (it can do 72hz) so I'm never expecting over 60FPS; I'm fine with 30FPS and could get by on 20FPS for non-shooters if I had to. My Ubuntu system with GTX1650, I so far (even with UE5 games) have been able to hit my FPS targets.

 

[Porkous]

Unreal proved absolutely nothing with that. It fixes none of the issues mentioned and Fortnite is definitely not a good example (there is a reason why most people play it in potato mode).

Why would the engine fix your issues with your game project ?

At about 10:22 they stated that the way these optimizations work, the game has to be built from ground up on with these features in mind. "You can't get away with too many simplifications and compromises".
Most titles have started implementing these features as of late, projects where already going, early, mostly, we will start to see things moving with PS6. I think we seen in the past glimpses of the future, but mostly were sponsored stuff, that at first ran heavy on the engine, with highest gfx card, but much later, these features were normal as hardware progressed for mainstream.

"Comparison: Normally, even a handful of dynamic lights causes huge performance drops in games. MegaLights changes this, letting artists fill whole cities or interiors with light sources freely."

If the dev gets something wrong, the opposite happens.

 

[Puiu]

Why would the engine fix your issues with your game project ?

At about 10:22 they stated that the way these optimizations work, the game has to be built from ground up on with these features in mind. "You can't get away with too many simplifications and compromises".
Most titles have started implementing these features when when projects where already going, early, mostly, we will start to see things moving with PS6.


"Comparison: Normally, even a handful of dynamic lights causes huge performance drops in games. MegaLights changes this, letting artists fill whole cities or interiors with light sources freely."

If the dev gets something wrong, the opposite happens.

It fixes ABSOLUTELY none of the issues mentioned above. Do megalights fix Lumen's requirement of mega blurry temporal AA?

 

[Puiu]

Yeah... Because they want 600fps, and grass or leaves hinders their aim. Those guys would love single colour flat textures, zero effects, just worst setting ever to get every little advantage. And it is not fortnite unique. War thunder "pro"streamers are playing potato as well. And most other competitive games where grass exists.

Then the many MANY videos of youtube posts and reddit posts complaining about performance issues must be just my imagination.

Fortnite has the same problems as other AAA games. It's just less obvious because few play at max settings and the game isn't exactly a graphics powerhouse.

 

[Loadedaxe]

"Many studios, he said, focus on building for top-tier hardware first, leaving optimization and low-spec testing until the final stages of production."
It has been done this way since forever and there is a good reason to do so. Devs want to show the game in all of its beauty which means it runs on the most powerful hardware if we are speaking about PC. They want to show the very best. Why would they show an optimized version that does not look as good as the games presented by other studios who go the same route that Tim described?
Development got more complex, also true. A studio, if it wants to present perfectly optimized game built on u5, has to hire people whose salaries will bankrupt the studio if the game does not sell well.
So what, are we supposed to go back to older engines or make games with graphics that look like it was made in 2015?
Last, it is so much easier to fix issues when you have even 5k free beta testers. It is so tempting that even the studios that have full in the house beta tester teams still use customers as beta testers.

I think the problem is less about aiming high and more about when optimization is treated as a priority. Yes, studios want to wow with visuals on top hardware, but when optimization only comes at the tail end, it often means anyone not on a 5090 ends up with a stuttery, broken mess at launch. And when you’re charging $70+ on day one, players expect more than “we’ll fix it later.”

Nobody’s asking studios to go back to 2015 graphics or abandon UE5, people just want better balance. Prioritizing scalability earlier in development doesn’t kill ambition ... it ensures the game actually runs on the wide variety of hardware that pays the bills.

As for using players as testers, it’s not that beta testing is bad, but it feels lazy when it replaces proper QA rather than supplementing it. Getting feedback from 5k users is great ... shipping a borderline unplayable game and calling it “early access” is not.

 

[arrowflash]

Since my previous comment was mostly off topic. About the actual article subject, I think it's a complicated matter with more than two sides, and all sides have their points.

Tim Sweeney is right, or at least makes very good points, but critics like the Threat Interactive guy are also right.

Another significant part of the issue is that video game developers used to be very passionate people who loved video games, and were very skilled programmers. Many video game "developers" today don't even like video games, have nothing but contempt to the gaming public, and aren't hired based on talent or skill. And this competency crisis - which today is also very strong in the indie scene - is really rearing it's head today.

 

[toooooot]

Since my previous comment was mostly off topic. About the actual article subject, I think it's a complicated matter with more than two sides, and all sides have their points.

Tim Sweeney is right, or at least makes very good points, but critics like the Threat Interactive guy are also right.

Another significant part of the issue is that video game developers used to be very passionate people who loved video games, and were very skilled programmers. Many video game "developers" today don't even like video games, have nothing but contempt to the gaming public, and aren't hired based on talent or skill. And this competency crisis - which today is also very strong in the indie scene - is really rearing it's head today.

The outsourcing at many once studios that at one point in time were huge is a big part of how the industry works. People who make their game often do not even know a lot about technical aspects because the job is done outside the studio. What do people who do that work instead care for? To finish it as quickly as possible. It is not their project, they are not passionate about it, they do not benefit more or less if it fails or is a hit.

 

[gamerk2]

Ther was an attempt to so that.
Its calles .. Direct X Raytracing (DXR).

Works in theory. But reality is different.

In June-July there were big waves about Radeon RX 8070XT strugling hard with some games on UE4. And Intel cards.
And, for better measures, some nVidia cards.
Reason? Games come out with UE library made not with DXR but NVRT code branch.

But that's what *always* happens: One of the GPU manufactures (usually NVIDIA) comes up with a new feature, and pays a ton of money to modify existing engines to make use of it. The other one (usually ATI/AMD) comes in later with a lesser featured alternative that runs anything built on the alternative code horribly. Eventually, Microsoft mainlines the feature in DirectX and over time everyone goes with the vender neutral alternative.

Heck, I'm old enough to remember when NVIDIA/ATI had competing Pixel Shaders, which didn't get unified until Shader Model 2.0 (I think in DX9?). This literally *always* happens.

 

[Melkor Unlimited]

One of the GPU manufactures comes up with a new feature, and pays a ton of money to modify existing engines to make use of it.

Except ... when one makes the use of technology so constraining that even its own product struggle at using it, just to do it harder for competing companies.

PS: https://wccftech.com/linux-still-isnt-ready-to-give-up-on-20-year-old-ati-radeon-gpus/
Show me one nVidia GPU supported after 20 years

 

[techstrike]

Day after day new devs are missing their touch with the real hardware.
they are depended by high programming languages and their libs and they're missing the real contact with the silicon.

 

[Stoly]

I don't really mind developers using top of the line hardware (sometimes unreleased)

Thing is that going down the ladder should only mean using scaling or lower res or lower textures or NO RT. Thing is that you have to do all of the above plus more to get the damn thing running reliably even on a RTX5070Ti or worse.

 

[gamerk2]

Except ... when one makes the use of technology so constraining that even its own product struggle at using it, just to do it harder for competing companies.

PS: https://wccftech.com/linux-still-isnt-ready-to-give-up-on-20-year-old-ati-radeon-gpus/
Show me one nVidia GPU supported after 20 years

Unofficial Linux developers hardly count as "supported".

Secondly, if talking Ray Tracing in particular, *everyone* knows it's expensive, that's why it's still used fairly sparingly and no one uses it as their primary light rendering source, since GPUs are still a good five to ten years from being powerful enough to brute force it.

 

[etoven]

Tim Sweeney is correct. I a in the UE Editor everyday. These hater claims that it doesn't run well is absolute nonsense. He should really stop pandering to those people. Per tick unreal is probably the most bang for your buck engine ever.

 

[Axeia]

Yeah this is gaslighting, pure and simple. It might be true that a lot of devs are not doing things as well as they should be. But the fact of the matter, some devs that started out developing their game as a UE4 game then switched it over to UE5, performance tanked on those as well (whether these games were well-optimized or not, that made it pretty clear that it was UE5 slowing things down.).

There might be different recommended approaches in UE5 to do certain things.

Usually many roads lead to Rome, and some might have run better in UE4 whilst a new UE5 road is far better but not utilised when you just port things over as quickly and easily as possible.

It's probably also part of the reason why there is so many poor performing UE5 games. I suspect people either copy pasting old code that's no longer the recommended way of doing things, or simply continuing to use the tools the same way as they have. P

Imo the fact that there is performant UE5 games is proof that the engine is fine. Now whether to blame Epic for bad documentation or developers being lazy* is a different but probably the real question.

* Don't think lazy is the right term. Probably a mix of people getting hired to do a job and simply not caring. Game programmers should be gamers themselves and often just aren't. And/or people getting pushed to deliver as fast possible.
If you get passionate people and give them a little bit more time the result is fine.

Imo it's up to gamers to make the difference, don't accept stuttery games. No demand, no supply.

 

[TimsLambo]

It is kind of tone deaf that he has the gall to tell developers that they need to be educated while unreal is a pain to even if you have a pc way above the recommended specs in the documentation, and depending on the project you might even struggle with incredibuild.

A lot of unreal "Features" are completely undercooked and anathema to performance. They are not only bad features. Some features literally obstruct the developers ability to think on terms of performance.

Education, for me, would actually mean doing your own toy engine to see how much unreal *actually* brings to the table. Those that have done the exercise will realize quickly that unreal does proportionally a lot of noise in loc for what it actually does.