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

[Alfonso Maruccia]

In a nutshell: Unreal Engine powers some of the world's biggest games, but its reputation has taken a hit as several high-profile releases stumble with performance issues. Epic CEO Tim Sweeney argues the problem isn't the technology itself – it's how developers are using it. His comments cut to the core of an ongoing debate about whether today's development struggles stem from engines, hardware, or studio practices – and how much responsibility Epic should bear in addressing them.

Recent high-profile Unreal Engine 5 releases have highlighted the challenges of game development with Epic's popular creation tool. Games such as Metal Gear Solid Delta or Wuchang: Fallen Feathers have suffered from significant performance issues, but Sweeney insists the blame lies with game studios rather than the engine itself.

Speaking at an Unreal Fest keynote in South Korea, Sweeney explained that the main cause of stuttering, slowdowns, and other graphics-related issues is the order of development. Many studios, he said, focus on building for top-tier hardware first, leaving optimization and low-spec testing until the final stages of production.

Ideally, optimization should begin early in development – even before designers complete the full game content. Sweeney acknowledged that the development process is far more complex now than it was ten years ago, which is why these issues cannot be solved simply by tweaking an engine or SDK.

"We're doing two things: strengthening engine support with more automated optimization across devices, and expanding developer education so 'optimize early' becomes standard practice. If needed, our engineers can step in," Sweeney said during the interview.

The Epic founder emphasized that engine makers and studios need closer collaboration if they want to deliver well-performing, broadly accessible games. Epic is also applying lessons learned from optimizing Fortnite for low-end hardware to improve the Unreal Engine SDK, helping more UE titles run better on lower-spec PCs.

The latest versions of Unreal Engine are indeed focused on improving performance, particularly in CPU-limited scenarios. Epic is offloading more rendering tasks to the GPU, reducing CPU bottlenecks and allowing developers to fully leverage features like Lumen global illumination and Nanite geometry.

Sweeney's vision for an ideal development workflow sounds promising on paper, but it could present challenges for indie developers.

Smaller projects often lack the optimization resources we expect from AAA studios. It remains to be seen how Epic will support smaller teams in adopting these practices and make UE development more efficient for everyone.

Permalink to story:

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

 

[dim0n]

Yep, that why all gamers are alpha (not a beta) testers today, playing "early access" (what a crap name) games... In my times, when games was released on physical medias, there wasn't anything like you, younger players know, in our times, games was properly tested and released as finished products, patches was mainly to maintain, balance difficulty... Maybe adding more rendering options like 3Dfx, OpenGL patches, nobody downloaded patches about 100 GB size (almost 100% size of game), while today, you redownload your game quadrillion times... Yep I'm pointing on you COD! If we had patch it was from 1 - 20MB big, while game had 500MB... And nobody tried to say, you need to destroy your physical media, not I point on Ubisoft...

 

[FaTaL]

The piss poor performance on UE5 across many games is attributed to 3 things:
1. Epics documentation on UE is poorly written.
2. Developers working on UE5 lack UE experience and are newbs so learning on the job. gg.
3. Epic sucks at prioritizing rendering, which they admit.

ultimately 2 out of 3, the finger points to Epic

 

[Eldrach]

Yep, that why all gamers are alpha (not a beta) testers today, playing "early access" (what a crap name) games... In my times, when games was released on physical medias, there wasn't anything like you, younger players know, in our times, games was properly tested and released as finished products, patches was mainly to maintain, balance difficulty... Maybe adding more rendering options like 3Dfx, OpenGL patches, nobody downloaded patches about 100 GB size (almost 100% size of game), while today, you redownload your game quadrillion times... Yep I'm pointing on you COD! If we had patch it was from 1 - 20MB big, while game had 500MB... And nobody tried to say, you need to destroy your physical media, not I point on Ubisoft...

I was around in the early days - and you're comparing games with the technical complexity of a milk cartoon compared to games today. Most people didn't use Gpus - so games were made for Cpus. And I remember the first "real" 3d games like Quake 1 - and trying to run RivaTNT or 3DFx on those was extremely touch and go - could work flawlessly for 2 hours, then it wouldn't even start.

Over all - it's much easier for users now and much less buggy than it used to be. But due to the fact that there are 10 000 different hardware combinations, optimization has become as pain in the behind.
I actually applaud the ones making games with the xx60 series in mind instead of trying to push the bar on the upper end for eye candy presentations. Gamers..just wants their games to run smoothly and for the game to be good. Even if I have a 5090, I dont really go around staring at the puddles in every game I play - technologies like Path Tracing and Neural rendering are cool and looks great..but do I REALLY care if a game has them if it looks good without it? Not even a little bit

 

[m4a4]

I mean, he is right. These AAA developers (though more specifically publishers) aren't giving themselves enough time to optimize their games for lower hardware. They just target high end (because it's easier) and then hope that they can make it work later.

Ultimately it's up to themselves to make it work, not the game engine (unless there's a specific bug). There are tricks and optimizations the engine can do, sure. But it was never meant to be used as a crutch (as they're finding out).

As for indies, they don't normally do dumb/wasteful things like focus on bleeding edge graphics (unless there's a good reason). And they usually have more flexibility to figure these things out as they go along.

 

[arrowflash]

Yep, that why all gamers are alpha (not a beta) testers today, playing "early access" (what a crap name) games... In my times, when games was released on physical medias, there wasn't anything like you, younger players know, in our times, games was properly tested and released as finished products, patches was mainly to maintain, balance difficulty... Maybe adding more rendering options like 3Dfx, OpenGL patches, nobody downloaded patches about 100 GB size (almost 100% size of game), while today, you redownload your game quadrillion times... Yep I'm pointing on you COD! If we had patch it was from 1 - 20MB big, while game had 500MB... And nobody tried to say, you need to destroy your physical media, not I point on Ubisoft...

I agree with most of you said, only thing I'd say you got wrong was the "balance difficulty" part.

If there's one thing video game developers back in the old days wouldn't touch or patch at all once a game got to market, was things like difficulty and gameplay balancing, except in some very rare extreme edge cases. And this was a great thing.

Nowadays I see many indie developers who can't leave their games alone even after they're out of early access, and every patch introduces "balancing changes" and "gameplay changes", often making the gameplay worse and less fun, not better. What I dread the most are "community feedback" changes, that often come from vocal minorities. Many devs especially indie devs, keep forgetting the famous words from Steve Jobs that customers often don't know what they want and you should be very careful if you choose to listen to them. I'm talking about offline single player titles here of course, not GaaS online competitive multiplayer titles where being in a state of constant change is more natural.

Once a game is on market and out of early access, balancing and gameplay should be immutable, except in extreme edge cases when there are serious issues and complaints by the majority of players (say, a boss or area that only 20% of players can beat, or a puzzle that only 20% of players can solve).

 

[opckieran]

Publishers are more interested in conforming to their financiers’ agendas than actually optimizing.

 

[toooooot]

"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.

 

[Theinsanegamer]

"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.

That is not how most games were built. Most games were built around the consoles first then ported to PC later, of if you go back further, were built for one or the other. Plenty of games on PC had free "4k texture pack" DLC to take advantage of more powerful hardware. (skyrim, warhammer, ece).

I was around in the early days - and you're comparing games with the technical complexity of a milk cartoon compared to games today. Most people didn't use Gpus - so games were made for Cpus. And I remember the first "real" 3d games like Quake 1 - and trying to run RivaTNT or 3DFx on those was extremely touch and go - could work flawlessly for 2 hours, then it wouldn't even start.

Over all - it's much easier for users now and much less buggy than it used to be. But due to the fact that there are 10 000 different hardware combinations, optimization has become as pain in the behind.
I actually applaud the ones making games with the xx60 series in mind instead of trying to push the bar on the upper end for eye candy presentations. Gamers..just wants their games to run smoothly and for the game to be good. Even if I have a 5090, I dont really go around staring at the puddles in every game I play - technologies like Path Tracing and Neural rendering are cool and looks great..but do I REALLY care if a game has them if it looks good without it? Not even a little bit

You talk of games from when CPUs were used for rendering yet bring up the "10,000 hardware configurations today" line? Did you forget what the gaming world was like in those days, before DirectX?

Let us also not forget that 10 years ago, PC games were optimized for just as many varied configurations without issue. UE3 games didnt run bad the majority of the time, and eventually the devs figured out UE4. Why is UE5 any different?

I think the loss of talented personnel the last decade is finally rearing its ugly head. Devs today are cookie cutter programmers that are thrown out when a project is done. They dont have the time, drive, or talent to really optimize games. These guys are not John Carmack, they're not the old DICE engine team, and so on.

 

[Subset6]

So by Tim’s logic the only talented developers that do focus on optimization are those that don’t use Unreal at all, like Decima.

 

[Axiarus]

I was around in the early days - and you're comparing games with the technical complexity of a milk cartoon compared to games today. Most people didn't use Gpus - so games were made for Cpus. And I remember the first "real" 3d games like Quake 1 - and trying to run RivaTNT or 3DFx on those was extremely touch and go - could work flawlessly for 2 hours, then it wouldn't even start.

Over all - it's much easier for users now and much less buggy than it used to be. But due to the fact that there are 10 000 different hardware combinations, optimization has become as pain in the behind.
I actually applaud the ones making games with the xx60 series in mind instead of trying to push the bar on the upper end for eye candy presentations. Gamers..just wants their games to run smoothly and for the game to be good. Even if I have a 5090, I dont really go around staring at the puddles in every game I play - technologies like Path Tracing and Neural rendering are cool and looks great..but do I REALLY care if a game has them if it looks good without it? Not even a little bit

Gaming was a much smaller industry back then. I am pretty sure a multi billion dollar industry can put a little money towards optimization. The problem is that its all about pushing the game out as fast as you can to spend as little on development as possible. Who needs optimization time when you can just slap DLSS/FSR on it and call it a smeary day.

 

[Watzupken]

When things don’t look good, it’s normal for people to point at others. In this case, yes, it’s possible for games to be better optimized when we have an example of Expedition 33 getting it right. But at the same time, performance is not great. It’s just ok. Visually, it’s an improvement, but the performance cost is too great.

 

[llamapanda]

Tim's right, and the advice of people like him and Carmack with genuine 'doing it at the sharp end down in the mud' old-school experience should trump the mewling of drag-and-drop 'devs' any day of the week.

The price we pay for having engines that anyone can use is that anyone can use them, including people with no real understanding of how things work. I work daily with devs who have an "It runs and doesn't crash, ship it" mentality in complete ignorance of the hardware screaming at 95% under minimal practical load. Education is the key, but so few seem willing to take it even when it's free.

Those devs most aggrieved by bad press about their game's performance should take Tim up on his offer to help. The collaboration would benefit everybody by not only making the games more accessible for a wider audience but would also help fix any documentation issues that are holding people back so would pay dividends forever.

 

[Puiu]

So much BS. This is how they made the engine and how they themselves recommend the "optimizations".

Their own recommendations are full of performance issues. This is Epic's problem, not just the dev's

 

[J95qRP7]

It might be the developers fault IF it was only a few games. Given that almost every UE5 game suffers from the same faults and issues, the "problem" must be more systemic, since it spans so many different development workshops.

 

[MSIGamer]

So by Tim’s logic the only talented developers that do focus on optimization are those that don’t use Unreal at all, like Decima.

Yep. Or they don't use the features of unreal that tank performance (nanite/lumen).
Take Split Fiction as an example. Magically if you don't use those features then you suddenly have a good running game.

You won't hear Sweeny talking about that though, since he's the CEO of the company that is trying to push this tech and then blame the developers on not having optimization in mind when it doesn't run well.

Meanwhile Epic's own matrix demo runs like **** too. Just ask a bunch of questions and see through the load of bull.
For example: where is the lumen/nanite enabled game that is optimized?
It doesn't exist, they all run like ****.

 

[Eldrach]

That is not how most games were built. Most games were built around the consoles first then ported to PC later, of if you go back further, were built for one or the other. Plenty of games on PC had free "4k texture pack" DLC to take advantage of more powerful hardware. (skyrim, warhammer, ece).
You talk of games from when CPUs were used for rendering yet bring up the "10,000 hardware configurations today" line? Did you forget what the gaming world was like in those days, before DirectX?

Let us also not forget that 10 years ago, PC games were optimized for just as many varied configurations without issue. UE3 games didnt run bad the majority of the time, and eventually the devs figured out UE4. Why is UE5 any different?

I think the loss of talented personnel the last decade is finally rearing its ugly head. Devs today are cookie cutter programmers that are thrown out when a project is done. They dont have the time, drive, or talent to really optimize games. These guys are not John Carmack, they're not the old DICE engine team, and so on.

I think you may be looking at the good old days with rose tinted glasses. During the era of UE3 games was for the most part - much shorter endevours and easier to optimize due to for the most part using the same type of technology across all vendors (MSAA). And while UE3 "pushed" the envelope of graphics..the complexity of antialiasing, shadows etc. was much simpler.
Our expectancy for "life like graphics" have pushed developers to focus more on making DLSS, XeSS, FSR, Raytracing, Pathtracing etc etc to look good and work properly than gameplay.

The fact that we dont have universal standards is hurting the industry badly. AMD made FSR open to "all vendors", but with FSR4 that is suddenly AMD only.
There should be a consensus to settle on 1 upscaler, 1 Raytracing teqnique etc. - I truly believe we'd see better results come of it

 

[godrilla]

When things don’t look good, it’s normal for people to point at others. In this case, yes, it’s possible for games to be better optimized when we have an example of Expedition 33 getting it right. But at the same time, performance is not great. It’s just ok. Visually, it’s an improvement, but the performance cost is too great.

The biggest critic of the engine doesn't believe Expedition 33 is as optimized as some are making it out to be and that Unreal engine 5.6 isn't the savor update Epic has us hoping for.

Tim Sweeney just admitted he loves the status quo and will do little for fork over resources to help all those failed projects and layed off developers that from lack of optimization and cluster fudge of an experience even on capable brute forced hardware.

The benefactors are Nvidia and AMD.
5090 and 9800x3d tweaked to play a palatable stutter experience anyone?

Like I said before at this point Epic would have to help those failed projects on optimizations via new tools, resources etc if it wants to build back bridges to Gamer's trust!

 

[Puiu]

That is not how most games were built. Most games were built around the consoles first then ported to PC later, of if you go back further, were built for one or the other. Plenty of games on PC had free "4k texture pack" DLC to take advantage of more powerful hardware. (skyrim, warhammer, ece).
You talk of games from when CPUs were used for rendering yet bring up the "10,000 hardware configurations today" line? Did you forget what the gaming world was like in those days, before DirectX?

Let us also not forget that 10 years ago, PC games were optimized for just as many varied configurations without issue. UE3 games didnt run bad the majority of the time, and eventually the devs figured out UE4. Why is UE5 any different?

I think the loss of talented personnel the last decade is finally rearing its ugly head. Devs today are cookie cutter programmers that are thrown out when a project is done. They dont have the time, drive, or talent to really optimize games. These guys are not John Carmack, they're not the old DICE engine team, and so on.

"Why is UE5 any different?" - because unlike previous versions UE5 focused on the wrong things and regular devs cannot fix it. It's broken at the fundamental level.

First off, the engine doesn't work without using the smeary and ultra blurry temporal AA. Lumen is broken beyond belief and requires massive amounts of blur to hide the artifacts. And Nanite is destroys edges (causes insane pixel crawls) which further exaggerates the need for blurry AA and is a performance hog.

This is a great video on how broken nanite is and how Epic is lying or pushing devs towards bad practices:

 

[Melkor Unlimited]

There should be a consensus to settle on 1 upscaler, 1 Raytracing teqnique etc. - I truly believe we'd see better results come of it

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.

 

[mrSister]

Of course it is devs fault. He is programming and selling the engine. He must know!

 

[Primey]

He's obviously not going to blame the engine.

 

[bviktor]

It must be extremely convenient in Sweeney's world where every issue is someone else's fault.

The guy's been doing literally nothing in the past decide except playing the blame game.

 

[Porkous]

The biggest critic of the engine doesn't believe Expedition 33 is as optimized as some are making it out to be and that Unreal engine 5.6 isn't the savor update Epic has us hoping for.

Tim Sweeney just admitted he loves the status quo and will do little for fork over resources to help all those failed projects and layed off developers that from lack of optimization and cluster fudge of an experience even on capable brute forced hardware.

The benefactors are Nvidia and AMD.
5090 and 9800x3d tweaked to play a palatable stutter experience anyone?

Like I said before at this point Epic would have to help those failed projects on optimizations via new tools, resources etc if it wants to build back bridges to Gamer's trust!

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.

 

[Puiu]

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.

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).