Unreal Engine 5, [UE5 Developer Availability 2022-04-05]

50fps at 4k on a flagship gpu released 3 years after the game's target console generation launched... Doesn't look too different from ue5 game perf (on 3 year past console launch cards) to me. On 2028 gpus I bet you'll find games ue5 games targeting the ps5 scream too.
The equivalent GPUs to the GTX 1080 are the 4070 and 7800XT. I’m willing to bet any amount of money they wont be pulling anywhere near 55 fps at 4k in a 2028 UE5 title by a large developer. In fact, I wont be surprised if they hit <60 fps at 1080p. Do you mind listing some current UE5 titles that hit just under 60 fps today on those GPUs?
 
Last edited:
If there are games using UE4/5 that run well and look good, then it shows the engine is capable, the rest is a matter of dev talent/time/cost etc...
The real problem since PS4/XB1 era, is that with the ability to patch games, a lot of publishers release a game in an unfinished/unoptimized state to make money as soon as they can, and then bother to let their devs optimize if the game sells well enough, if not, it's just forgotten.
 
I’ll drop it but if I may respond to the frostbite points. I can’t speak to Squadrons in VR, but the regular version screamed in terms of performance.
My work with Frostbite was specifically for Squadrons so I'm well aware of how it performs. Unlike the public narrative, I think it was indeed pretty good, but there's no disputing that it required a GPU that didn't exist at the time to run in VR at high settings. Only since the 4090 has that been possible. But of course at the time people were like "my 1080 can run HL:Alex great, Squadrons VR unoptimized, lazy devs", not dissimilar to other comparisons folks make now.

The UE games you mentioned were UE3 which always performed well.
Some of them but the Frostbite games you mentioned are mostly on older versions of that engine too (I did qualify "especially if you go back a bit to similar time periods as those Frostbite games"). In any case I wouldn't get too caught up in when marketing decides to call something a new major version... in many ways there's more differences between Unreal 4.0 and 4.27 than there is between 4.27 and 5.0 and similar goes for Frostbite. It's all continuous development after all.

Anyways let's just move on and talk about cool tech and games we are playing again.
 
Last edited:
Thought it was interesting that "Satisfactory switches to UE5" has hit the headlines a little today. Think it shows how news worthy UE5 is, good or bad. :)

Coffee Stain have had it in testing for a while. They're using Natite for cliffs. You can turn on Lumen, but they've no plans to relight the whole game to support it fully. The more useful thing for them is world partition. They were having to load in whole areas before, but that's not really a sexy UE5 story.

Their cheesy update video, with about 30 seconds of UE5 comment at the start.

 
BF2042 was the rare FB game were performance was all around dreadful.
And even that game doesn't run so bad if you stick to 64 player modes. We knew from games like BF1/BFV how much scaling player count affected performance in these Battlefield games, as 32-48 player modes on console could run at a pretty reasonable 60fps while 64 player modes would usually be in the 40-50fps range often enough.
 
And even that game doesn't run so bad if you stick to 64 player modes. We knew from games like BF1/BFV how much scaling player count affected performance in these Battlefield games, as 32-48 player modes on console could run at a pretty reasonable 60fps while 64 player modes would usually be in the 40-50fps range often enough.

BF2042 was absolutely horrendous at launch. Massively cpu limited. The improvements they've made are tremendous, but don't forget how it was when it first launched. I think BF1 is the only one that had a decent launch since BF4. BF4 was a disaster that became incredible after maybe a year of updates. Battlefield V had a rocky launch from memory, but I don't remember specifics.
 
Coffee Stain have had it in testing for a while. They're using Natite for cliffs. You can turn on Lumen, but they've no plans to relight the whole game to support it fully. The more useful thing for them is world partition. They were having to load in whole areas before, but that's not really a sexy UE5 story.
They have a bunch of neat custom tech too for stuff like rendering conveyor items and similar as well. The game runs really well (both before and after UE5 tbh) and is fun, they've done a really good job with everything so far. There's more details in some of their other community updates if folks are curious.
 
I’ll drop it but if I may respond to the frostbite points. I can’t speak to Squadrons in VR, but the regular version screamed in terms of performance. Other games would be BF3, BF4, BF One, BFV, both Battlefront games etc. DX12 was bad but you didn’t have to worry about using it. DX11 ran great and scaled to as many cores as needed. To this day performance scales great on new CPUs and GPUs in all of these titles. BF2042 was the rare FB game were performance was all around dreadful. The UE games you mentioned were UE3 which always performed well. People really only complained about the similar look a lot of titles had in the beginning.


Problematic for gameplay systems as the engine was not designed for RPGs.

I think you have selective memory on the Battlefield games. Pretty much every single Battlefield game released was a buggy mess at launch. Sure, over time those bugs were addressed and each game was generally quite performant and relatively bug free by the time the next BF game was released at which the cycle repeated.

Now some BF were far more buggy than others, but all of them were buggy to some extent at launch, even the original BF1942. BF2042 was just epically buggy. But if you look through various BF threads here you'll find people having bugs or performance problems with just about all the BF releases. Again, some better or worse than others.

Now some of them were also very nice visually, but BF games were almost never without their problems.

Regards,
SB
 
I'm still rather impressed that visas like this

Jusant 13_11_2023 19_37_24.png

have a level of detail that seamlessly goes down to this

Jusant 15_11_2023 21_21_41.png

The difference from non Nanite titles is that it's pervasive. Wherever you go there's just stuff scattered about, which other that 'made of chunky clay' stone, has really high poly counts. The artist have really leveraged Nanite to great effect.

I've probably failed miserably to convey that in a few screengrabs. I'd have climbed all the way from the bottom platform in the first shot.
 
InZOI's graphics aren't grimdark. It doesn't count. :)

Only been in development for a year? Impressive that something like that can be put together in that timeframe.
 
If that actually turns out to be a fully capable sim game and gets westernized, it would certainly kill EAs Sims business. Good riddance to Sims if it does, since EA have no interest in improving it ever each version and just using it as a revenue source.
 
EA's Sims business targets potato laptops. Will a UE5 game like this scale down that low?
Yeah, that's a good point. There will certainly be a barrier of entry many wouldn't be willing to invest in. If it gets released on console maybe.
 
The Sim's 5 (whatever it ends up being called) does have it's own GI solution though!

There's the barest glimpse of it at this time stamp. Don't watch the whole show though. These people are far too happy.

 
The Sim's 5 (whatever it ends up being called) does have it's own GI solution though!

There's the barest glimpse of it at this time stamp. Don't watch the whole show though. These people are far too happy.


It's hard to remember from 4, but The Sims 1-3 were technical marvels for the time, heck 3 has it's own cubemap capture system for each room for reflections, in what was essentially a fully dynamic open world game from 2009. Showing up Crysis and Rockstar at the same time is kind of amazing.

That 5 is going back putting in effort is a good, that they can probably make it pretty scalable because it only needs to be semi dynamic, needing to take into account construction of static objects but can just get away with scaleable cheap solutions for dynamic objects, means they'll be able to scale it down pretty well.

The Sims 5 is a good showcase for why not everything can just be put onto UE5 or Unity without very obvious sacrifices today. Photoreal characters are a straight detriment to the sims, as one example. A need to grow the game indefinitely with who knows what types of mechanical overhauls means having complete control and understanding of the engine from the ground up can mean easing a lot of future work/making future work a lot more open than relying on a codebase from others.
 
Last edited:
Back
Top