Unreal Engine 5, [UE5 Developer Availability 2022-04-05]
- Thread starter mpg1
- Start date
cheapchips
Veteran
Quite surprised that when the credits rolled on Jusant it was a relatively big team. Not because of quality, more just that they found budget for a title like this. About 80-100, depending on who you count.
I think the next game from DontNod is UE4. Presumably that's reflective of Banisher's development time as a bigger title. They have their own technology team, who can carry forward lessons from Jusant into their first big UE5 title.
I think the next game from DontNod is UE4. Presumably that's reflective of Banisher's development time as a bigger title. They have their own technology team, who can carry forward lessons from Jusant into their first big UE5 title.
Last edited:
DavidGraham
Veteran
CDPR is strongly hinting that their upcoming game will ship with Path Tracing using UE5.
wccftech.com
Cyberpunk 2077 Dev Talks About Upcoming Path Tracing Improvements, FSR3 Support, and CDPR's Switch to UE5
We've interviewed CDPR Global Art Director Jakub Knapik to discuss some upcoming Cyberpunk 2077 improvements and the switch to UE5.
wccftech.com
Frenetic Pony
Veteran
CDPR is strongly hinting that their upcoming game will ship with Path Tracing using UE5.
Cyberpunk 2077 Dev Talks About Upcoming Path Tracing Improvements, FSR3 Support, and CDPR's Switch to UE5
We've interviewed CDPR Global Art Director Jakub Knapik to discuss some upcoming Cyberpunk 2077 improvements and the switch to UE5.
A bit silly, by the time this game ships UE5 will essentially have the same features as Nvidia's pathtracing. Restir Direct lighting is already in experimentation, Lumen already handles all the GI/reflections pretty similarly and will end up with more bounces by 2028 or whenever this game ships. So, bit of a "whatever", thinking about what you're shipping 4-5 years from now definitely doesn't have "Public relations partner" as a priority.
They aren't implementing RESTIR. They're experimenting with "stochastic shadows" which is a subset of RESTIR that only covers shadowing and Epic Games still expects developers to use VSM with directional lights even if it does come to fruition ...
Last edited:
DavidGraham
Veteran
A very good demonstration for the upcoming inZOI game running at medium graphics using UE5.
cheapchips
Veteran
David Adams and Ben Gabbard of Gunfire were on Game Maker's Notebook last week (excellent podcast). If it wasn't obvious given it's limited UE5 featureset, it was a late in the day switch to UE5. Not much UE5 detail beyond that. Good listen for the usual chat on developers' careers and games though.
The new season of Fortnite is coming maybe this weekend? They have a big event on the 2nd which is supposed to launch Chapter 5. It's an in-game event with Eminem ... not sure if it'll be one of those in-game performances like they've done in the past. Anyway, Fortnite OG was version as UE 5.4 and Chapter 5 will have a brand new map so it'll be a good chance to evaluate performance. I'll be playing it for sure, and I'll probably take a look at performance on my system by slowly turning up the nanite, lumen, vsm stuff to see how much they cost relative to baseline performance with them off.
DavidGraham
Veteran
The current state of UE5 games on consoles.
atpkinesin
Newcomer
With all the talk about UE5 performance I got curious so I decided to benchmark Fortnite on an old AMD FX-8320 system I still had laying around. Apparently I had been running it at 4.0 GHz with 2x4GB@2133 as that's what it successfully booted with from a sata ssd. I put in my 1070 with an old 1680x1050 LCD and played a few matches to get "playable" settings that worked out to be TSR low w/100% render scale, no nanite, medium shadows, ambient occlusion, SSR, medium view distance/textures/effects, and low post-processing. I then benchmarked a portion of a replay (a 90 second third-party combat encounter at rifle range) with various UE5 features on or off to get an impression of their performance impact. The data are below and qualitatively the "playable" settings were genuinely so on the 60Hz monitor, but came without any of the headline features.
With the above settings the system was CPU bound and GPU utilization averaged mid-80s with no peaks above 90. With nanite enabled and TSR at 100% the GPU utilization was high indicating a potential gpu bound (impressive to manage that with the 8320, given that the 1070 is better than a series S GPU but the 8320 is so much worse), so I also tested the "recommended" resolution scale setting of 77% (equates to roughly half the total render pixels of 1080p or a bit more than 720p). That did increase the fps as seen in the chart while gpu load averaged 66 % with a max of 90 %, so a likely cpu bound and decreasing render res to 50 % did not see an increase in fps. Given that there was some gpu headroom at the lower render resolution, i also tried enabling all the ue5 headline features by setting vsm to high, gi to lumen high, and reflections to lumen high. At 77% tsr, gpu load increased to an average of 76 % with peaks at 97 % (reflex was on), but decreasing render res to 50 % saw no increase in fps and 100 % render res saw a decrease. The chart below shows the naive frametimes throughout the capture in the top panel while the bottom panel show the "GPU-Busy" frametimes (selectable in capframex).
I personally find the nanite on results very surprising in a positive sense and am more optimistic about future UE5 performance as work on the engine continues and devs develop better understanding/practices. With all the UE5 features active Epic devs have basically managed to eek out the performance needed for a console 30 FPS-lock on the meme cpu so i would have felt fine using a lower case e at the start of this sentence.
During actual gameplay there was quite a lot of stutter during the first few matches (and after turning on nanite, so some shader comp?) but it pretty quickly subsided (after 3ish matches) to being "fine for a free game" with stutter mostly restricted to the jump out of the bus and skydive with about ~1-2 "annoying to me, pascal hold-out" stutters per match while on the ground.
With the above settings the system was CPU bound and GPU utilization averaged mid-80s with no peaks above 90. With nanite enabled and TSR at 100% the GPU utilization was high indicating a potential gpu bound (impressive to manage that with the 8320, given that the 1070 is better than a series S GPU but the 8320 is so much worse), so I also tested the "recommended" resolution scale setting of 77% (equates to roughly half the total render pixels of 1080p or a bit more than 720p). That did increase the fps as seen in the chart while gpu load averaged 66 % with a max of 90 %, so a likely cpu bound and decreasing render res to 50 % did not see an increase in fps. Given that there was some gpu headroom at the lower render resolution, i also tried enabling all the ue5 headline features by setting vsm to high, gi to lumen high, and reflections to lumen high. At 77% tsr, gpu load increased to an average of 76 % with peaks at 97 % (reflex was on), but decreasing render res to 50 % saw no increase in fps and 100 % render res saw a decrease. The chart below shows the naive frametimes throughout the capture in the top panel while the bottom panel show the "GPU-Busy" frametimes (selectable in capframex).
I personally find the nanite on results very surprising in a positive sense and am more optimistic about future UE5 performance as work on the engine continues and devs develop better understanding/practices. With all the UE5 features active Epic devs have basically managed to eek out the performance needed for a console 30 FPS-lock on the meme cpu so i would have felt fine using a lower case e at the start of this sentence.
During actual gameplay there was quite a lot of stutter during the first few matches (and after turning on nanite, so some shader comp?) but it pretty quickly subsided (after 3ish matches) to being "fine for a free game" with stutter mostly restricted to the jump out of the bus and skydive with about ~1-2 "annoying to me, pascal hold-out" stutters per match while on the ground.
I'd understand that for PC gamers like the DF crew FSR looks OK on console but only because consoles is not their main platform and they can play those games with much better quality with DLSS.
But for me TUAA, even if with a bit lower final resolution, was far much better than FSR and possibly TSR. I am currently playing The Talos Principle 2 and FSR2 is overall a terrible solution (I actually prefer the overall look of TP1 using in-house AA). Looks fine in the best case scenario: lots of nanite rocks / walls without reflections, transparencies or vegetation. But most of the time the game has plenty of vegetations, transparencies, reflections and those FSR artefacts completely kill the overall quality of the image, making the game... ugly, there is no other word.
Devs should not use FSR and should instead use TUAA (used in Kena). It looks a bit lower resolution, yes, but the IQ consistency in most cases and lack of horrendous artefacts make it look way more pleasing for the eyes.
Isn't Fortnite currently still on the legacy map with far reduced geometry?
Fidelity Super Resolution looks great in Starfield on Series X. The proof is that if you have the native 1440p and the code is good, then it is perfectly sharp 4K
and provides an anti-aliasing image. There may be some shimmering in certain places, but it is hardly noticeable unless someone is looking directly at it...
So the properly used FSR is very good, of course you need a stable render resolution.
Has way less to do with FSR being "properly used" than with the content a game has.
No it doesn't as evidenced by the multiple videos Alex has made. FSR 2 completely falls apart during any and all motion.
atpkinesin
Newcomer
yeah, I should have said that scott_arm's post above pushed me to do the benchmark now so i can compare it to the new map when it comes out. i think its interesting as a lower bound for ue5/dx12 cpu demands as it is, though.
I'm not looking for errors in videos, but I'm playing the game on a 65" TV, and I'm completely blown away by the picture quality. Many people I know who also play the game have never mentioned "falling apart", they just say the image quality on Xbox is nice.
It's one thing to critically see what you can see if you focus on it, but that has little to do with the judgment of normal gamers who play the game and enjoy the graphics.
I hope many games use FSR this well in the future.
Similar threads
- Locked
