Marketplace environments and performance

I don’t know what the problem is, but after trying dozens of Unreal Marketplace environments found that none of them are production ready.
I’m working in an open world game so I use trees, grass and that kind of assets, and by just adding foliage to my level I go from steady 120 to 15 FPS in the worst cases, 40 at best, that’s without having anything else in the scene and in a machine capable to run RDR2 at max quality without any problems.

I would like to read you guys experience, but It seems to be something really wrong with the engine, with those numbers it’s imposible I can think to make it to production.

2 Likes

It’s not the engine. It’s the products.

Some… of them are pretty good, but most are really for production work only.

Something like RDR2, all the assets are custom made, right down the last poly, and there’s a lot of trickery going on. Including writing their own engine.

PS: When I say ‘production’, I mean for cinematics. Game is a whole other thing.

2 Likes

Clockwork (at least in my experience) is absolutely correct. Most of the environments I’ve used are made to look as pretty as possible without the thought or care of how they could be interacted with or explored in a gameplay scenario.

An example is one pack where there are HUNDREDS of lights, many of which do virtually nothing, that have HUGE attenuation distances. The optimization view of the lighting shows the entire scene to be bright-purple/white. It would run at around 50FPS, but when I delete and optimize the lighting, the FPS jumps up to around 120.

So, you’ll need to do a LOT of work with most of the assets.

So far, the most optimized assets I’ve come by are from FRESHCAN. They’re expensive, performance-wise, but VERY well optimized. I found that I could do very little to gather extra FPS/MS.

If you want tips on optimization, you can check out my list, here:

I’ve gotten very good at it.

Good luck.

1 Like

Clockwork is a community hero. Some of the anti-heroes or villains on here might just disagree with that, and say when it comes to performance gotchas, BOTH unreal engine and marketplace products ARE the problem. :stuck_out_tongue:

Suggestion: Spend some time doing a bit of wider reading, then decide. Follow posts by MostHost_LA… TheKJ… Roy_Wierer.Seda145… And others, to get a counter view. :wink: UE is great at sci-fi type scenes with high-mesh counts.

But landscapes and foliage? Its real easy to kill UE stone dead! :rofl: Out of the box, UE is a mixed bag. So don’t be surprised if scenes require a huge amount of tweaking, testing, profiling, optimisation [rinse-repeat]. That’s just how it is. The performance of your rig may be deceptive here, even meaningless. :wink:

It was @MostHost_LA, in fact, who pointed out to me that’s it’s possible to get large terrain with grass and trees running on a Galaxy Note. ( With UE4, to be fair ).

I was also helping someone with a project the other day, and they had grass all the way to the horizon. It wasn’t affecting the frame rate at all. It was stylized, but even so. It’s was just using the foliage tool and LODs, although it was an efficient mesh and LODs / imposter.

A lot of that could be the virtual shadow map invalidation to be fair - you could try going into your foliage mode and changing the “Shadow Invalidation Type” to “Static” to compare the difference in performance. You can swap that back and just change the “WPO Disable Distance” to something reasonable if you find that puts you up to over 100FPS

I do agree with you here, though. The engine is supposed to be general purpose, and so can’t really be totally optimized for any particular use.

It’s hardly surprising the RDR engine works better, but RDR is all it does.

2 Likes

Well, its not… it also does GTA5 and a few others.
RAGE is probably one of the best propriatery engines for the end product - I know nothing about working in it except the videos they release detaling the features it brings (such as predictive AI for npcs).
Regardless of that, the game-s shorts are better cinematic level than most things out there (even if the light/shadows is a bit worse than it could be).

The long and short is:
If you want a game to work with landacapes ue4.18 is the one to get before they messed the LOD optimization by changing it from distance based to size on screen.

In 4.25 you have a drop of 20% fps just for the pleasure of downloading the newer engind.

In 4.27 that drop is even harsher - but, because between 4.20 and 4.25 some id10t broke the options to precompute transparency in the basepass, you do get some of thaf performance drop back on foliage heavy scenes.

If you know nothing about anytning, then its probably impossible for you to even try to make a store bought scene perform better.
Conversly
If you know everything (which trust me, you never do), then it’s probably impossible for you to buy a scene and not return it - or to find any use for it.

That’s partially why I never released dynofoliage updates.
Made the environment and assets. It runs “great” at 60fps on a 1080ti. But it also looks bad compared to what I envisoned.
(Also the marketplace approval id10ts really ■■■■■■ me off by flagging as needing to fix the speed transitions on Dynocamera, when the whole point of the things is that you can tune the transitions to taste. EPIC is just bad. The employees are not like us, threy are like the leeches who ask for help and then when they get help cry and ask you to share your code so they do nothing).

Anyway, point being.

People releaseing these assets have to fight against knowledgless id10ts to get them released.
And many times have to put extra work in to bypass silly sh*t - almost worse than the Apple app aproval process (and there aint nothing worse than that!).

So, even brushify and people who get grants to release stuff probably won’t produce or deal with the hasssle of publishing to the marketplace more than once or twice.

I suppose we should start bugging Tim on twitter about how sh*t the engine and everything has become…
Kinda like how he bugs Apple about publishing (and gets pubslishing bans for tweets).

I started with UE when stable version was 4.14 I think, and I never had big performance problems then, even when I’m sure I made silly mistakes due the lack of knowledge.

I would give 4.18 a try assuming the repo is still there, the challenge would be to setup the required toolchain considering how old this version is.

i was perform lot of testing in last 5yrs since ue4.21 version. regarding marketplace assets (large environment), almost none is game ready (production game ready).for those who are always ready to argue, i note that with production game ready i mean 150+ fps on rtx3060 scalability 1. it is true that any single game have different and unique need so we cant expect from templates to bee ideal for ours. asside all ups or downs you can evaluate templates according to the complexity to adapt them to your needs. its not true that yu cant make open world games with solid performance (fps). but the sad truth is that in order to get those same performances to a satisfactory level, a lot has to be done. so first you have to know that whenever you want to get something you have to give something up.forget about the wonderful kinematics ready marketplace assets. in the best case, if you have a graphic engineer at your disposal who is able to directly adapt engine source, then of course you will be able to get the best speed/look ratio. as for the rest of us, you have two the most common options. 1st you must give up of all super duper advanced visual high fidelity …, go to extreme reduction in any possible way and you get your open world but keep in mind that the graphics will probably look like games from 10 or so years ago. 2nd you can try to keep some level of visual quality, but then you have to use other tricks, practically you have to somehow organize and intersect your spaces in owg, to limit the scenes to localized zones, and display all panoramic distances in hlod. anyway on begin first download and study epic games boy and his kite template.
keep in mind that even bahk is not optimized enough to be used directly for the game either. but its god example of owg level and yu can make additional optimisation and make them so. then you will probably get a clearer idea of ​​what you need to do and what it might look like in the end and with what performance.