Of course I don’t expect HDRP to just by default be perfect and render everything faster, it’s not magic. I do get that with builtin things would not be easy and performance would need a lot of work. With the current HDRP and what BOTD showed us, we naively thought that we could go for 60fps and HDRP features on PS4.
HDRP 5.x seems to have Switch in mind (from what I read in the changelogs).
Just a small correction to what you are thinking, yes we may have wrongly configured HDRP hence the bad framerate but I suspect that even with the perfect configuration and usage, it would not give us enough time for everything else (gameplay, UI, network, etc.). Again, it’s only temporary and it will, at some point, work for this kind of use case.
Yes, I did not said it wasn’t, like I said, we just thought that it could be used for a 60fps PS4 game now since it’s been a few months since BOTD.
I did not post in that forum yet, I still want to run more tests to maybe pinpoint the reason why it’s so slow. Just like you said, I don’t want to join the ignorant group by saying that I can’t prove with facts. I should find time to do these tests in January. My dream project would be a simple scene with some basic values for lighting and everything, latest HDRP in a release build PS4 which would output useful information about performances. That may help the Unity team to improve the HDRP package and / or help me better configure future projects.
Do not forget that the scriptable render pipeline only works with HDRP/LWRP, with it alot of specialized optimization can be done that can not be done under the classic pipeline
All rendering passes are turned off on hd pipeline
GTX 550 ti
Unity 2018.3.6f1
PC
Ultra preset on Built In
Deferred rendering on Both
SRP Batcher is On …
Hi, I’m new to Unity, I saw this topic and I wanted to ask this question at a distance, I tested the default example both in the 3D template with Extras, and with the HDRP template at 1920x1080. I made a build of both using FPS_Counter, and I also noticed a difference of about 100 FPS more in the classic pipeline. Then did anyone understand if this gap could be reduced or are the advanced features of HDRP to reduce performance? Thank you. Attached the screenshots
This is because HDRP is an AAA renderer - it does all the heavy lifting up front so it can maintain that framerate long after builtin has died… does that make sense? it does a lot of caching and work up front in the frame.
HDRP does a hell of a lot more work so you can throw the kitchen sink at it and expect stable fps.
First of all, thanks for the answer, the doubt came to me because even with the same conditions, the classic rendering is more than twice as fast. so you can deduce that if you are satisfied with the graphics in general of the classic pipeline go to that one, otherwise if you want to do something even more realistic go to the HDRP. The fact in my opinion is that there is too much performance gap between the two in the face of a significant graphic difference, but maybe there will be further optimizations, is that correct?
Deckard is a temporal & spacial super sampler designed to get similar AA, DoF and motion blur as offline renderers (aka non-real-time renderer) … because it is in effect an offline renderer. It can take multiple seconds per frame if you want it to. Also, you can use it with HDRP.
HDRP is designed by people who did Frostbite and Destiny’s rendering. It follows a similar pattern: you get all the AAA features but you have to stop thinking like a hobbyist and start thinking like AAA developers do by viewing the problem mostly as “how do I reduce bandwidth”? because bandwidth is the main limiting factor.
With builtin, for a simple scene, the bandwidth requirements for builtin are very low by default. But when you add more lights and geo and post effects to builtin, it will get stomped by HDRP’s performance because the heavy work HDRP does up front is designed to make bandwidth cost much less once it is under a real load in a real project.
Here’s some made up numbers to give you an idea if the scaling, I’ll use fps as a guide so it’s relate-able (we really should use millisecs):
Prototype boxes
HDRP 200fps
Built-in 600fps
2.5D platformer
HDRP 200fps
Builtin 400fps
Next gen FPS with all post effects with VFX compute and shadergraph:
HDRP 100fps
Builtin 20fps (or not able to work at all due to pipeline limitations)
See? once you cross a certain threshold, HDRP keeps on going - for real world jobs, not little experiments. Also if you know how millisecs and fps really work, you’ll see FPS as a guide is misleading you badly. It’s not linear!
So basically if something is running at 200fps or 300fps, it’s hyper meaningless. There’s not actually a big performance difference. But the number looks big. When you look at the reality with millisecs though, there’s not a big difference.
Also HDRP allows for using raytracing with one click in 2019.3 so there’s that.
For me, right now, HDRP is easier to use, has more functionality, looks way nicer, the lighting is more realistic, and experimenting with basic shaders is much more streamlined (all in the box). I’ll take it any day over Standard at this point, even if there is a performance hit.
If you need performance, use what is good for you.
