My experience and thoughts on HDRP since its release.

Hi everyone, I just wanted to share my experience so far regarding HDRP since its release. I’ve been using HDRP as a base platform to build graphical assets for my next project that would be using HDRP in the future.

I’ve used HDRP for about 3 weeks and figured I’d try to share my thoughts on it. I hope this post can share a bit of insight to those who have yet tried HDRP or have just fired it up but not spending a lot of time with it. I also hope that the Unity HDRP devs can benefit from this in any way they can.

At the time of this writing, I am basing this post on Unity 2018.2.0b7 and 1.1.1.10 preview version of HDRP. I will be referring to the current Unity as “Legacy” and HD render pipeline, “HDRP”

Point 1) It is almost ready for production - sort of.
Despite what Unity Devs say, I think it is relatively ready for production. It has many things constantly changing and breaking, but in general, publically released versions available via the Package Manager is good enough. I can feel the rapid improvement of HDRP week after week, and so far I’ve found that it is actually on par with the Legacy Unity in terms of stability. The number of patches released after its launch is quite astonishing. Given that HDRP is handling the graphical side of things mostly, starting on a full-blown project seems like a viable option. It will be a bumpy road, but every project is, and I am willing to bet that HDRP will pay off in the long run given a 2-year project versus working on a next-gen graphics project on the Legacy with Asset Store shaders and FX slapped on.

Point 2) It is definitely a huge upgrade to graphics.
SSS works well and is quite performant. Post-processing seems better (I found TAA and motion blur showing better results) and the PBR seems way better, especially with low gloss materials. I also find it much easier to achieve photorealistic results compared to the Legacy Unity. It is difficult to point out why, but it feels right when I import it to HDRP and I often find myself adjusting the texture a few times compared to Legacy, where I’d adjust for days and still feel something is off.


See what I mean? The transparent backboard and the specularity on the rim always felt “off” on legacy, but with HDRP, I get results!

Point 3) Real-time shadows are not quite there.
I’ve found that the realtime shadows, in general, are not that different from the legacy Unity shadows. Personally, I appreciate the Contact Shadows, but the directional light shadows are just lacking. There are cascade settings that one can play around to control quality - especially near camera - but there is very little wiggle room in those settings that work without glitching shadows of other areas. I tried many different settings but always ended up going back to the default settings. I find the realtime directional light shadows quality to be very poor - not particularly bad compared to other engines, but given the premise that HDRP is aiming to be the HD platform, I find the lack of better shadows strange.

Point 4) Lightmapping is still generally broken and is in dire need of GPU mapping.
I’ve had issues with lightmapping from day one of Unity5. Truth be told, both Enlighten and Progressive Mapper is not production ready. It breaks, then glitches, but worst of all, it takes me 5 hours to bake a 100m x 100m scene, only to find out that there is something wrong with the configuration, or the lightmapper just acting out in general. Then another 5 hours and repeat until I flip my desk.

Obviously, lightmapping is not HDRP’s issue, but given that there is no other lightmapping/GI solution available on HDRP either, I thought I should still mention it. Whatever you do, currently on Unity, even on HDRP, you can’t achieve lightmapping at its highest potential. You just can’t for a production level scene. It is only viable for a small indoor sample scene. Anything beyond 10 x 10 is going to make your life miserable.

Unfortunately, Unity doesn’t seem to realize how bad the situation is. There was a presentation done about a year ago with the following title: Bake it til you make it.
Here is the video link

It does seem that lately, Unity is working on the GPU mapper
Here is the video link

While I am glad that they are working on this, it is almost 2 years late. It should have arrived with Unity 5 in the first place. CPU baking was never viable in the first place and HDRP a few years later is going to suffer. Unity said the GPU baker will be 10x faster, and it sounds great, but if you do the math, the 5 hour bake time will go down to 30 minutes. the 10x statistic could be higher or lower than 10 exactly, but 30 minutes is still a long time for a 100m x100m scene. I just hope that the GPU lightmapper coupled with PLM will be flexible enough for production.

Conclusion
Pros:

  • Best graphics hands up and will dominate the future.

  • HDRP team is rock solid and is working hard based on Github activity and release history.

  • HDRP is already amazing this early.

Cons:

  • Realtime shadows aren’t quite there yet.
  • Lightmapping deteriorates the whole experience in general and turns a great product into a mediocre one. Mostly renders the whole graphics department of the engine useless for large projects.

I hope my review/thoughts helped other users. I only wrote this lengthy post hoping that HDRP will not lose traction over time and carry on pedalling to the metal! HDRP guys are doing great.

12 Likes

“Point 1) It is almost ready for production - sort of.
Despite what Unity Devs say and other people moaning, I think it is relatively ready for production”
That is not the right message.
HD has been released as preview. Maybe it works for your requirements , but there is an entire world that is about custom shaders. There are a lot of challenge and Unity HD srp Devs are doing a great work, but it is far from production ready .
Unity Devs are the one that know the status of their product, so community stick on their recommendations. No moaning.

Feedback is not moaning.

13 Likes

I have to respectfully disagree for the following reasons.

“Maybe it works for your requirements , but there is an entire world that is about custom shaders.”
HDRP is a template preset made for the SRP. The approach is to create a high graphics pipeline using the power of scriptable pipeline. It begins with the premise to serve a very specific requirement, not mine, but for creating visuals suited for realistic rendering that is physically correct. Custom shaders would be nice, and the shader graph atm not working with the HDRP is not helping, but you could write your own. To be more exact, if custom shaders is what stops you, then just create your own template using the SRP. HDRP is just a template, and that template in its current status, seems to be quite ready to serve as a template taking everything I said above in account, not to serve an entire world of custom shaders.

“That is not the right message.”
I am noticing that the use of the word “moan” seems to cause uncomfort, and I have thus removed it. But, other than that, I don’t see how my evaluation of HDRP is not the “right message”? In simple terms, my words can be rephrased as “wow, it was better than I expected. I think HDRP could be used at production level given the release history, my project duration and the possible trade off of not using HDRP and trying to achieve same visuals with the Legacy Unity”. I honestly fail to see how that sends the wrong message. I also don’t get that it is the wrong message because Unity Devs know best. If Unity Devs know best, then why does this whole community and its forum exist? If Unity Devs know best, don’t you think every feedback would be a moan? If you had different opinions to share even if it was against my opinion, we could have had a constructive discussion, learn from each other and the devs could also have benefited from our conversation, but what you said leads nowhere.

3 Likes

Xour feedback is excellent helping the entire users community. Please don’t get me wrong.

HD SRP is a preview. SRP is a preview.That means that is experimental.
Production means also stable enviroment and business support. Of course everyone is free to do whatever they think is best for them, but it is far from affirm “it is production ready”:
Feedback from community helps to address development and find bug., helping other users,
A good feedback does not mean that is ready for production.
We as users, cannot say *despite what Unity devs say" They are SRP stack developers, so if the consider as preview, that is the only point that count.
Your wrong message is only your statement that is production ready. That can put wrong expectation in users and can lead to misunderstanding.
Thank you for your understanding and have a great day

I use it, but it needs work (XR/some platforms/broken features/graph). Can I ship with it? Yes. That means for some people you can use it now, but for Unity newbies or people without a strong background, there is not enough support material or docs, so they will struggle to even use it right now. By default, it’s not ready for Unity’s main customer base :slight_smile:

Nice review though (which is what I took your post as).

And for me, contact shadows are broken but dir light shadows (up to half a mile) are just fine (once heavily tweaked).

1 Like

Exactly. And I read few one star reviews to excellent assets because people expected to get it working with new SRP
Thats the “wrong expectation”.

1 Like

I understand the caution and I agree that the documentation isn’t great, but if I may, you may be cautioning more than it needs to be. HDRP has enough flags to make people realize that it is WIP (thanks to the preview flag everywhere), what I just wanted to share was the details about it, and for the curious minded, how viable it would be for using it.

And on that note, I still think it is “relatively ready”. It is stable enough graphically and has the benefits of the new 2018.1/2

Most of the critical issues - for example, lightmapping - is the same in other Unity releases anyway.

What could work against a newcomer is documentation, shader graph and the general non Asset Store assets incompatibility for shader based assets. However, for me at least, production readiness is largely dependant on whether an engine can do its job on regular basis. Engine issues, in my experience, is a common thing during development, regardless of the brand, history, and age. It just happens due to the complexity that any modern engines have to work with. Heck, I am still running into issues with my shipped game that uses the LTS version.

To be more exact, maybe I did want to re-examine the expectation that people have established about HDRP - that it is not ready yet. I really really think it is good enough start. Honestly, comparing it to some Asset Store solutions it already is better in many ways.

Other than that, I do have a question for both hippocoder and antoripa. I obviously opted in to work with HDRP, but what would your evaluation be if you were starting on a high graphics project (first or 3rd person) aiming to release the game 2020 Q3-ish on PC? Would you choose to work with HDRP, or stick with Legacy? I am genuinely curious and want to re-evaluate my decision. Don’t want to be regretting my decision two years late.

2 Likes

When @SebLagarde says it’s ready I will believe him!

I have already ported my long term project to HDRP. I understand clearly the talent of the people working on it and have properly evaluated it for my needs. It is mine!!! :slight_smile:

As far as I can determine they are aiming for a Q1 2019 release but do not quote me!

For 2020 I would certainly be looking at HDRP. Most of what could scupper this is very hard to anticipate because it will likely be devilish detail that spoils its suitability for a particular project, and that stuff is hard to predict. How many 3rd party assets the project is going to use, and whether these all get updated to work with HDRP in a timely manner, is another factor.

I am dancing already :slight_smile:

One thing that makes me choke at the moment is terrain. I am pretty sure HDRP terrain is going to happen one way or another before 2020, but I am not sure about the viability as no details have been shared. I started working on props and other things that wouldn’t be effected by the terrain, but still I am a bit worried.

2 Likes

There’s some activity on HD SRP terrain here https://github.com/Unity-Technologies/ScriptableRenderPipeline/commits/gpu_terrain but it’s using the old directory structure + no official word on this. I’d expect it to happen during 2018 cycle as it’s kinda big limitation for many atm.

Initial Shader Graph support for HD SRP is already there, it’s just not in the version we get through regular package manager atm. But for example latest SRP releases on github have HD support for Shader Graph already.

It still got some issues, like for example shader graph shaders for HD don’t currenly show up correctly in Forward Only mode (from pipeline asset) but they seem fine in deferred.

Terrain is getting a mega overhaul and, along with VFX requires special treatment for HDRP. As everything is still separately trying to complete it makes little sense to bring them together before then. Meanwhile we are happily doing “terrain” with the HDRP lit shader (it’s just meshes) and it works wonderfully. We don’t have vast amounts to render though.

I just hate the fact the GPU lightmapper is only being release in private beta… like release it to the public already. :frowning: The videos showing it off like over a year ago was cool and all, but now it’s like who cares at this point.

HDRP is great. Performace is great, but the only thing holding it back is… little to no much needed 3rd party plugins to support it, like Anti Atlasing, Custom Shaders, etc…

1 Like

Check the RSS feed. Whatever it really was that they worked on has been scrapped now anyway.

Also where did you guys hear of that “mega overhaul” ??
I must have missed that.
Because the pain points I have with it are not getting resolved last I heard:

  • less popping with terrain LoD
  • no seams when two terrains of different LoDs are next to each other
  • support for overhands/caves and actual holes in the geometry
  • either not taking so long to calculate the geometry, or preferably simply doing it off the mainthread when loading in a scene asynchronously.

What I heard from the video where it was last mentioned is that none of those problems are getting resolved.
And the thing they are working on will be a better editor and general UI tools for terrain.
Can’t find the vid right now, but that was basically it. And all of that split into 3 phases.

But now that even this is not (from what we can see, not what we guess) even in development anymore… :stuck_out_tongue:

Personally I find it is not a big problem. Sure the terrain as it is is unusable for me, but I’m building my own solution that is simple and does what I need it to do for now. Having a fancy terrain system that takes care of all of those things would be really cool and probably save me quite a bit of work, but I’m grateful for all the other things they’re already doing right now (which are - for me personally - much more important because I would not be able to work around them).

First off, I assume you mean the Roadmap RSS feed. I would certainly not assume that just because a feature has vanished from the roadmap for now, that it is scrapped, no longer being work on, etc. It probably means something has happened, but it is in no way safe to assume exactly what.

You did indeed miss something about terrain having a large overhall. It was mentioned around the 15 minutes 30 seconds onwards mark of the GDC Unity 2018 Roadmap talk. The terrain stuff is clearly considered to be a multi-step mission and whilst some of the planned features are all about editor tools, others are not, they are about performance, rendering and overcoming limitations of the current system.

Anyway I mention the detail from the roadmap talk only to draw your attention to what they had planned, because various timescale/version details in this talk went out of date rather rapidly. A number of things have slipped since, including terrain and the new VFX system. Maybe we will find out more about terrain next week at Unite, or maybe the confusion, lack of detail and unstable timescales trend will continue.

1 Like

To be clear about my slight moan at the end there - I have no problem with the occasional inevitability of features slipping a version or two. I do think that by the time the roadmap video was released (there was quite a long delay releasing that particular video), it was already out of date, and until very recently the roadmap on the website was even more out of date.

Nor do I have a problem with terrain being a multi-step development process, rather than something we get all in one go. Its really mostly communication I am complaining about, eg not only has info not always been updated in timely fashion, but also a clear sense of what different features are planned for which Unity (or pipeline etc) release are not always readily available. This is bound to cause more confusion when we are talking about a fairly large system like terrain and when we already know that the plan was to deliver different improvements at different stages. The phrase ‘this is the year it gets its major update’ was used in the roadmap talk, and then performance was mentioned (in the context of what, at the time, they thought they were bringing with 2018.2). Then the multi-stage nature of these terrain improvements was touched on, then talk moved to brushes (which at the time were scheduled for 2018.3, not sure about now), but there really wasnt any detail about what was supposed to come in 2018.2.

And then, just to make it even more confusing (to me at least), we have the whole HD Pipeline terrain support issue. In my mind I had kind of assumed that they were waiting for a new terrain system before even attempting proper HD compatibility, but a lot of the talk more recently sounds more along the lines of bringing some basic support in HD pipeline for the existing system. I’m not sure if this is in response to delays with the engine side of terrain, and is seen as a bridge to cover the larger gap that delayed terrain improvements has caused, or whether this is a dodgy assumption on my part.

1 Like

Or to put it another way, I’m hoping for a better roadmap talk at Unite Europe, or at least some clarity about specific things, including what has changed since that last roadmap talk.

Since I am a bit tedious when it comes to detail, I should also have mentioned that in the (now inaccurate) GDC Roadmap talk, ‘Terrain Rendering Improvements’ also came up on a later slide, around the 46 mins mark. Again this was mentioned in the context of 2018.2, so the info is already out of date, and I’m only mentioning it in the context of what peoples expectations of terrain improvements were.