Eleven Hour Render With This .tgd?

[RichTwo]

I got an invitation from Planetside Staff to share my concerns over an over-lengthy render time only when I choose Path Tracing.  It could be my machine which is several years old.  It could be due to a displacement factor.  I've used some fairly complex node arrangements and radical warp / displacement settings and have seen it have an effect.  

So here it is, and y'all place nice...

[Oshyan]

Thanks Rich, we'll take a look! PT is definitely going to be render-intensive, but I'll bet we can optimize this at least a bit, if not a lot.

- Oshyan

[gao_jian11]

The quality of the cloud is too high, and in this scenario 0.5 seems to be fine. I think the time is mainly from the rendering of the cloud.

[Oshyan]

Thanks for taking a look. You're right that the clouds are definitely part of the problem, but those will not make Path Tracing significantly longer, so they would affect even normal renders. The terrain also takes a long time to compute here due to many different shaders, and that will be more of a problem for path tracing.

Update: Matt has informed me that clouds actually do affect PT calculations and speed directly, so reducing that does help a lot. Still the terrain is tough to calculate and is a big part of the rendering challenge in this scene. Fortunately reducing cloud quality is a much easier improvement to make vs. reducing complexity of the terrain shading which would involve a fair amount of work.

More details and specific recommendations soon.

- Oshyan

[Dune]

Too many shaders can indeed take a long time more. Some time ago I rendered a file with about 100 shaders for a particular ground effect, which took 7mins for a certain crop, while replacing that with about 7 shaders that got a reasonably similar effect (reasonably being that you won't lay awake over it), which took less than 2mins. More than 3x as much on a complete render is significant!

[digitalguru]

oo many shaders can indeed take a long time more.

I had many fake stones shaders in a scene chained through merge shaders and that added 50% to my render time. I split them via a few Surface layers and it reduced the time by about 30% so it's definitely worth experimenting with shader setups to see if they can cut render times.

[RichTwo]

FYI - with the Standard Render it took around an hour (camera orbit close to ground level).  As for the complexity of node arrangement, I don't consider the one for the .tgd to be overly so. I actually removed quite a lot from the original. 

Thanks to all for your help and suggestions!

[Tangled-Universe]

I noticed there are quite a few XYZ compute nodes in the setup and there are also quite a few nodes with internal networks containing displacement functions like cracks. So at first glance it looks very neatly organized and "lean", but I think this is quite a displacement intensive scene all togheter. Anyway...about the "Tex coords from XYZ" nodes...

I do know that computing normals (compute terrain/normal nodes) are expensive, causing the renderer to parse the network upstream 3 times for each normal computation.
This is why the general advice is to perform displacements before compute terrain and preferably just once. Displace -> compute terrain -> displace -> compute terrain etc. will make the number of shader calls skyrocket.

However, I do not know how this behavior is for "Tex coords from XYZ" nodes, which returns the texture coordinates at that renderstage, useful for height restrictions, but not slope restrictions.

Long story short: what happens to your rendertime if you disable all of those?
Yes I know, your render will probably look vastly different and not what you want, but we want to find the cause, right?

[Dune]

I get the impression that an XYZ shader doesn't cost much, certainly not as much as compute nodes. I use them quite a lot too. I wonder, btw, if patch sizes differ much in cost.

[D.A. Bentley (SuddenPlanet)]

I downloaded that tgd and rendered it on a 16 core Threadripper and it only took just under two minutes (1:45) using the standard renderer, and with the standard renderer with Defer atmo/cloud checked on it took 13:14.  When you said 11 hours were you increasing the resolution to something greater than the 800x450 which was how the file was set?

There is a lot of complexity in various node sections as discussed here, but I am wondering how much RAM your PC has, and what kind of CPU.  Have you rendered the image with Task Manager open to watch Hard Drive and Memory activity (lots of activity probably means your swapping memory to disk especially if you see your memory max out).

Update:  Oh after looking at it more I realized the file was not set to path tracing by default, so I am seeing the slow render that you were seeing, and it seems to be using only 8GB of total memory on my PC, so not memory intensive at all.  Final Path Tracing render time on the Threadripper was 21:41 so roughly 10 times longer than standard renderer.

-Derek

[Oshyan]

I know it's taken me a while to respond in this thread, but trust that I have a big post coming.  I ran into some unrelated computer issues (which I'll post about separately!) while working on testing and optimizing, so it has delayed me a bit.

- Oshyan

[Tangled-Universe]

Interesting Oshyan!

[Oshyan]

Hey guys, just a heads-up, I split this topic as the Compute Terrain, Tex Coords, etc. discussion was beginning to be a bit more in-depth than needed here, and seemed like useful discussion on its own. Here's the new thread: https://planetside.co.uk/forums/index.php/topic,26887.0.html I'll post an update there as well.

- Oshyan

[Tangled-Universe]

Thanks for splitting it up Oshyan, I was starting to get worried I got carried away by the topic instead of helping out Rich. Sorry for that Rich!

[RichTwo]

Thanks for splitting it up Oshyan, I was starting to get worried I got carried away by the topic instead of helping out Rich. Sorry for that Rich!

Yes, thanks to Oshyan for splitting that part of the conversation away.  I didn't understand much of it, but it still seemed like a valid concern.  No problem, TU!  I know we each go our separate ways in how we approach TG.  Mine tends to be as simplistic as I can manage, though I wish I had the knowledge that a lot of you do.  Any bit of help is valuable to me.  

I patiently await Mr. Greene's evaluation and news...  all good, I hope!