[rusteberg]

persian rug looks great.... just noticed little buddha

what about rag loom?:

[feynman]

Mr. rusteberg, your render with these loose threads is what seperates the men from the boys; I have never seen carpet or fabric rendered so life-like...



Mr. Polyxo, if you don't render them, do you photoshop them in from an image library with matching perspectives?

I don't ever have to render actual carpets.

Mr. bograt, how would you add such "pub-at-last-orders-lads" details like fag butts, crushed KP nuts and Bishop's finger spillovers to a carpet MXM? With separate BSDFs using more channels? Are you in posession of a SHEFFIELD_PUB.MXM? You seem to posess vital first-hand knowledge!

[eric nixon]

Rusty, I would make such a rug with a modelling disp combo. I would arrange a line of intersecting cylinders, map a stripy disp cloth, lock the uvs and twist the geo... too easy try harder....

In theory (because I'm not gonna bother + I'm scared ) you would prob need 25 gigs ram.

For a dirty pub floor make a seperate geo plane + alpha mxm for the 'crud' based on a photo of crud, similar concept to this disp pizza mxm (original maps + mxm made by M-que) but with loads of alpha, imagine someones been sick on the floor..



More geometric elements like fag buts could be modelled.. I know I'm taking your question a bit seriously, but its an interesting area of 3D - the crossover between using displacement and modelling depending on how close the camera gets.

[feynman]

So, with only displacement at my disposal, a bit of brighter specularity at the tips and some R2, that's as tufted as my carpet wants to go; 1 hour and 40 minutes to SL 19 with pretessellated selected and 128 subdivisions. That must do now, until I can use grass with 2.7 again.

[eric nixon]

How much ram is it using, because 128 subdiv sounds high, I think its easier (not sure) to use finer geo and lower disp subdiv.

The other thing that will definately help is to consider that most carpet hairs are not very diffuse/rough theyre smoother, try roughness 60-80 on the main bsdf with a rougher fill bsdf.

But the main thing is getting the size of the mapping right to get the most out of the disp, and this will vary from geo to geo...

[feynman]

When rendering, the task manager shows an increase of 0,8GB RAM used.

[eric nixon]

Is the disp map fine enough - tiled enough? Turn smoothing off...

Maybe try testing on the simball, and then use a subdivided plane which has the detail of the simball (put the plane under a subdivision modifier in your host app)

[feynman]

Smoothing is off, size of map is 2000x2000 pixels. Tiled 2x2 across the plane which has 4096 triangles. Might try with 3000x3000 pixels or a plane with twice as many triangles - but eventually, the render has not enough pixels to show that much detail.

[Polyxo]

feynman, you used a pretesselated plane with 4096 faces and subdivided it 128 times, used pretesselated displacement and you only see an increase
of memory-use of less than a gb? Something can not be right. Using 128 subdivisions should let every machine run out of memory.

[feynman]

That's odd - that's the number I typed in in the MXM's displacement component Subdivision field, and my plane is definitely consisting of 4096 triangles. Apart from the museum room, which is only a few triangles, and the PLOPP stool, which is 64000 triangles, there's nothing else in that scene. Actually, I don't get much faster rendering using pretesselladed versus on the fly; but that's probably another issue.

I've read all information from this thread and similar threads and looked at all given explanations in http://support.nextlimit.com/plugins/vi ... Id=1475519

[Polyxo]

Well, say you feed in a quad-mesh with 2048 faces which transfer to 4096 triangles, every subdivision increases the count of virtual faces by 4.
Now do this 128 times, each time with the output of the previous multiplication as input. The output-mesh was of stellar size. Something must be wrong.

[feynman]

So, why is this scene not supposed to render? Meaning: because it does render with the settings I triple-checked, somewhere there must be a bug? There is no doubt that the plane has 4096 triangles. There is no doubt that I entered 128 in the Subdivision field. Therefore, I conclude, that the Subdivision input field is buggy.

[Polyxo]

So, why is this scene not supposed to render? Meaning: because it does render with the settings I triple-checked, somewhere there must be a bug? There is no doubt that the plane has 4096 triangles. There is no doubt that I entered 128 in the Subdivision field. Therefore, I conclude, that the Subdivision input field is buggy.

Best was if you simply posted an mxs-file with the rug.
128 subdivisions on an already dense mesh would not fit any ram money can buy and no displacement-map would provide enough detail to
justify such subdivisions. One needed to feed in a map of terrabytes in size to do such a triangle-count justice. But that's theoretical.
If everything is set up correctly your machine simply would crash.

In your case Subdivisions up to maybe 8 or ten with a good map appeared reasonable but would already need a lot of Ram. Staying lower probably
would do. Just look at Erics experiments to get a reference, he always mentioned the ram-consumption.

[Polyxo]

Do you maybe have "cache meshes" activated?
Then you can dial in what you want but the renderer still exports the cached version of the mesh.

[feynman]

Don't know where "cache meshes" is to be found in Studio. But I post the MXS when back at work. With only 16 subdivisions (which is what I began with) for example, it looked worse. Then, I used 32, 64 and finally 128. Maybe that MXS I'll post will reveal something.