Hardware rendering versus "software" or AI rendering?

Rendering within Rhino is super convenient. It’s hard to learn new programs. Maybe not “hard” but we are required to learn so many different apps these days.
My problem is that my results suck. I have a very basic understanding of lighting, materials, material mapping, etc… I could and will invest time to get better. My hardware is also dated; I’m about to upgrade.
But regardless of all the above, I was thinking back to when I did some rendering in Revit with next to zero knowledge of lighting, textures, among other things… and achieved pretty decent results. That Laptop had intel integrated graphics and 4GB of ram (it actually ran Rhino 6 okay).

Up until last month I had the opinion that I just needed to get better and I just needed to upgrade. But noticed some things since then:

  • While some people achieve photo-realistic results, other’s are struggling.
  • Flamingo seemed way more forgiving both in relation to hardware and ability (I’m basing this off hearsay).
  • My scenes often look better in “rendered” or Visual ARQ’s “realistic” mode. Even a basic ray-trace should look better than the rendered mode, which is a rendered preview if I’m not mistaken?
  • Sometimes stuff doesn’t work right, which makes it harder for beginners to figure out if they’re doing something correctly or not. For example, recently I had major texture mapping issues. This might be due to my hardware.

Rhino’s Ray Traced hauls a** for certain things like glass, liquids, etc… and I’m not going to deny that my ability and hardware are limiting. I guess I wish there was a more beginner-oriented method of making my work come to life. Maybe AI is the solution? I know there’s a new(ish) AI rendering app on Food4Rhino somewhere - I’ll be giving that a whirl soon enough. How hard would it be to bring back Flamingo from the dead (or even Penguin :smiley: ).

I guess it depends what you want to achieve.

If your primary and dominant focus are for provision of actual parts and finished products, then AI rendering may be a great choice. Concepting and such may lend itself to AI better than fighting with lights, environments, and Rhino 8 Cycles errrrm… unique behaviours.

However, AI rendering does tend to use the same hardware. Most of it is GPU accelerated, and you will come up against the same limitations; perhaps even more so with VRAM on occasion.

Personally, I think rendering and renderers have stood the test of time as a proven method. I think while a great concepting tool, I prefer learning the skill, and understanding why something is happening. The AI doesn’t actually understand why it is doing anything, as it is just chucking a load of averaged results it has stored as “correct” from other peoples’ work from over the whole Internet. You are likely to notice that your work may begin to take on the “sameness” element as is often regurgitated by other AI artists. Often wishy-washy, unresolved, and unrealitic high contrast renders.

Cycles is great for many applications; and I would certainly start there and persist for a few months. Especially if you are on new hardware and can get something like an RTX 4070 Super; it’s really good.

Beyond that, my bias would go towards bella renderer, owing to its 1080p trial mode. Maybe a bit extreme for your case, and is still awaiting guidance in help files, but world class. Then of course there is Vray and Octane; both of which sport fast GPU modes. Chaos Vantage is often instant on the right hardware… if you are willing to pay the high cost.

If you have something you see as very typical for your work, I’m happy to see what Cycles and Bella can do if it helps; but I’m sure others can do far better. And there’s no substitute for doing it yourself I suppose.

2 Likes

Hi @keithscadservices
Can you be a bit more specific? “Better” is so subjective, and hard to help with… but that’s not going to stop me from trying :joy: I’m going out on a limb here, so if I’m wrong, just ignore me! There are a lot of differences between what “Rendered” and “Raytraced” presents, but one of the main differences is that Raytraced has inter-object and environment reflections, where Rendered only reflects the environment map. Unfortunately Raytraced (aka Cycles) can’t reflect the environment map on a white background/groundplane, so if you opt for a background colour (instead of the environment), reflections tend to be boring and flat for those parts of the image that are reflecting anything below the horizon. Another thing is the tone mapping, where (at least in my eyes) the default Clamp is a tad dull - I usually go for Filmic set to medium contrast.
Another thing to look out for, is the actual environment. The default HDRi isn’t really suitable for Raytraced renders, I think. No contrast and no colours. Another thing to be on the look-out for is “mathematically” flat surfaces. The tend to be really boring in renders compared to real life. This can often be worked around using a large scale bump map on the material. Another thing - and this is something I really wish would be addressed - is that Rendered and Raytraced look nowhere nearly the same, and they should. It would make fine-tuning renders much faster, if the colours, contrast, overall brightness etc. was the same for the two.
But please post examples (3dm or renderings) of something that looks “better” in Rendered than in Raytraced. Here’s a quick example of Raytraced (upper) and Rendered (lower) where tone mapping, contrast etc. have been adjusted, (hopefully?) making the Raytraced rendering look a little more interesting.



HTH, Jakob

I was never really somebody who did professional rendering but I frequently had to „sell“ my work and that involved an appealing presentation. Since I was not directly being payed for doing so, it was critical to be efficient and make it look professional without much effort. One thing I noticed is that this not necessarily means to render something ultra-realistic.

Its all about transferring an emotion, telling a story. Therefore composition, lighting and coloring can have a bigger impact than anything else. There are many books about this matter.

Other than that, the first time I loaded a production-ready car-part into Rhino, I was surprised of the looks with the default shader.

The reason, is basically that in automotive design people care about reflections. So surface/model quality affects the results to a great extend.

1 Like

This almost exactly describes my application.

The information in your post will be super valuable to myself going forward. I definitely won’t be ignoring all the useful information you’ve shared :wink: . Thank-you!!

I will in the future! Anything I’m currently working on is more client-oriented and wouldn’t really make a good subject anyways. I’ll dig out some past projects when I have time within the next couple weeks.

I might end up with an RTX 4070 laptop GPU (they aren’t that much slower than the desktops - and at least I avoid the CPU issues). I’m thinking about either spending a bit more and getting the 4080 (laptop), or going with an interim solution (3070ti or 3080) if I see a good deal on those. My needs might change in the next coupe years and at that point I might opt for a desktop.
Even on my GTX 1650ti I’ve gotten acceptable results with Enscape. Actually I was really impressed with Enscape. It’s just a bit expensive and not available for R8 yet. I basically need things to look “good” but not necessarily photo-realistic. Enscape had a really good feature for drawing in edges (silhouette edges) and I could get a really well-defined semi-realistic look.

One thing I was interested in is if the AI/Software rendering could ‘fudge’ lighting somehow. Video games can do this. The lighting previews in non-ray-traced modes aren’t the greatest. I’m still trying to figure out how I got better results in Revit (with intel integrated graphics) than with cycles on my current PC. I’m sure when I figure my sh*t out, cycles will give much better results.