Rhino CAD Rendering Comparison Test (Part 1)
For the new year, I’ve taken on a big project to help the jewellery CAD community. Introducing my Rhino 3D CAD Rendering Comparison Test!
(Update 22/3/21: Some corrections to information on Thea Render, and V-Ray licensing options.)
(Update 21/1/21: I’ve added the render settings I used, as well as whether the render was done with CPU, GPU, or CUDA.)
What is the Purpose of this CAD Rendering Comparison Test?
When Rhino moved to Rhino 6 in mid-2019, and Gemvision switched from Matrix to MatrixGold (and notoriously replaced V-Ray with Cycles for Rhino), we now have quite an upheaval in the rendering software and plug-in market for Rhino and MatrixGold users. Now, one year on, there are several renderers vying for the Rhino/MatrixGold user market, and online forums have become very…animated over which renderers are the best value for money.
In order to showcase what each of these renderers can do in one place, and to help my colleagues and students make a more informed decision on which rendering tool to pick, I decided to do my own comparative test run of each software. The final instalment of this will end up eventually in my Frequently Asked Questions section.
I knew right at the start this could get ugly as each of these tools has its vested interest followers (and their first comment would always be “If you know what I know then this render would look so much better…”), I will be taking pains to make the test seem as fair and balanced as possible. So I will be running three different renders, one of which will be done by the rendering company itself in the name of good sportsmanship.
Who Am I to Run This Test?
As I brace for the ad hominem and moving goalpost attacks based on my saying their favourite renderer isn’t the best, I shall lay out my relevant background:
I have been teaching jewellers how to use CAD since 2006, and I have been working with CAD modelling and rendering tools for real time animation and product design since 1998. I teach all major 3D CAD modelling software packages used for jewellery, and have been using Rhino rendering plug-ins since Flamingo version 1.1. Previously I have run courses in Flamingo, V-Ray, and Keyshot rendering tools for Rhino, Carveco/ArtCAM, and Zbrush, as well as Deep Image for 3Design, and I have used other renderers before such as Brazil and 3DStudio Max.
In addition, I have also made and photographed a bit of jewellery before once or twice in the past 18 years…
So, I can only hope that will be enough to satisfy the majority of readers.
My Test in Three Stages
The following are the three renders in each software. Each will get their own separate article here on the blog:
- Render Test 1 “15 Minutes from Opening” – Two “out of the box setup” renders, based on what I can achieve within the first 15 minutes of starting up the program. I will be using only standard pre-sets included (or easily downloadable) within the renderer, and a minimum of environment options changes. One render will be in yellow gold, and the other will be in platinum. Both will use white diamonds.
- Render Test 2 “The Experts Give it a Go” – I’ve asked an expert or exponent for each rendering software to perform this render. They could be either the developer, a sales rep, or trainer depending on who is willing and able. I’ve given them some very specific requirements (which I will list below) which they must cover to maintain consistency, and they will have 2 weeks to deliver the render from the model arriving to the finished piece.
- Render Test 3 “Best I can do in 1 week” – One “best that I could reasonably achieve” render within 1 dedicated week learning to use it, based on my actively pursuing support and learning currently available at the time of this article’s creation. I will endeavour to use every setting I can find (except for Depth of Field) to make the render as good as I can get it. It will be platinum with white diamonds, and I anticipate these results may prove the most controversial.
- Render Test 4 “The Experts Show Off” – Finally, out of fairness, I’ve asked the experts to “take their best shot” at making this model look pretty in a render. They’re allowed 2 weeks, but they can do anything they want with the model and use any tools or tricks they like so long as it’s included within their own software (ie no third party post-production). This time, however, I will allow Depth of Field if they want to use it.
Side Note – Why I’m leaving out Depth of Field
I have chosen to leave Depth of Field controls out of the first two render tests for every software for several reasons. Firstly, I found in my experience that Depth of Field can distract from inadequacies in a render’s quality. Secondly, it greatly increases setup and rendering time, making it unhelpful for Stage 1 render tests. Thirdly, it makes more sense to give the experts a chance to add some pizzazz to their renders somehow, and they will likely have the giant rendering farms required to take most advantage of it.
Comparative Test Rules
For all three tests, I have laid out the following requirements:
- All renders will be made in the most recent version of Rhino 7 as of this printing, with one exception– MatrixGold is currently only available in Rhino 6 as of this writing.
- All renders will use the same model.
- All renders shouldould be made to Instagram standard: 1200 x 1200, and at least 72 DPI. If they can achieve greater DPI, I will specify that and use it (300 DPI maximum).
- Renders can use only the built-in tools within the rendering software. No external post-processing or compositing allowed in Photoshop.
- I have strived to match all renders to the same setup as much as possible. The baseline for this setup is based upon two control images I have generated. The first was made with Matrix 9’s old V-Ray rendering engine (shown on the right), and the other render comes from a colleague using Maxwell (which I’ll use as a baseline for Render Test 3). I am aware that this ring does not include coloured stones, but that is intentional—dark coloured stones can obfuscate how light bends when passing through the gemstone.
I have, however, opted to make two different stage 1 metal renders for yellow gold and platinum. For my Stage 3 renders, I’ll be aiming specifically for platinum.
Additional Requirements
Test 1
- All my Test 1 and 3 renders must be made on the same PC.
- Aim to create an image as close as you can to the one in the baseline above, using only the already included HDR maps which come with the software.
- All renders will be timed.
- The render setup must take place within 15 minutes of starting up the software for the first time, using my existing knowledge of rendering software.
- If there are built-in post processing options in the render screen, they are allowed.
- No depth of field. (see above)
Test 2
- The experts must aim to match the benchmark render in terms of positioning and materials. The environment must be similar (pale background, studio photoshoot lighting). Beyond that, anything goes.
- Aim to create an image as close as you can to the one in the baseline above, using only the included HDR maps in the software.
- The expert will have 2 weeks to deliver.
- Post Processing is allowed within the software only.
- No depth of field.
- Each render test must also include information on the following:
- Computer Specs, specifically Graphics Card, RAM, Processors, OS, and cache settings.
- Setup time
- Global environment settings used
- What pre-sets for materials, environment, lighting, etc used. If anything had to be made from scratch, specify that.
- Time taken by the Rendering Engine
Test 3
- All my Test 1 and 3 renders must be made on the same PC.
- All renders will be timed
- There is little point in timing this render setup in total, as it’s likely it will take the entire week to get it working. But I will time how long it would take to set it up again on a new file from scratch once I get it right.
- No depth of field.
Test 4
- The experts must only use the tools included native in Rhino or their own rendering plugin or standalone app.
- The expert will have 2 weeks to deliver.
- Post Processing is allowed within the software only.
- Depth of field allowed, but if it is supplier must specify it is being used.
- Each render test must also include information on the following:
- Computer Specs, specifically Graphics Card, RAM, Processors, and OS.
- Setup time
- Global environment settings used
- What pre-sets for materials, environment, lighting, etc used. If anything had to be made from scratch, specify that.
- Time taken by the Rendering Engine
Overall Comparison Ratings
Once we have the data from all tests we’ve been able to gather, I’ll distil it into final verdicts. The categories I will be discussing will be as follows:
- Price: How much it costs. This will be the only category without a star rating.
- Setup Speed: How much time it took to setup the render in both cases.
- Rendering Quality: I’ll be discussing how clean and realistic the image looks, and how sparkly the gems are. While I know this isn’t a particularly empirical measure, I will be backing this up with images.
- Material Controls: How much control you have over material properties (beyond built-in pre-sets).
- Lighting and Environment Controls: How much control you have over lights and environment
- Render Speed: How fast the renders are relative to each other. A higher star rating is better.
- Interface: How easy it is to use, and the presence of helpful features.
Final Disclaimer
Beyond this, I can only say that I have no vested interest in any of these renderers, and am merely looking to examine what is commercially available now for Rhino CAD jewellery makers.
I have included all the renderers I have been able to identify which have a working following amongst jewellers. If I have missed one, let me know and I can add it in as a separate test.
Now, on to the different renderers!
Jewellery CAD Rendering Comparison – Render Test 1
With the exception of Rhino 7 and MatrixGold (which I put first because one is the default and the other uses a variation of the same plug-in), all other rendering tools are listed in no particular order.
For reference, my my Computer Specification is as follows:
XMG laptop with Windows 10 Home, an Intel Core i7-10750H @2.60GHz, 32 GB RAM, and an NVIDIA GeForce RTX 2060 (supports CUDA).
Rhino 7’s default renderer (Cycles for Rhino) –
(Render Test 1 run in version 8/12/2020)
Rhino 7 now includes a new and vastly expanded rendering engine called Cycles for Rhino. Depending on the power of your machine’s graphic’s card, you can choose between CPU, GPU rendering, or CUDA (if your graphics card allows it).
It’s interface is built into Rhino, so in theory it should be the best of its kind. Sadly, as of this writing there are still some bugs with setting materials up, especially when using the raytraced view.
15 minute render setup isn’t too bad to do. The closest we come to default materials in rhino is the Type menu, so everything is made at least partially from scratch. But the quality was surprisingly good once you get the colours right. One issue is that the gemstones always look a bit glassy, no matter which advanced custom or physical material options you use. There is an option for an automatic ground plane which you can turn on as well.
The rendering feels really slow compared to the others, but having checked the time it was only slightly below average for the Good render. Final render, however, was very sloooow.
Nevertheless, compared to all previous versions of Rhino, it’s nice to see they finally have a built-in renderer which can hold its own in terms of quality and ease of use.
Rhino 7 Cycles Render Test 1 Results:
Render Times and Settings:
Gold render time – 1 min 47 seconds
Options used: Rhino 7 default diamond and physical yellow metallic material, Gamma of 1, Studio scene, default ground plane material. CUDA rendering. Good quality render.
Platinum render time – 1 min 37 seconds
Options used: Rhino 7 default diamond and physical white metallic material, Gamma of 1, Studio scene, default ground plane material. CUDA rendering. Good quality render (500 samples)
I tried using Final Render (1500 passes) for one image, it clocked in at a whopping 4 min 52 seconds, with only a modest improvement on the quality.
MatrixGold Cycles –
(Render Test 1 run in 2.3.20183.1002)
MatrixGold uses its own heavily modified and streamlined version of Cycles for Rhino. All the materials for jewellery are there, and they’re set up properly without having to fiddle with anything. Like Rhino, you can choose between CPU, GPU rendering, or CUDA (if your graphics card allows it).
15 minute render setup is quite easy. There is no drag and drop, but the rendered preview is much better than vanilla Rhino 7. The ground planes are already made for you in every file, so they don’t need to be turned on.
But while the interface is quite easy to set up and adjust, it still has annoying quirks. If the images look inconsistent, that’s because they are– One particularly annoying quirk of the interface changes the defaults for post-production in between every render. This means you have to continuously adjust exposure gamma and tone mapping for every image, even if you have good settings for one, getting them again isn’t guaranteed. Upon my saying this, I’m confident that some power-user will tell me I missed some switch or another. If you know how to fix this please let me know!
Also, given the location of the Render button, it’s far too easy to switch to Side Viewport when selecting materials.
Renders behave much the same way as Rhino. However, for some reason, the default settings tend to gravitate back towards dark and soupy without post-production every time.
Having said all that, I found it interesting how the render was actually better in MatrixGold than Rhino even when using their default with fewer passes.
I did notice the stones had to be cut out from metal to look right. But when they are, they certainly look among the best for all the renderers here.
MatrixGold Cycles Render Test 1 Results:
Render Times and Settings:
Gold render time – 1m 12 seconds
Options used: Diamond and Platinum Ruthenium Textures, Studio Diamon environment, CUDA rendering, 300 passes
Platinum render time – 1m 12 seconds
Options used: Diamond and Platinum Ruthenium Textures, CUDA rendering, 300 passes, Studio Diamond environment.
I didn’t test 1500 pass render.
Maverick Studio –
(Render Test 1 run in demo version 2020.14)
Maverick connects to Rhino via a Bridge software plug-in, but it also works completely standalone. You can buy the program via either monthly payment or standalone license (with annual update subscription). They’re currently maintaining an intense bimonthly update schedule, so they’re working hard to improve it.
It not only uses GPU rendering with CUDA, but cannot even be installed unless your graphics card supports CUDA. Therefore, it might be worth your while making sure you check your graphics card before using this software.
Maverick features what must be one of the easiest to use interfaces of the lot. Stones look very good with minimum fuss or fiddling.
Applying materials and environments is as easy as Keyshot. It’s not perfect though– Items don’t highlight when you drag a material onto them like they do in Rhino, and manipulating the environment and the position of the render is a bit awkward and relies on its own unique interface of buttons on the side. But you get used to its quirks fairly quickly.
The material and scene editing options for Maverick behave more like Modo’s rendering tools (also used in 3Design’s Deep Image) than V-Ray. They don’t seem to do layered materials, so it offers less detailed modification options in exchange for being comparatively easier to learn. The ground plane is also automatically generated for you.
Draft renders are also very fast to complete. However, much like Keyshot, setting up higher quality renders is slower, and computing those final renders is also comparatively slow, but still faster than Rhino Cycles. Annoyingly, the “Remaining time” estimate during renders keeps changing so it’s not a reliable indicator of actual rendering time. Fortunately, renders are faster than what it predicts.
In terms of stability, I found Maverick relatively bug-free compared to some of the other renderers. There were no technical glitches with the software that I encountered.
The 20 minute out of the box render was decent enough quality, although a bit grainy and relatively slow to create. The stones looked glassy, and it’s hard to figure out how to make them not look that way. Like most of these renderers, you have to cut out gemstones from the metal to make them look right.
For their demo, Maverick Renderer covers their rendered and preview images with watermarks, but at least it lets you test it.
Maverick Studio Render Test 1 Results:
Render Times and Settings:
Gold render time – 2min 50 sec
Options used: Diamond and Gold 18ct polished texture, Denoiser on with default settings, “01 floor SC” floor type used, 1200 x 1200 rendering pane, Final Render
Platinum render time – 3min,
Options used: Diamond and Platinum polished texture, Denoiser on with default settings, “01 floor SC” floor type used, 1200 x 1200 rendering pane, Final Render
Thea for Rhino –
(Render Test 1 run in demo version on Rhino 7 2.2.118.1875)
Thea comes fully embedded within Rhino. You can lease Thea directly from their site.
Thea runs renders using CPU, bit it can run in “hybrid mode” that means it will use CPU GPU at the same time. It’s also one of the few which can take full advantage of either NVIDIA or AMD cards.
(UPDATE: I only learned after the fact that the latest version of Thea was mainly built for Rhino version 6, and while version 7 can run the software, the developers have not finished development on version 7, so I was running a kind of beta version even though this was not communicated by the site or developers. Given how much I found out about this software a month after I ran the test, it is very telling of the gaps in the initial documentation.)
While the renderer does work well with all of Rhino’s existing rendering interface and panels (including Rhino’s drag and drop), the Thea interface otherwise looks Half-finished. Also, there’s an unfortunate lack of a working preview shading mode in Rhino. That does fit with my general overall impression of Thea– it can produce good images, but there are still a lot of bugs.
The 15 minute out of the box render was pretty good quality, even though there are no materials for jewellery initially available. It was definitely one of the fastest renderers for its quality out of the box. The ground plane works the same as vanilla Rhino 7.
The stones were very glassy, and it’s hard to figure out how to make them not look that way. The interface itself is relatively hard to work out.
Like most of these renderers, you have to cut out gemstones from the metal to make them look right.
Thea offers layered materials, which are potentially extremely powerful, but they make the options for changing materials as extensive and difficult to learn as V-Ray’s if you need to create your own materials.
For their demo, Thea Renderer covers their rendered and preview images with watermarks, but at least it lets you test it.
Thea Render Test 1 Results:
Render Times and Settings:
“Gold” render time – 60 seconds
Options used: Default Rhino textures for Gems and yellow metal in Rhino 7, Uniform Illumination turned on with intensity of 1.2, Gamma of 1.2, Studio scene,
“Platinum” render time – 61 seconds
Options used: Default Rhino textures for Gems and grey metal in Rhino 7, Uniform Illumination turned on with intensity of 1.2, Gamma of 1.2, Studio scene,
(There weren’t any jewellery materials with the default libraries).
V-Ray 5 –
(Render Test 1 run in trial version 5.0001)
V-Ray comes fully embedded into Rhino, and can be purchased either with a single payment with an additional optional price for perpetual upgrades, or via renewable term licenses of monthly or annual. With V-Ray 5, the many myriad previously available license types have consolidated into one much easier to use piece of software.
One interesting feature: like Rhino Cycles and V-Ray, you can choose between CPU, GPU, and CUDA supported rendering. But V-Ray also supports RTX rendering within its options (which I used here just to try it out).
I found setting up Chaosgroup’s licensing server for the first time to be a bit of a headache, but their tech support is good.
The interface has certainly come along in leaps and bounds since Version 3.6, and is much easier to use. And while they fixed the compatibility errors with old V-Ray materials imported in with older models, the software still has a few bugs– I could not find a built in time, so I had to time the renders myself. Also, I couldn’t get the software to apply materials directly block instances (an possible issue with gemstones in older Matrix models), but once I removed the block instances from the stones it worked.
Unlike the others, V-Ray still uses physical objects for the ground planes. Considering all of others seemed to have moved away from this, it’s a bold old school move. But they have an automatic physical ground plane button built in to their interface (which builds a giant surface under your object).
The 15 minute render setup involves downloading the entire textures library from Chaos Group’s servers, which sadly takes much longer than 15 minutes, so I’ll assume we’re starting after the download of the 9GB library finishes.
V-Ray’s handling of gemstones and transparency has always been top quality, and with this newest version it does not disappoint. Even at medium quality, the renders were well above the average. The colour of the yellow gold on their default material was a bit more saturated than I’d like (more like Asian 24ct plated wedding jewellery), but that’s pretty easily adjustable. Also, the exposure was a bit dark by default, so that had to be adjusted as well.
One unique feature is that it’s possible to make the stones render nicely without being cut out. Once you figure out how to do that, the gemstones look among the best of all renders here.
Its price for V-Ray perpetual is slightly above average.
V-Ray 5 Render Test 1 Results:
Render Times and Settings:
Gold render time – 49 seconds
Options used: Diamond and Gold 14k blurry texture, Denoiser on, RTX rendering engine, Medium Quality, Exposure value 9.7.
Platinum render time – 49 seconds
Options used: Diamond and Silver blurry texture, Denoiser on, RTX rendering engine, Medium Quality, Exposure value 9.7.
Keyshot Pro –
(Render Test 1 run in Keyshot Pro 9)
Keyshot attaches to Rhino via a bridge software plug-in which allows import and export.
One interesting feature: like Rhino Cycles and V-Ray, you can choose between CPU, GPU, and CUDA supported rendering.
What they don’t tell you, however, is buying into Keyshot for Rhino or Zbrush is a dead end, and if you upgrade one of those dedicated renderers to Pro, you won’t be able to upgrade into a standalone Keyshot Pro license without buying the software all over again. Considering it is also the most expensive renderer on this list, this seems rather thoughtless of them.
Unfortunate licensing and payment structures aside, Keyshot still remains one of the fastest rendering setups there is. Only Maverick can keep up on rendering setup speed. The interface takes some getting used to since it behaves so differently from anything else out there (especially when it comes to manipulating objects in the scene), but I was still able to figure it out pretty quickly.
They generate an automatic ground plane for all imported models, but like all the other objects in Keyshot, you control from the scene control editor.
One big advantage Keyshot has over the others– unmatched preview quality. They made Progressive Render and drag-and-drop interfaces popular in product design, which explains why these features in Keyshot remain more mature and well-implemented than even Rhino’s default. The Progressive Render itself was almost instantaneous, and often nearly as good as the final render. Given how the final render’s speed drops down close to average comparative speed, I can see why people often don’t bother with doing final renders with this software.
Oddly enough, saving full renders is weirdly slow though, as is closing the rendering window. Also the Denoiser tends to cause occasional crashes, as I had one during the rendering section.
During the 15 minute render, the stones looked better than average, although it did take some fiddling to make the metal not look so cartoonish with its colour and reflections. The images still look pretty good though.
If you use their demo, Keyshot Pro covers their rendered and preview images with watermarks, but at least it lets you test it. Because I already had a copy of the software, I didn’t need to worry about this.
Price-wise, the Pro version of the software is nearly double the cost of any of the others on this list.
Keyshot Pro Render Test 1 Results:
Render Times and Settings:
Gold render time – 1 min 21 seconds
Options used: Diamond and Gold 24k polished texture, startup studio environment, CUDA rendering, custom render size of 1200×1200.
Platinum render time – 1 min 19 seconds
Options used: Diamond and Platinum Polished Texture, startup studio environment, CUDA rendering, custom render size of 1200×1200.
Final Summary
I’m not going to pass any judgements yet on the various packages until I’ve had a bit more time with them, and I’ve had some imagery back from the various expert users. But it is interesting to see the differences between them.
In the next instalment, I’ll share the results of the first render test from the professionals and developers.
Jack Meyer
Bespoke jewellery designer, and specialist in jewellery CAD/CAM and emergent technologies that affect jewellery.
Related Posts
All3DP Article – Best 3D Printers for Jewellery
I ask kindly how to do it “you have to cut out gemstones from the metal to make them look right”
Hi Giuseppe.
Simple, really– make a cutter shaped like the gemstone using Revolve, Rail Revolve, or Loft, and then Boolean Difference the surface with that cutter.