Hello everyone,
We are currently in the process of updating our hardware structure in order to implement Rhino 8. We were recommended the Nvidia graphics card. Are there any Rhino 8 users among you who have had good experiences with a specific type of this graphics card?
Thanks in advance.
Kind regards,
Jacqueline Schneider
Hi there, and welcome.
Many users here would actively promote the usage of Nvidia GPUs for use with Rhino. They are generally rated as the most stable GPUs, with the better side of driver support.
However, that doesn’t exclude you from using an AMD GPU, which can also be used.
Depending on your workload, you can always ask more about which one may be right for you; as there is no point in overspending on GPUs if you can avoid doing so, if you simply do not need the power for rendering or viewport.
Additionally, consider which renderer you are using, as some have specific hardware vendor requirements.
What is your budget for a graphic card?
Btw, you can run rhino8 on anything that runs rhino7
If you want to render out things then you should go for an Nvidia GPU. The more powerful and more memory it has the better, but of course your budget will be the limiting factor.
Out of curiosity, if HIP is working in Rhino 8, is there something specific that makes an Nvidia card a definite recommendation specifically for Rhino 8 over an AMD?
My only guess was because HIP-RT isn’t there.
You know that post sounds a little weird right? And some of the replies here too.
Get the best Nvidia you can afford, it’s no more complicated than that.
That’s all that matters, and it’s not due to CUDA, it’s due to Nvidia being the only company that is even half-heartedly attempting to provide adequate OpenGL drivers. It’s been…a decade? since Nvidia’s put out a driver that’s caused Rhino to crash and require rolling back. It’s routine for AMD.
Hmmm, no.
I’ve never had any regular problems with other OpenGL applications on any GPU beyond Rhino. Even Blender doing reasonably heavy rendering, viewport work has been very stable even on my obscure Intel Arc GPU last year.
In fact, the only applications I have ever encountered serious regular graphical problems with are Rhino and Solidworks (some welding work I was doing, it really, really didn’t like).
I’ve used 3DS Max, Inventor, Siemens NX, SmartCAM, Creo, Blender… all of which I have used between light and very heavy use, and never had any particularly jarring OpenGL issues beyond something like a random flicker, or very rare crash; even when shunting all of the work onto the CPU (where I have a “meh” iGPU).
Not that it really matters, but since you are keeping track.
RH-81020 Crash in CRhinoDisplayEngine_OGL::ShowBackBuffer
→ This one is NVIDIA-specific and has been causing a bit of a headache…
But, yes…
-w
In my experience Raytraced and Rhino Render still are much faster on Nvidia hardware. I have a Radeon Pro W7800 to test with, but it is clearly slower than my Nvidia GPUs
Interestingly, I posted some complaints last year on this forum about my Intel Arc GPU that I thought were very specific to that new GPU architecture. Since I swapped to an Nvidia (not because I disliked my Arc, but purely for render compatibility), I have been encountering exactly the same problems doing exactly the same style of work (CurvatureGraph and MatchSrf work). I can inconsistently cause Rhino to crash out without crash dumping or error reporting, directly to desktop, with precursor display problems.
Sorry to derail the OP’s topic (sort of).
This is contrary to what we’re experiencing.
Try ZLUDA
When comparing opencl vs price performance, amd is a clear winner. Visualization is where Nvidia shines. But even there, if I’m not after real-time rendering, I’d probably use a cloud service. At current costs, it makes no sense to render locally.
There is a nice project called zluda which basically makes cuda code run on amd/intel hardware.
ZLUDA unfortunately seems a rather dead/hanging project that nobody wants ownership of. It’s a cool concept, but unfortunately seems unlikely to work out, and certainly I couldn’t see major or minor software developers using.
I imagine as well, that Nvidia will specifically not permit these actions for long, and just add more phrases to the 2000 page EULA for CUDA.
However your point about price/performance is interesting, which is why I raised the point in the first place. Nvidia has both AMD and Intel beaten for now generally in rendering, but not always in price/performance. But One big problem here is that many raytrace renderers will additionally use OptiX; the AMD and Intel alternatives to which can’t seem to perform anywhere near as close, objectively, right now.
the real story is that intel and amd stopped funding the project - i guess for legal reasons - but the guy seemed ambitious enough to continue on his own. besides he started the project without any funding, so… there is reason to be cautiously optimistic.
last i heard, it worked well with blender
true. but when it comes to rhino, my cpu single thread/ cpu-ram has always been the bottleneck. again, i don’t render in rhino (i’m sorry to say, but it’s a pain to try that - so much so that i switched all rendering to blender)
btw, unreal’s lumen does an ok job on amd as well.
I think for now this will go in the not-supported category for Rhino. If a user wants to try it, sure they can do what they want. If it works with Rhino, great. If it doesn’t, well, not surprised.
I doubt the starter app will be able to handle child-processes that do kernel compiling. And I’m not sure if it will be able to show Rhino a CUDA device. Anyway, I don’t have currently the resources to even start investigating this beyond looking at the depressing known issues section GitHub - vosen/ZLUDA: CUDA on AMD GPUs
Going back on topic.
Nvidia GPUs will currently give best performance at least for Raytraced and Rhino Render.
I have had good experience overall with the 4090RTX …
Many thanks to all of you for the great response to my post. I really appreciate all the feedback and the exchange of experiences. We were recommended the Nvidia graphics card mainly because of the rendering (which many of you also referred to) and also because of our additional use of Exactflat. We used the Flamingo renderer until it was discontinued and now use the Rhino Renderer.
In that case definitely go with Nvidia GPUs 
Apple also bites the dust. My new laptop has Nvidia RTX 5000 Ada GPU. Raytraced display mode is very fast - around one thousand samples per minute, but its rendition quality is bad - apparently my rendition skills are mediocre.
Nvidia GPUs: https://tinyurl.com/44yc3rkb
Nvidia GPUs for desktops: NVIDIA RTX in Professional Workstations
technical details: Professional Visualization Product Literature | NVIDIA