Cinebench score for 4D Cinema (real world performance).. is it sufficient?


#1

Hi,
I just did a cinebench: Version R23

My score was:
Multi: 13,100
SIngle: 1,550

What kind of performance will I get from Cinema 4D when it comes to rendering, etc?
I’m wondering up till which point my computer will be sufficient, and at which point of complexity I would need to upgrade certain components of my computer.

I dont need lightening fast rendering, but I also dont want to get frustrated with slow rendering.

Opinions are welcome :slight_smile:

=======
Computer spec:

CPU: Ryzen 7 (5700X)
Motherboard: ASUS PRIME B550-PLUS
RAM: Corsair VENGEANCE LPX 32GB DDR4 3600MHz (AMD Ryzen Tuned)
GPU: EVGA GTX 1050
2x Crucial SSD drives
Windows 10 (Home)


#2

hi. for standard/phisical render 13к is good. or for Corona, i guess its not bad.
with standard renderer you can make short animation but without GI option.
for next level of performance/quality you need 3060 or better GPU and migration to Octane/paid Redshift renderer


#3

I’ve long thought Cinebench is a very dated benchmark which doesn’t provide a useful benchmark for using C4D in production let alone any other 3D App. AMD’s massively multi-core CPUs always bench well but unless you’re actually doing CPU based rendering they’re the wrong choice for most 3D artists.

Even if your 3D App is impeccably multithreaded the law of diminishing returns kicks in surprisingly quickly around 6 cores and after that the assistance from further cores tails off rapidly during interactive workflows like character rig playback and deformer playback. A high clocked 8 core beats a Threadripper in interactivity all day long.

I was discussing fluid sim and multicore performance with a developer and he’d tested his solver rigorously and said 8-10 cores was a sweet spot and don’t bother after 16 cores you’ll just be throwing money away for very little improvement unless you were going to spool up multiple instances.

A good 8 core like the 5700x is plenty good enough for the vast majority of 3D tasks for years to come. The 1050 is a bit weak and would be the first thing to upgrade especially for GPU rendering down the line but it’ll be fine for running the viewport.


#4

The thing is, you don’t really have to choose between good single-core & high core count nowadays.
A 5950x or 12900k gives you pretty much the best of both for a semi-affordable price.
I’d agree, Cinebench is really only valid for CPU rendering, but I think a lot of people are using VRay, Corona, Arnold etc mostly on the CPU. It’s probably a good approximation for fully-threaded sims too, again, if they’re running on the CPU.
So it has it’s place.


#5

Ignoring CPU rendering Cinebench doesn’t give the user any means to choose the best CPU and GPU for combination for working interactively, you can’t say for definite which is better them 5950x or the 12900K. The point of a benchmark is being able to determine which is better.

Multithreaded CPU rendering is a very poor measure of sim baking, the memory access profile is completely different. Sim baking is much more limited by memory bandwidth.

Cinebench used to be a good benchmark of typical C4D performance for hardware 15 years ago but it hasn’t moved with the times. When mid-low end GPUs can easily beat Threadrippers in rendering most professionals and hobbyists will be using GPUs to render.

C4D used to be almost entirely single threaded except when rendering, that has slowly changed over time. I believe the new Project Neutron scene graph is heavily optimised for modern multicore processors so it would make sense to build a Project Neutron benchmark scene into Cinebench so users can see expected interactive performance of future C4D. Blimey, it might even generate some interest in Scene Nodes along the way!! This type of benchmark would answer which modern multicore CPU and GPU combo is best for interactivity.

Maxon should take a leaf out of Blender’s book and create a CPU and GPU rendering benchmark, they now have Redshift on CPU so they should include Redshift rendering in Cinebench. They want people to rent Redshift GPU well create beautiful but taxing scenes to render, the images will end up in GPU review articles all over the interweb.

I’m pretty sure Cinebench wasn’t originally released out of the goodness of Maxon’s heart for all in sundry to test their CPU hardware, it was a clever marketing tool to get the C4D name out there. As it stand Cinebench has little relevance to anyone using modern C4D or thinking about using it, it’s just a benchmark of CPU rendering in a old dated and ugly renderer which fewer and fewer people are using.

Make Cinebench more relevant, add a kickass Scene Nodes project running on top of the high performance scene graph, show what C4D’s got under the hood. When Project Neutron was announced McGavran said it could handle millions of objects and offered next level performance, why not show it in one of the most ubiquitous benchmarks? Why would anyone not take advantage of Cinebench to market the modern state of C4D?


#6

It would certainly be cool for them to develop it into a more rounded benchmark suite, with more catagories.
I get the impression larger studios are still mostly using GPU just for lookdev, with CPU farms doing final renders ie Arnold etc.
But I don’t work in a large studio, so I might be behind the times in that impression.


#7

Yes you are quite correct large studio use CPU for rendering due to the enormous detailed scenes being rendered.

When I said ‘most professionals’ I was thinking of most freelancers and small outfits. CPU rendering is by far the most popular form of rendering at the high end and with those who have their own render farms. When you invest millions of dollars in a renderfarm you aren’t doing it on the basis of Cinebench though, bigger picture factors such as long term support contracts dominate these decisions. I would be surprised if most hollywood studio render farms weren’t Intel based despite AMD trumping them on performance.

Making Cinebench more relevant to C4D users and prospective C4D users should sure see it shift from a dated CPU bench to one which is far more in tune with contemporary hardware could be a far more effective marketing tool. Maxon is obviously moving to incorporate GPU acceleration within C4D to speed up the timeline and physics so why doesn’t Cinebench reflect that?

You said in the other thread Redshift CPU is 10x slower than the GPU version, Blender’s OpenData corroborates that, the fastest CPU based system 2x EPYC 64 core is matched by a 2060 Ti across various test scenes. When this level of performance is available to the more constrained budgets then one has to ask what really is the point of Cinebench in its current form?