OTOY Forums

This is kind of meaningless without a comparison to anything Nvidia

Good job improving render times
But it'd be more helpful to know how different Macs compare to various CUDA cards.
Is an M2 Pro or whatever as fast as... a 2080ti? 3050? 4020?? Who knows.

Hi,
Nvidia GPUs are still greatly faster than AMD or Apple Silicon GPUs, but the great advantage of M1/M2 chip is not in rendering speed, but in the incredible capacity to handle huge complex scenes without issues or slowdowns

I had to work in C4D on a huge complex scene with CAD components on PC Intel/Xeon 96GB of RAM and it was literally a nightmare with continues C4D slowdowns and memory errors!

Then I have moved the scene to mac mini M2 Pro 32GB, and it was a real pleasure, like working with a normal scene with great responsiveness and instant feedback

So the perfect combination is M1/M2 for preparing and editing the scene, and RNDR or Nvidia PC for rendering.

My two cents.

ciao,
Beppe

Hi,
for C4D 2023 users, please upgrade C4D to R2023.2, or c4doctane plugin cannot work correctly, please go to your MyMaxon/Downloads page and download the update.

ciao,
Beppe

bepeg4d wrote: So the perfect combination is M1/M2 for preparing and editing the scene, and RNDR or Nvidia PC for rendering.

100% agree. Add in headless rendering tech and you've got the ultimate 3D setup.

bepeg4d wrote: So the perfect combination is M1/M2 for preparing and editing the scene, and RNDR or Nvidia PC for rendering.

My two cents.

ciao,
Beppe

That's the ideal we are still hoping for, Beppe, but with network rendering. We finally had to move to PCs for our workstations last year because we just couldn't wait anymore. I had MacGyvered our cheese grater and trashcan Mac Pros as far as I could with various hardware and software hacks, and we have a bunch of network PCs for number crunching. It's amazing how long the Macs were still good enough for production work.

If/when network rendering happens, this blasted Windows machine is getting kicked headless out of my office and into the render room! I haven't had to work regularly on Windows since running Maya under Windows NT in the 90s.

bepeg4d wrote:Hi,
Nvidia GPUs are still greatly faster than AMD or Apple Silicon GPUs, but the great advantage of M1/M2 chip is not in rendering speed, but in the incredible capacity to handle huge complex scenes without issues or slowdowns

I had to work in C4D on a huge complex scene with CAD components on PC Intel/Xeon 96GB of RAM and it was literally a nightmare with continues C4D slowdowns and memory errors!

Then I have moved the scene to mac mini M2 Pro 32GB, and it was a real pleasure, like working with a normal scene with great responsiveness and instant feedback

So the perfect combination is M1/M2 for preparing and editing the scene, and RNDR or Nvidia PC for rendering.

My two cents.

ciao,
Beppe

Agree that working on the Macs would be great. We currently work with older Macs but would like to update them all to new Macs. We use PCs for main rendering (as a mini farm).

But that doesn't mean a comparison of current Mac speeds on latest build, would not be helpful. We would prefer to keep working on Mac, but need to know what kind of interactivity (if any) we would get.
As compared to Nvidia cards on older Macs (with CUDA drivers of course), which mostly means 1080Tis or 2080Tis as the last cards that were able to be used on older Macs.
How does an M1/M2 pro or Studio Ultra compare with 1x 1080Ti? or 2x 2080Ti? If it's worse, there would be no point updating all the Macs. But 2x2080Ti is enough to build scenes with comfortably, and look-dev, with main rendering happening on PC (wow wouldn't Network rendering from Apple Silicon to PC be great here?????? Yes, yes it would, probably years away I expect).

So, if someone could provide some actual benchmarks to compare Apple Silicons with Nvidia times on the same scenes, ideally the Octane Bench scenes, that would be really helpful for us to decide how to go about updating our workstations. Obviously dumping Mac entirely is the "raw power" solution but if we could get decent look-dev ability on a new Mac, we don't need the daily-driver workstation to be the hottest machines in the studio. Save those for the mass rendering.

I hope you understand what I am saying here, because every time I ask for some kind of update or comparison, everyone shouts "DID YOU KNOW NVIDIA WILL ALWAYS BE FASTER". Yes. Of course. My god, really? I had no idea. Gosh.
No, simply we need to understand the trade-offs between systems and some kind of cross-OS Benchmark would be REALLY appreciated from Otoy. But then, I can understand they don't want to publish the comparison, because it IS going to make Mac look weak. But it's so obvious now that Otoy are avoiding it. Just bite the bullet and publish an Apple Silicon version of Octane Bench please.

Hi northalex78,

we are not ready for an unified OctaneBench CUDA/Metal version, maybe when Apple Silicon M3 and OctaneRender 2023.1 stable will be out

Apple is relatively new to GPU rendering, and I still think that it is like comparing artichokes with strawberries

Nvidia GeForce GPUs are designed to get the best rendering speed whatever it costs, while Apple silicon GPU are designed to give the best rendering power per watt.

Probably we should need to compare only with Laptop or Quadro Nvidia models to be more fair

Anyway, as said, rendering speed is not everything, especially if Apple Silicon is faster in all the other tasks, and you have PCs for final rendering.

Probably the best test you can do is to rent a MacBook Pro M2 Max or Apple Studio M1 Ultra with macOS 13.3, and try to work with your scenes in OctaneX 2022.1.1.

ciao,
Beppe

bepeg4d wrote:Hi,
Nvidia GPUs are still greatly faster than AMD or Apple Silicon GPUs, but the great advantage of M1/M2 chip is not in rendering speed, but in the incredible capacity to handle huge complex scenes without issues or slowdowns

I had to work in C4D on a huge complex scene with CAD components on PC Intel/Xeon 96GB of RAM and it was literally a nightmare with continues C4D slowdowns and memory errors!

Then I have moved the scene to mac mini M2 Pro 32GB, and it was a real pleasure, like working with a normal scene with great responsiveness and instant feedback

So the perfect combination is M1/M2 for preparing and editing the scene, and RNDR or Nvidia PC for rendering.

My two cents.

ciao,
Beppe

Totally agree, same experience here! Unfortunately, we still cannot match Mac + PC with the same latest Octane + C4D version. Hopefully, the long away announced headless version isn't going to come out in a long time ahead. We really need a combination (and I am meaning network enabled) of Mac M1/2 version with matching PC based render slaves soon, even if in beta stage to get advantage of both worlds. Crossing fingers

3dworks wrote: Totally agree, same experience here! Unfortunately, we still cannot match Mac + PC with the same latest Octane + C4D version. Hopefully, the long away announced headless version isn't going to come out in a long time ahead. We really need a combination (and I am meaning network enabled) of Mac M1/2 version with matching PC based render slaves soon, even if in beta stage to get advantage of both worlds. Crossing fingers

Since Otoy is not ready to do it themselves, how about people with Apple hardware use the .orbx file that comes with OctaneBench and run their own tests in OctaneRenderStudio under Metal? Does that file open/render properly on the Mac? Unfortunately I don't have Octane compatible Metal hardware available to me, or I'd do it myself. It's certainly not as fair/controlled a test as OctaneBench itself would be, but all the scenes are right there in OctaneBench2020_1_5/benchmark_data/benchmark.orbx. If necessary, the Metal incompatible elements could be removed and the resulting standard used on both systems for informal testing.

Yes, those scenes have not been updated to account for Metal, and Octane has changed quite a bit since OB 2020.1.5, but at least it would give us some decent data to share.

I'll start with what I can. These are times generated by simply selecting the appropriate render node in OctaneRenderStudio and letting it rip. I'd be happy to re-run these if someone makes a more Metal compatible version.

1x stock NVIDIA RTX 4090
OctaneRenderStudio 2022.1.1
NVIDIA driver 31.0.15.2802

ATV Info: 1:05
ATV DL: 4:22
ATV PT: 5:05
Total 9:32

Box Info: 1:00
Box DL: 5:30
Box PT: 6:21
Total 12:51

Idea Info: 1:08
Idea DL: 4:25
Idea PT: 4:54
Total 9:29

Interior Info: 1:03
Interior DL: 4:02
Interior PT: 8:19
Total 13:24

Grand Total: 45:16

Clearly, OctaneBench is using lower settings than OctaneRenderStudio uses by default on the same data (100,000 Max samples ORS). When I run the suite in OB, the total run time is 7:06 with an OB score of 1241.58.

northalex78

You wrote that you can use a 2080ti in an older mac - is that true? I thought a 1080ti and High Sierra was the upper limit?

I still run a 1080ti in my Mac Pro, and still use it for quite a lot of jobs, even though I have a dual x 3090 PC sitting here! Just prefer working on the mac...

If I could use a 2080ti I'd get one