I am running a CAD workstation with a TR 2970WX, 2x Radeon Vii, with 64gb of 3200mhzC16 RAM.
I got the Radeon Vii for Vectorworks 2020 for Architectural work for my startup firm, and large BIM models replacing a 1080ti which had problems with Vram overfilling and generally being unstable, its been a great replacement for this being able to open lots of files without crashing and being able to import massive non-optimised models which sometimes is unavoidable in Architecture, I bought a second card so that I can run Lumion 10 on a separate monitor and render scenes independently of the card running CAD. This setup works very well. I can now render a scene or a video of the model with one card maxed out with the other working freely doing other stuff.
However after a little more research I'm a little frustrated with the Radeon Vii performance in Lumion 10 compared to other equally priced cards, mainly the rendering performance.
Now there are two aspects to performance in Lumion.
1. The build mode FPS.
They recommend running only in 1080p as its a heavy program 1440p and 4k are not worth the effort even with a Titan RTX.
I'm getting 32 fps with the 3 star 100% resolution settings, a 1080 ti is around 42fps by comparison and a 2080ti is 60fps or just under. This isn't great for the Radeon, but its ok and perfectly usable to work in.
This is where the Radeon Vii really is getting let down. It's getting beaten by Nvidia 1080 and RTX 2070 cards, not even close to a 1080ti. The 2080ti has almost exactly double the performance in rendering. Radeon Vii is basically hovering around the performance of a 1070!
I just think there is someting seriously wrong here, and the cards performance is not being leveraged at all.
BTW they do a benchmark using one of the example house scenes as it seems to be the only way to measure performance properly the built in benchmark and passmark scores are irrelevant, benchmark process is below:
How To Compare:
1. Make sure that the monitor resolution to set to 1920x1080 on the screen that the Lumion window is on.
2. Open the Settings screen in Lumion and use the following settings:
- Editor Quality: 3 stars
- Enable high-quality terrain: On
- Enable high-quality trees: On
- Editor Resolution: 100%
3. Load the Example Scene - Villa Cabrera. Don't move the camera.
4. In Build Mode, move the mouse cursor to the white square in the top right corner. What is the 'FPS' number?
5. Go to Movie Mode and render Clip 1 to an .MP4 file. Render settings: 1920x1080, 5 star quality, 30 frames per second.
6. After it's rendered about 10 frames, hold down the CTRL key and let us know how many seconds it takes to render each frame. Alternatively, you can render the whole clip and note the total and per frame render times.
7. Repeat the render steps 5 and 6 with resolution at Ultra HD (4K) 3840x2160.
8. A comparative test in Photo Mode of Photo 1 at Poster 7680x4320 would also be great.
The benchmarks below give an idea, however there are no scientific benchmarks done on one system yet.
Radeon Vii :
5. 7.5 seconds per frame HD quality
7. 26.5 seconds per frame 4k quality
8. 3:30 poster quality render
5. 4.5 seconds per frame HD quality
7. 21 seconds per frame 4k quality
8. 2:10 poster quality render
5. 3.4 sec per frame HD
7. 14 sec per frame 4k
8. 1:30 poster render
I've discussed this with Lumion on their forum, and I'm going to keep asking, but by the sounds of it there stance is that it is simply what it is and they don't actually work on optimising for any cards and the results are what they are.
A little bit frustrating to be honest as I bought this card for the purpose of rendering. I am not massively bothered as I can render things now without affecting my other work. I.e. for the price of one 2080ti i have 2 Radeon Vii and more efficient workflow.
Is there anything that can be done about optimising in the settings from AMD's end?
Lumion 10 apparently uses the Direct X 11 API