I read this status report on WiAB. There are solutions to this problem. -
He can use two RTX4090s together and get the same 48 GB (rendering software can handle multiple graphics cards in one system and doesn't need any additional technology). It will still be cheaper - $3000 for 2xRTX4090 vs $5000-6000 for one Ada RTX6000. Of course, the power consumption will be much higher, but the savings when buying + fast rendering in my opinion are more profitable pluses.
- Scene optimization. Reducing the resolution of textures for objects in the background plus hiding objects that are out of view. I do this in DAZ, because I have weak hardware, I think there are similar utilities for Blender, which he uses. He uses background blur in most scenes, in which case I see no reason to make these objects in high resolution.
4090s do not share VRAM if put together. It doesn't matter if you use 1, 2, 4, or 10. It's still just 24GB. If it was that easy to stack VRAM I'd never go for a Professional card.
For some reason, people think I don't know how to optimize scenes... I do everything to keep the VRAM and calculations as low as possible.
Don't listen to some artists/devs that render on SubD1 tell you that I don't know how to optimize properly.
I've been doing this for years and started with a 1060 6GB, I've gone through VRAM hell before SG was even out and know how to make it work.
I know of all the tools to do it, be it scene optimizer which I used back in 2019 for the old WiAB 0.2, and learned to never ever use it on a char, hide unused assets, lower res, or be mindful about the number of lights and caustics.
I know you meant no offense, but I've been getting kinda sick of hearing it.
I'm getting better and better cards to not have to compromise on quality.
Just some items to consider - ultimately people can do whatever they want; I don't want it to come across like I'm saying they can't do what they choose to, just putting my views on it out there.
I know if I was a creator, I'd put a dedicated 30A circuit in for my rendering rig to run off of in a well-ventilated room and design the workstation to be able to hold (4) Quadro cards.
Sadly, the A6000 ADA has a lot less value than the old A6000.
And it comes down to a single thing... They removed the ability for the A6000 ADA to pool the VRAM to 96GB.
And this was the whole reason why I really wanted them... And I have done a lot of research to find if it still works via PCI-E (Which some claimed, others denied) and I doubt it does.
Meanwhile, the Ada costs 9500€, while the old one cost ~ 5000€ with the ability to scale up. (For the US people, keep in mind I live in Germany and GPUs are usually much more expensive.)
75% of the scenes I can render on 24GB.
The other 25% need more milk. (I also want cards to be future-proof for Blender.)
But in the past weeks, I came up with a better plan... I'm gonna talk about it in the next Dev Log.
Ocean did about 2/3 of the rework for CH 2 in the most recent release (CH 4.5 Final). There is a significant difference when you bump from the reworked CH 2 content to the old CH 2 content.
The rework will be completed with CH 5 Beta or Final (not sure about the upcoming Steam release).
It will be completed with the Full.
