Just a thought:
I’m a gamer and I’ve been testing game-streaming services like Google Stadia and Nvidia GForce Now the past few weeks.
For those unaware of what these do: the actual game code is run on high-spec servers with huge GPU arrays “in the cloud” and the resulting video/sound is streamed to your client at home.
I have to say, I’m pretty impressed: I consistently get 50-60fps and input lag is negligible.
The obvious advantage is not needing to invest in a PC rig capable of running these games on high graphics settings.
This got me thinking: it would be perfectly possible to run Blender in the same way and stream the user interface to a client. This would enable an ordinary user with low-end hardware to run Blender and render projects on high-spec hardware. I’d be interested in this type of service.
For something like this to work there needs to be solution for projects with sensitive data. You have to send your data to some other companies computer and a lot of customers will not be amused if you do that.
I already use Blender like that a lot on Paperspace. Here is a previous thread about it: *Streaming* Blender
Also, look into Xesktop, you’ll get the same but on steroids with a 8 or 10 GPU cloud pc.
The Foundry used to offer an Application cloud streaming service, Athera they even included Blender in the list of Apps they offered. It was the most expensive way ever of using Blender as the service was not cheap!
Cloud based Applications have been used in VFX studios for a while, it allows them to scale up and down as the work requires without investing in a lot of hardware that could potentially go unused a lot of the time.
I’m not at the state of play with these services, Athera obviously failed to gain traction, but I’m sure more of this is on the way for more mainstream markets.
Thanks for your reply, this was a valuable tip. I’ve used Blender on a Paperspace VM over the weekend and I’m very impressed.
I started off a model locally and when I got to the heavy lifting I took it to the VM. There is a lot to be said for a pay-per-hour model like Paperspace vs investing in a local high end rig.
One caveat: I’m in Europe running a Mac and keyboard mapping in the native Paperspace app was a nightmare. Luckily they also support plain old RDP, in the Mac client I was able to map my keyboard correctly.
Sounds like in order for this to be worth it you would indeed need to do work that simply requires hardware so expensive as to be basically out of the question for the average user (e.g. terabytes of RAM at the cost of a decent car or unusually high amounts of video memory that you can only get with the highest end video cards).
Because in any other case the cost of even a high spec PC built from common components is surely going to be less than whatever bill you are racking up when working on your scenes for many hours over the course of weeks/months.
If you use it a lot Shadow can be a viable option with its relatively low flat fees (unfortunately they have recently postponed all RTX options to 2021).
I think the real game changer is gonna be when I am going to have a little ultraportable laptop, connect to a remote workstation on the cloud, open blender, access my project library on the cloud, work on the project, to some test renders, save, open 20 render nodes instances on the cloud, render, get the files on the cloud.
I love assembling hardware, but when you think of the scalability and costs benefit, I think having a physical workstation will be less and less appealing, or even feasible.
That is perfectly possible. I have a local folder on my machine mapped in my remote desktop client, which allows me to save my project files locally when working in the cloud.
So: I start modelling a project locally on my crummy Mac and when it gets too CPU/GPU intensive, I open the same project from my 16 core Xeon Quadro 4000 30Gb workstation in the cloud and take it from there.
So you’d install Drive sync on your remote workstation basically
On a general level, does anyone know if there’s a way to seamlessly work with a huge gdrive library withouth having to physically have it on a workstation, local or virtual?
: