The Spirits Within (2001) was completed over a period of four-years, during which approximately 200 people put in a combined 120 years of work on it. A render farm of 960 workstations was tasked with rendering each of the film's 141,964 frames. Think about the processing power we have now in comparison.
These videos inspire me to keep going so I can afford something ridiculously fast like this one day. I can only imagine the type of ideas I'd come up just because I can render it with ease.
Now with this I feel I got a potato/bread toaster PC with my 3090 and 64 GB of RAM😂 Jesus lord this is a dream come true for renderes and super hardcore designers. Cant wait to see the next gen of the A6000 with the Blackwell arquitecture
Not a hot take, but was expecting a £90K machine to render faster. Remember when Nvidia/Epic showed off the storm trooper Reflections demo on a $60K machine, and a few months later it was like "it runs on a single 2080ti now lol"? They just keep introducing new hardware/software combos that destroy their previous gens, which is great for us but not always great for people who spend like a $100K on a single machine...
This is bonkers. Would love to see a test with Blender as well. Also, I wonder how it'd be with something like Embergen as it primarily works with GPU only. Also, octane has bench marking tool and the new Cinebench also has redshift built in.
We could honestly have 1080p or 1440p photorealistic graphics with a laptop gpu right now, if optimized correctly in terms of lighting and poly count, not to mention nanite&lumen. Why do we push for pixel count instead of light/mesh quality, is beyond me. This just feels like justification for exorbitant gpu prices...
My understanding was that Omniverse Interactive Path Tracing was competitive with Unreal Lumen in terms of quality vs. render speed. Is that the case with the right settings? I was considering using Omniverse as an alternative to rendering with Cycles, but a beefy setup like the one shown here would achieve similar results in Cycles as your other video demonstrates, so not sure what the benefit of Omniverse is vs. just rendering in Cycles.
Thanks, makes sense. Guess Lumen is still the undisputed king of path traced quality GI rendered in seconds per frame instead of minutes / hours (without a monster of a machine). Omniverse looks interesting, but RTX render mode GI is crap from what I’ve seen, and I was really hoping the path traced mode would be a real innovation in terms or render speed, but guess it’s no better than Cycles. EEVEE Next GI looks disappointing as well from what I’ve seen. Damn you, Epic!
I think its a misconception that these are needed to render a high quality CGI feature. Experienced CGI/compositing teams carefully plan out what needs to be rendered in scene to determine what minimal samples rates are required for each layer. Some layers/scenes that are heavily stylized might not need Path Tracing to look convincing. You could be wasting a ton of processing headroom due to rendering samples that are not necessary. I'm sure they have all kinds of cheating tricks in major studios and that's what really saves them time and money. I'm sure they hate it when everything in the scene is green screen and requires CGI lol. Lazy directors and production teams are what cost the most because they use CGI to compensate for simple things that can be done practically.
$100k PC seems like a massive waste of money, considering gpu tech will catch up relatively quickly and make the investment look pretty poor in a few years
get your wallets ready. P.S. Tried it in Unreal engine 5.3 and 5.4. Unreal crashes. I guess multigpu is not a priority for them 😢
The Spirits Within (2001) was completed over a period of four-years, during which approximately 200 people put in a combined 120 years of work on it. A render farm of 960 workstations was tasked with rendering each of the film's 141,964 frames. Think about the processing power we have now in comparison.
exactly dude exactly
These videos inspire me to keep going so I can afford something ridiculously fast like this one day. I can only imagine the type of ideas I'd come up just because I can render it with ease.
same man rendering kills momentum tbh wish we can just focus on creating vs waiting life is too short
Now with this I feel I got a potato/bread toaster PC with my 3090 and 64 GB of RAM😂
Jesus lord this is a dream come true for renderes and super hardcore designers. Cant wait to see the next gen of the A6000 with the Blackwell arquitecture
yea man pathtacers dream rig also did one in blender
It will be super cool to see how Blender handles 8x A6000. I got a couple of heavy Blender files/projects I can share to you if you like. @@Jsfilmz
Hardcore designers? I wasn't aware designers needed render farms made for extreme cases of CGI/Compositing.
Not a hot take, but was expecting a £90K machine to render faster. Remember when Nvidia/Epic showed off the storm trooper Reflections demo on a $60K machine, and a few months later it was like "it runs on a single 2080ti now lol"? They just keep introducing new hardware/software combos that destroy their previous gens, which is great for us but not always great for people who spend like a $100K on a single machine...
oh dude 4k real pathtracing 2048 spp is very fast its insane speed tbh. my 4090 would die if i try
Well it actually depends a lot on the renderer, samples etc.
This is bonkers. Would love to see a test with Blender as well.
Also, I wonder how it'd be with something like Embergen as it primarily works with GPU only. Also, octane has bench marking tool and the new Cinebench also has redshift built in.
ill try today mang. cinebench doesnt support miultigpu it seems i did an octane one though
@@Jsfilmz maxon can't get anything done
We could honestly have 1080p or 1440p photorealistic graphics with a laptop gpu right now, if optimized correctly in terms of lighting and poly count, not to mention nanite&lumen. Why do we push for pixel count instead of light/mesh quality, is beyond me. This just feels like justification for exorbitant gpu prices...
More killer science. Thank you for your findings @jsfilmz !
Only 7 hours. Crazy! Thanks for the overview
yea blender coming too too bad unreal whack with multigpu
We have all this tech yet we don’t have any new 3D anime movies .. like I wonder what final fantasy spirits would look like with all the new tech.
I mean it was a static shot. No explosions or rain .. etc going on
50k usd in GPUs. I would love a pair with nvlink. That NVidia has killed nvlink on workstation and gaming cards really sucks.
What about the new Nvidia remote workstation service RTX vWS?
Everything looks crisp in this video except your chair which looks like it have a low res texture on it 🤣
its old lol
My understanding was that Omniverse Interactive Path Tracing was competitive with Unreal Lumen in terms of quality vs. render speed. Is that the case with the right settings? I was considering using Omniverse as an alternative to rendering with Cycles, but a beefy setup like the one shown here would achieve similar results in Cycles as your other video demonstrates, so not sure what the benefit of Omniverse is vs. just rendering in Cycles.
nah path tracing in omni is pathrracing mang lumen is hybrid raytracer yea blender users are so lucky
Thanks, makes sense. Guess Lumen is still the undisputed king of path traced quality GI rendered in seconds per frame instead of minutes / hours (without a monster of a machine). Omniverse looks interesting, but RTX render mode GI is crap from what I’ve seen, and I was really hoping the path traced mode would be a real innovation in terms or render speed, but guess it’s no better than Cycles. EEVEE Next GI looks disappointing as well from what I’ve seen. Damn you, Epic!
Can you do a 4090 vs 6000 Ada comparison for the same image
i cant my cpu is butt on my 4090 pc its 3 years old
Only the grass has some uncanny valley
The RTX 6000 ADA is not a gaming card. RTX 6000 ADA for workstation VFX Movie, Graphics Designer, Adobe all, Unreal Engine, 3D Max, more.
Ummmm.. Duh?
I thought RTX 6000 ADA doesn't support nvlink.... how are they able jam so many GPUs?
Supermicro makes server cases and servers that will let you mount 8 GPUS. This is running through the PCIe bus sadly.
Hey!
I think its a misconception that these are needed to render a high quality CGI feature. Experienced CGI/compositing teams carefully plan out what needs to be rendered in scene to determine what minimal samples rates are required for each layer. Some layers/scenes that are heavily stylized might not need Path Tracing to look convincing. You could be wasting a ton of processing headroom due to rendering samples that are not necessary. I'm sure they have all kinds of cheating tricks in major studios and that's what really saves them time and money. I'm sure they hate it when everything in the scene is green screen and requires CGI lol. Lazy directors and production teams are what cost the most because they use CGI to compensate for simple things that can be done practically.
$100k PC seems like a massive waste of money, considering gpu tech will catch up relatively quickly and make the investment look pretty poor in a few years
I can only image what kind of machines Marvel Studios own.
they render with farms so just multiple weaker ones vs one i think
Great googly moogly!! How???
carefully 😂
Can it run Crysis?
Exactly