If I don't know what happens at the smallest scale and how they are connected, I don't understand it. But always asking why something happens helps me break down complex topics. Thnx for the support 😉
Really nice didactic. You are better than most teachers. But 1080p in 2024 isn't acceptable, Holt that compression for the VRAM on TH-cam we don't pay for the size lol
Thanks for the words of support, I appreciate that 😃 If you are talking about 1080p for the texture, I think it can really depend, since 4K textures for diffuse are definitely important, but for height or bump it can be an overkill, since too much VRAM is used but the details may not be as visible. Or are you talking about 1080p for the video and YT compression, because that's a whole other story 😃
That is an excellent suggestion, thanks for your contribution 😉. Perhaps, someone can take this challenge, since I am not that profficient in either to offer good insight.
Do you mean "what" like, I did not know this or "what" as in what are you talking about, this is false :D I have a 4TB external hard drive for storing things, I since they are the cheapest and the slowest to tranfer data, they can pack a really punch when it comes to capacity :D
@@markitekta2766 he could mean as in who owns it and how expensive is it... Always remember, especially if you're American, everything you own is less common and more expensive for people around the world. Among the people I know I am a storage king, but I have between my 2 computers like uh... 3.5 TB of HDD and SSD, and that's lumping everything together. My friend has a 500gb external HDD but less internal storage than me. Now that's just my personal context, zoom out and buying external HDD here is insanely expensive, zoom out more there is no one I know from my online list of friends and gamers who own large storage like this, there is one but he runs a website and gets donation support to help.
If you embrace the full virtualized pipeline, you should use virtual shadow maps, virtual textures and virtual geometry (nanite), that means using full geometry assets and opaque materials, which means a radical change in any 3D related workflow for real time rendering. After you embrace such big change, full virtualized pipeline will give you better performance and visual fidelity. Nice video though.
Thanks for sharing your take on it, it is pretty interesting 😃 It is true that only opaque materials can go through the deferred shading or lighting but it's nice to have a different set of tools for different occasions, like virtualizing large textures that should be used for large opaque geometry assets, like open world spaces. Maybe the entire pipeline can be virtualized, but not for all approaches, right?
ok... but how it actually works? how can you stream a texture? how do you create and use sticky note texture? this vid create more questions than giving answers
Thanks for watching the video and for your thoughtful questions! 😃 I wanted to clarify the theory behind virtual textures, but you're right-applying this concept in practice can vary depending on the software and tools you use. SO, to try and sum it up briefly Streaming a texture: This involves dividing large textures into smaller "tiles" and loading only the visible ones into VRAM as needed, rather than loading the entire texture at once. It's like loading just the relevant pieces of a map while navigating through it. Sticky note texture: This metaphor refers to how the system keeps track of which texture tiles (or mipmaps) are loaded and where they belong, using pointers or lookup tables for efficient access. If you're interested in exploring this further, I'd recommend checking out the subchapter on Renedring Large Scene, page 866 in the Real-time Rendering book Fourth Edition by Akenine-Möller, Haines, and Hoffman Lastly, practical implementation will depend on the rendering engine you're working with-Unreal Engine, Unity, etc., each have their own methods and documentation. Thanks again for your curiosity! This helps me improve and refine the content for future videos. Let me know if you'd like me to cover practical examples in another video! 👍
but gold isnt that wrong? SSDs are used with directstorage (actually its sampler feedback), GPU and CPU can be used at the same time, so you cannot add that up... I know that video may be for dummys but its still somewhat missleading
@@PrefoX Thnx for sharing your thoughts. 😃 I'm not sure what part you are referencing here in particular, but I'm guessing the part at the start where we are talking about access time. My intetion was to have a shared ground and point for comparisson, so we can get a better grip of why we need to use DRAM or VRAM and why we need to diminish the moemory consumption there. Did you have anything specific in mind here?
You're welcome, and thank you for watching! 😁 Next video on virtual geometry system like Nanite is available here th-cam.com/video/RRKCqmctxLs/w-d-xo.html
You break down complex ideas really well. This is great
If I don't know what happens at the smallest scale and how they are connected, I don't understand it. But always asking why something happens helps me break down complex topics.
Thnx for the support 😉
Really nice didactic. You are better than most teachers.
But 1080p in 2024 isn't acceptable, Holt that compression for the VRAM on TH-cam we don't pay for the size lol
Thanks for the words of support, I appreciate that 😃
If you are talking about 1080p for the texture, I think it can really depend, since 4K textures for diffuse are definitely important, but for height or bump it can be an overkill, since too much VRAM is used but the details may not be as visible.
Or are you talking about 1080p for the video and YT compression, because that's a whole other story 😃
Brilliant content, dude!
I'm glad you found value in it, thank you 😄
Excellent video. Waiting for the next one!
It's being research and structured at the moment, but until then I'll do a follow up on Nanite in practice 😃
Great videos man
Thnx for commenting and supporting 😉
Pls do more performance daddy! A unity & unreal tutorial to do it practically
That is an excellent suggestion, thanks for your contribution 😉.
Perhaps, someone can take this challenge, since I am not that profficient in either to offer good insight.
4TB hard drives are usual? what?
Do you mean "what" like, I did not know this or "what" as in what are you talking about, this is false :D
I have a 4TB external hard drive for storing things, I since they are the cheapest and the slowest to tranfer data, they can pack a really punch when it comes to capacity :D
@@markitekta2766 he could mean as in who owns it and how expensive is it...
Always remember, especially if you're American, everything you own is less common and more expensive for people around the world.
Among the people I know I am a storage king, but I have between my 2 computers like uh... 3.5 TB of HDD and SSD, and that's lumping everything together.
My friend has a 500gb external HDD but less internal storage than me.
Now that's just my personal context, zoom out and buying external HDD here is insanely expensive, zoom out more there is no one I know from my online list of friends and gamers who own large storage like this, there is one but he runs a website and gets donation support to help.
they are like $40 lol
@@top.of.reddit Yup, pretty cheap 😃
Americans only account for America not the rest of the world where 1TB would be the usual for HDD and 500GB for SSD.
If you embrace the full virtualized pipeline, you should use virtual shadow maps, virtual textures and virtual geometry (nanite), that means using full geometry assets and opaque materials, which means a radical change in any 3D related workflow for real time rendering.
After you embrace such big change, full virtualized pipeline will give you better performance and visual fidelity.
Nice video though.
Thanks for sharing your take on it, it is pretty interesting 😃
It is true that only opaque materials can go through the deferred shading or lighting but it's nice to have a different set of tools for different occasions, like virtualizing large textures that should be used for large opaque geometry assets, like open world spaces.
Maybe the entire pipeline can be virtualized, but not for all approaches, right?
@@markitekta2766 i think the engine offers a wide range of tools, you would not want to hammer a nail with scissors and viceversa.
@@MarioCola I agree, the diversity of tools and the knowledge and experience of how they work help in achieving optimal results
ok... but how it actually works? how can you stream a texture? how do you create and use sticky note texture? this vid create more questions than giving answers
Thanks for watching the video and for your thoughtful questions! 😃
I wanted to clarify the theory behind virtual textures, but you're right-applying this concept in practice can vary depending on the software and tools you use.
SO, to try and sum it up briefly
Streaming a texture: This involves dividing large textures into smaller "tiles" and loading only the visible ones into VRAM as needed, rather than loading the entire texture at once. It's like loading just the relevant pieces of a map while navigating through it.
Sticky note texture: This metaphor refers to how the system keeps track of which texture tiles (or mipmaps) are loaded and where they belong, using pointers or lookup tables for efficient access.
If you're interested in exploring this further, I'd recommend checking out the subchapter on Renedring Large Scene, page 866 in the Real-time Rendering book Fourth Edition
by Akenine-Möller, Haines, and Hoffman
Lastly, practical implementation will depend on the rendering engine you're working with-Unreal Engine, Unity, etc., each have their own methods and documentation.
Thanks again for your curiosity! This helps me improve and refine the content for future videos. Let me know if you'd like me to cover practical examples in another video! 👍
This is gold
Thank you for the support, I really tried to make it comprehensive and easy to digest 😃
but gold isnt that wrong? SSDs are used with directstorage (actually its sampler feedback), GPU and CPU can be used at the same time, so you cannot add that up... I know that video may be for dummys but its still somewhat missleading
@@PrefoX Thnx for sharing your thoughts. 😃 I'm not sure what part you are referencing here in particular, but I'm guessing the part at the start where we are talking about access time. My intetion was to have a shared ground and point for comparisson, so we can get a better grip of why we need to use DRAM or VRAM and why we need to diminish the moemory consumption there. Did you have anything specific in mind here?
I cant believe you have 700 followers. This needs to change 🙏🏼
Thank you very much, I appreciate the support ;-)
@markitekta2766 no problem. I am learning a lot from your videos
Thanks for content (=
Wait for new videos
You're welcome, and thank you for watching! 😁 Next video on virtual geometry system like Nanite is available here th-cam.com/video/RRKCqmctxLs/w-d-xo.html
my head is smoking
😃 Don't get too knowledgeable in a short amout of time
"How to render on a buget"
...ignore the D 😂
😃 thanks for pointing it out, that could have been a nice pun - rendering on a baguette, keeping up with the story 😆 .