I honestly haven't seen it. I'll go watch it but it seems like a misunderstanding on your part because the SSD wouldn't know what assets to ignore or cull and neither would the software programmatically until it's processing in the rendering pipeline. The SSD doesn't and cannot possibly have that type of logic there before rendering.
I'm looking at the transcript now is there's nothing, verbatim on your quote, on " entire game logic" unless you are paraphrasing and not quoting? No mention of the word "logic"
You mean the part about the player in game " turning around"(Hence your comment on turning)? He's talking about loading the assets, the textures really fast via the ssd( I quote "possible to load textures for everything behind the player in that split second"), not any type of culling. He's talking about what SSDs do....read and write, which it does fast like most SSDs, nothing to do with game logic or the rendering pipeline.
www.videogamer.com
Just read it, and yes it looks like a misunderstanding on your part:
Lysandros
It's a misunderstanding compared to what you ORIGINALLY said, you were describing something like culling , I quote from you:
Sure okay friend, my bad, no need to write more. You convinced me that the technique described by Cerny is ancient, inefficient thus redundant compared to the velocity architecture. Thanks for your replies, have a good day.
"the GPU will have to render only what is in the player's/character's vision right? Rendering only 120 degree of game logic, geometry, assets etc. out of 360, thus saving significant amount of RAM, bandwidth and cycles. I have hard time seeing how this doesn't improve performance. "
Now you're changing your phrasing to plain ol loading from the device...loading texures. You are now saying:
"I was not talking about Geometry Engine or Mesh Shader culling or even fustrum culling, did you watch Road to PS5? Cerny talks about the 'entire game logic' + assets being loaded and discarded (around 4-5GB of data) at 0.5 second as the player/character turns. This technique is enabled by the very fast/low latency, custom SSD/IO complex. "
Now this texture loading from the storage device, albeit fast(for console just getting SSDs now), is ancient. Now if you add in the custom hardware for Sampler Feedback Streaming and involve the XSX SSD with Xbox Velocity Architecture, NOW we have NEW AND something EFFICIENT....streams only the part of the texture it(not just a mipmap) that needs to....now THAT'S efficiency!
That’s still just cerny saying how great it is. It’s nothing that Dolby Atmos already doesn’t do. Sony don’t want to license Atmos so they just made their own of the same thing. There’s nothing groundbreaking about 3D audio or how the PS5 is doing it. The Series S/X have the same thing. Windows Sonic is 3D audio. Dolby Atmos is 3D audio. The Xbox supports both. The Xbox one supported both. The Xbox has HRTF. The Xbox has hundred of sound sources in 3D. There’s nothing special about tempest, it’s just Sony’s name for 3D audio.
His job is to make the ps5 that he designed sound like it’s the best at everything and that it has no weaknesses. Everything the other console has more of is actually not important because the ps5 has such and such. He’s selling his console. He’s not unbiased. He’s pr.
yeh he put an apu with some ram and an HDD. There was nothing exceptional about the PS4 design. The XB1 would have been pretty much exactly the same has GDDR5 been guaranteed to be available for 8GB of chips.
Now this texture loading from the storage device, albeit fast(for console just getting SSDs now), is ancient. Now if you add in the custom hardware for Sampler Feedback Streaming and involve the XSX SSD with Xbox Velocity Architecture, NOW we have NEW AND something EFFICIENT....streams only the part of the texture it(not just a mipmap) that needs to....now THAT'S efficiency!
Let me ask you a hypothetical question. I'm not going to use the actual numbers too to simplify things. Let's say, hypothetically, that texture streaming on the ps5 is 10Gb/s and the XSX is 5Gb/s. Lets say we have a 10Gb texture they both want to stream. For simplification, the ps5 streams that into RAM in 1 sec(NOTE: From here the Ram bandwidth for textures is higher on the XSX btw). Now this gets more interesting with the XSX, with the custom Sampler Feedback Streaming HARDWARE, unique on the XSX, it streams in only the part of the texture it needs. Let's say it needs just 50percent or 5Gb of said 10Gb texture. So it ends up streaming in only what it needs to RAM in the XSX in 1 second. So let me ask you, in only needing to load in the part of the texture it needs on the XSX, is the XSX with XBVA and SFS 1)more, 2)the same or 3)less efficient, efficient is the word we've been using, in streaming in textures to RAM than the ps5 from their respective SSDs?
Because SFS is custom featured hardware on the XSX.
I understand bottlenecks in a system...overall that's not what we were talking about....we were discussing efficiency in texture streaming from storage vs brute force. The Ps5 is likely faster, but from what we know, in that particular aspect, texture streaming, it is less efficient.......THAT, efficiency, was the conversation.
Yes and he said (how do I not know that ps5 has something like or similar to patented custom SFS hardware?)....hence the analogy. I hope the analogy helped to point out why I was skeptical about the ps5 having that same type of custom hardware and we don't know about it?...I apologize if it wasn't a perfect analogy
