Pscht, dieses "Konzept" durchbricht aber die Jammertiraden ;-)jube schrieb:Niemand wird dazu gezwungen, die Karten zu kaufen
Und wir wissen erst seit Monaten, dass das nur 16 Gb VRAM werden. Die MIR auch zu wenig wären. Also zwingt mich Nvidia die 5090 zu kaufen, merke ich gerade. ASUS, bringt die 5090 ASTRAL ran!
![Breites Grinsen :D :D](/forum/styles/smilies/biggrin.gif)
Aber nochmal im ernst, bezüglich des Whitepapers:
Die 5080 hat fast den Durchsatz, den die 4090 hatte aber 16 GB Limit.The GeForce RTX 5090 ships with 28 Gbps GDDR7 memory and delivers 1.792 TB/sec peak
memory bandwidth, while the GeForce RTX 5080 ships with 30 Gbps GDDR7 memory, delivering
960 GB/sec of peak memory bandwidth.
Das Thema Texturen Kompression steht erst am Anfang, Nvidia hat in den Marketingseiten folgendes:
We've integrated neural networks inside of programmable shaders to create neural shaders. RTX Neural Shaders will drive the next decade of graphics innovations. They can be used to compress textures by up to 7X, saving massive amounts of graphics memory. And be used to create cinematic-quality textures, and even more advanced lighting effects in games.
Im Whitepaper finde ich auf Seite 40:
NVIDIA has worked with Microsoft to create the new
Cooperative Vectors API. When combined with
differentiable shading language features in
Slang, Cooperative Vectors unlock the ability for
game developers to use neural techniques in
their games including neural texture
compression, that provides up to seven-to-one
VRAM compression over block compressed
formats, and other techniques such as RTX
Neural Materials, Neural Radiance Cache, RTX
Skin, and RTX Neural Faces
Es mag sein, dass "demnächst" das VRAM Problem bald keines mehr in der "Form" wie bisher ist.
Nach AI Frames kommt AI Ram, und ich will das (tm) darauf hier im Forum Zwinkersmiley