What was amazing about the 8800 GTX is that for years Nvidia was downplaying the importance of unified shaders while AMD was pushing hard in this direction. And then boom! They took everyone by surprise and delivered a card with unified shaders and on top of that it was so powerful. Masterful move!

In the end the N64 sold a little bit above 2M units in its first 12 months of availability in Japan.

In the same spirit on macOS for years I've been dragging files to the terminal to copy their path and I've just learned that by pressing option and right clicking on a file there is a 'copy path' option

I'm sure you know all of that and just didn't want to add more complexity to an already confusing topic.

As you explain Re-Z is not only less efficient (because you defer ZWrite) and it also has some overhead of performing the Z test twice so it is only enabled when the shader is sufficiently complex. Some public info about Re-Z here www.x.org/docs/AMD/old...

Great explanation! Just a quick note, Early-Z is "always" disabled with depth write what you are describing is what AMD calls Re-Z. Basically Early-Z is an atomic ZRead/ZComp/ZWrite while Re-Z is just ZRead/Zcomp and deferring ZWrite to after the fragment shader.

À Bordeaux 😊

It's funny to see that from a raw specs perspective, the just revealed 5080 is not THAT far from the 3090 Ti. Of course, there are several architectural improvements and a huge increase in clock speed.

Don't pay too much attention to these people who just need something to be angry at. The silent majority loved what you've shown. Keep up the good work, the trailer was awesome.

Nice! How does it compare performance wise with per pixel sorting (fixed size kBuffer or Per Pixel Linked list ABuffer).