The articles on this site are really informative. I always enjoy it when new ones pop up on Reddit.
What current graphics hardware does bilinear texture sampling at a slower rate than point/nearest sampling? I was under the impression that any GPU made in the past decade or so did both at the same speed unless you went really far back (like Permedia era), and you had to go to trilinear or aniso before you'd see extra clocks per texel.
Overall, we should measure this kind of things on a specific hardware before making any conclusions (for example, even bilinear seems to require a bit more of memory bandwidth, and if so - and if it is already close to bottleneck - it may backfire easily). Going into guessland: on PCs - probably you're right (except maybe in some corner cases), but if we take into account all the 3D-enabled mobiles out there - I don't think we're there yet. Just IMHO...
For PC, the graphics vendors outright tell you the speed tiers in their documentation for various formats, and for the a long time the standard for the big three was full speed for nearest/bilinear and half-speed penalties for trilinear or wide formats (64-bit). I also heard an ex-driver guy say that some hardware he worked with always had bilinear filtering hardware on and just forced the tap weights for point sampling, so there was no fetch bandwidth savings by using point. Very interested in hearing if people know of specific mobile GPUs that have this issue since that's less advanced than PC hardware.
There's a couple of outdated things in some of his other graphics, but not his network, articles too. I get the impression that game graphics is not his primary field of expertise. He does however have a good book here, apart from the stupid "I'm a middle aged British chap who escaped from the enterprise" logo.
I get the impression that game graphics is not his primary field of expertise.
Yep; one cannot be an expert in everything, and out of all the things, 3D graphics is probably my weakest point. :-( Accordingly, within the book, I'm trying to speak about graphics only to the extent which is needed to be referenced in other parts of the book (well, probably a bit more than that to keep the text flow more or less smooth). For example, I need to speak about meshes in the context of protocols - but to do it, I need to define "what the mesh is"; in the context of interest management - there is a question of frustum (which is AFAIK never used for interest management at least in currently existing games, but it CAN be achieved - and needs to be discussed), and so on.
