Recent content by Kaotik

Didn't notice the fine print? It has OEM/IP revenue excluded for NVIDIA

2 APUs, Phoenix and Phoenix2 are different chips. Also should RDNA "3.5" be counted as separate? That's coming to at least 3 APUs by the looks of it (Strix Point, smaller version of Strix Point and Strix Halo)

Unlike they claim, Weapon Debris isn't broken. It works on GTX 10 and earlier, it's Turing and newer (RTX and GTX 16) issue only.

On look, Ham is useless once again when he doesn't have a car seconds faster than the rest :rolleyes: (yes I realize it's irrelevant what the rest does since they've decided Verscrappen wins whatever the biggest sponsors want few years back)

Huh? Since when is "real chiplets" dictated by shader engines split between chiplets?

It used to be almost the norm. Everyone wanted 9500 Pro "L-version" because it could be flashed to 9700 (Pro) for example.

Something's wrong, it's Netflix yet Senna still looks like Senna.

I meant the pure physics side of it, bigger die means longer traces

The only differences that could really be are related to which units are disabled and which enabled and whatever possible latency penalties (probably nothing measurable in practice) it might result in if enabled/disabled units are stacked together vs spread out (which then could bring miniscule...

Intel has been using matrix accelerators ("AI" "tensor cores", in Intels case "XMX cores") since day 1 edit: or did you mean frame generation instead? That's something Intel hasn't released yet.

Not all GPUs support compute if you go back far enough!

Yep, IO scaling hit a wall first and now SRAM scaling hit a wall at N3. Logic still scales but that's it.

Just to nitpick, it's "off-computex" as in not official part of Computex and a day early

That's GeForce RTX problem that has apparently been around always according to PCGamer...

According to Korean media AMD has struck a deal worth over 2.8 billion euros with Samsung for HBM3e memories