Comments by "IIIRattleHeadIII" (@badass6300) on "The PC industry is in crisis... and can't continue this way..." video.

@halfnattyboomer354 No 7nm and 5nm have just not been big improvements over 12/16nm in terms of performance and GDDR6(X) has been terrible compared to any previous generation of GDDR memory, which is why GPUs are filled with tons of cache instead of stream processors/cuda cores. The gtx 1080TI was a slightly above mid range GPU that should have costed 450$ and people bought it at 700$+. Nvidia made 90-138% profit margins during the gtx 1000 series, which was up from the 40-50% during the gtx 900 series and that was up from the 30-40% during the gtx 700 series. Nvidia's stock skyrocketed in 2017 even before AI.
2
@halfnattyboomer354 High end is determined by the die size of the chip relative to the reticle limit. If the chip has a die size of 85-95% of the reticle limit then it's a high-end chip, a mid-range is around half of that +/- 5%. The reticle limit of 12/16nm was 851mm^2, the gtx 1080TI was a cut down titan XP(I think it was called) which was 471mm^2, so it's a bit above mid-range, but way below high-end. The RTX 2080 TI is high-end and 754mm^2 die size. The problem that generation is that AMD was stuck on Global Foundry's horrible 12/14nm process node that was on par with TSMC's 22nm... Also Nvidia skipped TSMC's 22nm and 20nm and went straight to 16nm. That's like going from the gtx 1080TI to the unreleased rtx 5090 and praising the rtx 5090 for being massively faster. I got 6 years out of a r9 285 4GB when gaming at 1080p.
1