Once again we see the classic PR hype machine tactic of comparing a newer chip which is only available as an engineering sample to other chip designs which are widely available and much older.

They also fawn over the chip’s TDP when all other chips have to support 16 bit floating point and thus must run much hotter.

They make the classic mistake of equating max TDP with in-use-watts, and praise this magnificent (fictitious) performance per watt at FP8 with other chips’ max-TDP at FP16, which draw twice the power.

Evidence that the IPO can’t be far away.