Feb. 21 at 1:14 AM
$GSIT
You must think people are stupid.
So your final answer is NO: there is no disclosure in the October 20 release that both the performance and 98% energy efficiency are based on a simulated HBM2e interface that the real chip, THE ONE THEY COULD ACTUALLY SELL, DOES NOT HAVE.
Let's not play dumb: if the real chip lacks HBM2e, it cannot perform as claimed in the PR or on your website.
People bought the stock believing that performance and 98% energy efficiency were real. They were never told these figures were based on a simulation. Cornell was very clear: without simulating that subsystem, the chip is utter crap.
In my estimation, they would need north of
$200 million (probably much more) and a complete redesign in at least 7nm process node, adding chiplet capacity and HBM2e, for a real chip to actually meet the simulated performance.
