There are still bright spots, but the space is so much larger it’s hard to find them. I still love all the companies that present at SC and Hot Chips.
One I like in particular is Cerebras. Wafer scale didn’t work when wafers were 3” but, somehow, they managed to make it work when they are the size of a manhole cover.
BYTE would run entire issues on the exploits of IBM in POWER and Z, on Intel and AMD’s latest tricks, and on the latest 250TB 2.5” SSDs.
Yes, Cerebras is very cool. But that sort of thing is also becoming the exception. I use the homepage of HN as a sort of thermometer, the trend seems to be towards 'more' rather than 'better'.
Same here. HN is an excellent way to take the pulse of our industry.
This "more over better" is a natural trend - a lot of money can be made for a relatively low effort. Cerebras did the Really Hard Thing, something many other brilliant companies and people failed at for decades. I remember Gene Amdahl failing at wafer-scale back when wafers were tiny.
One I like in particular is Cerebras. Wafer scale didn’t work when wafers were 3” but, somehow, they managed to make it work when they are the size of a manhole cover.
BYTE would run entire issues on the exploits of IBM in POWER and Z, on Intel and AMD’s latest tricks, and on the latest 250TB 2.5” SSDs.