Really a pity that Oracle killed off SPARC. They already had 32-core CPUs almost a decade ago but Oracle never really understood the value that SPARC and Solaris brought to the table.
Solaris seemed to be much more solid than linux overall.
Solaris 10 was intended as a catch-up release but the package manager took years to mature. I remember getting racks of systems in the mid-2000s where simply running the updater on brand new servers rendered them unbootable. By then the writing was on the wall.
SPARC was similar: the hardware had some neat aspects but what they really needed was the equivalent of a first-class LLVM backend now. Statistically nobody was hand-tuning assembly code to try to get it to be competitive with x86, especially since that meant not only your own code but most other libraries. The reason ARM is doing so well now is that two decades where the most popular computing devices are phones means that you can just assume that anything popular will compile and run well on ARM.
Like, to update cron. Even though through 3 versions of the same patch id, it does and does not fix cron.
Pure insanity with no product focus on user experience.
[1] https://azure.microsoft.com/en-us/blog/azure-virtual-machine...
Nearly all programs have a constant memory usage regardless of how fast the core running them is. If you deliberately make your cores half the speed, and double the amount of them, you have approximately doubled the cost of the memory you need. In aggregate, memory already costs so much more than CPUs that this is rarely if ever useful, even if it meant your cpu is free.
The starting point of a design that tiles a lot of cores just needs to be ones of the fastest cores available, or it is not commercially viable.