Intel Corp (INTC)

[mas]

Our site used PA-RISC for over a decade with the last model being the 1 GHz PA-8800 in a Superdome which was only replaced by a Itanium (Tukwila) Superdome 2 in 2011-2012.

Enterprise server chips do die eventually but they take a long time over it as software migration testing is often the limiting factor as every crucial line of all your critical systems have to be checked in anger first to make sure everything will work as before when you switch over your systems and as I have found bugs in illustrious software such as Oracle before this is not always a given. You will still find Itaniums in production use a decade from now I bet. Ultimately as long the server chip handles your workload you are more concerned about its reliability, accuracy and serviceability - RAS.

Mainframe software and hardware used to be a lot more robust in my experience but then again Mainframe ecosystems were more expensive than RISC Server ones so you get what you pay for I guess with ARM tablets now at the bottom of that computing robustness/reliability hierarchy with x86 in the middle to medium-high now.