Talk:IBM - Telum

In principle, IBM Telum is just a single chip (similar to the difference between the 13900k and Raptor Lake and Xeon). (Maybe) used Telu on the IBM Z16 series Instead of Telum using Z16, because Telum is similar to what was used on IBM LinuxONE Emperor 4, or rather a set of units (dual chip/socket drawer/drawer system). ）

Before Telum, the chip of IBM's old server was called IBM Power Number (Like Power8, Power9), similar to the previous generation of IBM Power z15 servers, and its chip had no name, so it could only be called Z15

Telum is a special generation, and in principle, Telum cannot be attributed to the original Z16 series Because Telum is not a z16, a brand new generation, but the Z16 uses Telum to see some shadows of power, but it is really not exactly power Because Telum favors its banks, it needs needs things like encryption.

It's a complicated thing, not the staff in this area.

age：Some may be fake news: sometimes z14 adds some units to the jumbo and becomes Z15, and adding some units to z15 becomes Z16, I doubt that, but maybe it's true, the added units are similar to computing units

Regarding the design of Telum, Telum canceled L3, L4 and replaced it with huge L2, which would cause its cache to have a huge access latency, but telum is magical, it puts the data previously placed in L3 in L2 of the remaining spare cores, L4 is the same, resulting in its meaning, a chip physically has 256M L2 (8X32), but a single core can read 32M L2, 256M L3 outside the core, And a whole complex (the entire IBM Z16?) Unit) of 32mx8 core x2 (2die package a sub) x16 of 8GB of L4, indicating that white point, a core data can not be stored in other cores, different substrates, or even different groups or different units of L2, as long as it is free.