On Macs that have HDs this is why WoW can take literally upwards of ten minutes to load into the world from the character select screen - you’ve got both APFS and CASC slowing things down, and then you have the added latency from physical head movement. With both APFS and CASC active simultaneously it’s like fragmentationception. With APFS you’ve got that and the b-tree node clusterfudge. At least with HFS+ you’re only dealing with CASC’s inefficiencies by themselves. The reason we tell people to format the SSD/NVMe drives as HFS+ is to avoid the double dipping of fragmentation causing slowdowns. NVMe hides it the best due to both more robust I/O than SATA SSDs and higher total throughput, both in sequential and random 4k reads (R4K is where all drive types are their slowest). Even on an SSD performance will degrade well below optimal after only a few patch cycles. Now pair this with CASC’s own fragmentation issues, which are almost identical to that of APFS and you have a recipe for severe slowdown over time. The more nodes that have to be enumerated, the slower the load becomes. That process is the same on Intel or M1 CPUs.
The process to load them so the entire file can be loaded properly is called enumeration. The more patching that occurs on any given file, the more nodes are created for that file. As CASC’s index files and data files are updated, new nodes are created pointing to the new data while the original file and node remains. APFS creates new b-tree (extents) nodes for a file whenever it’s updated. You’re essentially seeing what SSDs do - hide APFS’ faults. Maybe M1 makes better use of APFS than Intel chips.
0 kommentar(er)
