The primary benefit of moving to 64-bit is the increase in the maximum allocatable random-access memory (RAM). 32-bit editions of Windows XP are limited to a total of 4 gigabytes.
That's the official statement, but there's a PAE patch which will enable up to 64GB of RAM. (There are also claims of a "128GB" PAE patch, but that is physically impossible as x86 never had a 37-bit addressing mode.)
Kinda interesting same thing applies to JVM. But the JVM compresses the 64 bit pointers to get you up to 32GB if memory with 32 bit space. After that it reverts back to uncompressed, but with custom flags you can do intermediate compression up to 64GB.
I don't think it's normally apparent, since PAE is needed for NX (which is enabled by default starting with XP SP2) --- so the majority of 32-bit systems at the time were running in PAE mode, just with an artificial limitation to 4GB.
I was working for a company that wrote device drivers for OEMs during the Windows XP era.
XP X64 was no big deal, and OEMs often requested 64-bit drivers, but I'm not convinced there was ever any significant install base of actual 64-bit-using end users. Not until Windows 7.
As a younger and less experienced computer user, I remember buying 64 bit XP for my first self-built desktop, because it seemed like it had to be roughly 2x better than 32 bit. Imagine my surprise when iTunes failed to install: Apple had no supported install path for 64 bit XP. I had to do shenanigans with the installer file to trick it into thinking I was on 32 bit.
That's no uncommon. I've tested hundreds of apps on every Windows OS from 95 to 11 over the years. A bunch that required drivers or tied in more deeply to the OS just failed on Windows XP 64-bit. It does make sense due to Windows XP 64-bit using a different kernel version than 32-bit.
That's the official statement, but there's a PAE patch which will enable up to 64GB of RAM. (There are also claims of a "128GB" PAE patch, but that is physically impossible as x86 never had a 37-bit addressing mode.)