@apiltch ,
As our tests show, it’s pretty difficult to use more than 4GB of RAM on Raspberry Pi, even if you’re a heavy multitasker.
Try
leaving open a few windows/tabs on tomshardware.com articles, without an ad blocker. Over time, you can use
about:performance
and watch the memory grow. Just the tab with your 8 GB review article is currently using 1.21 GB, on my machine. Your home page is using 545.9 MB.
Last year, I upgraded my Win7 64-bit machine from 8 to 16 GB because browsers burn so much RAM, these days. Some of it is clearly the fault of leaky, client-side adware.
A new OS could also take advantage of the RAM to improve performance.
One of the main performance benefits would be simply using the A64 ISA, which is more efficient all on its own.
You can sort of compare AArch32 and AArch64, here:
...except for this:
All the preloaded apps are 32-bit, so they don’t really take advantage of the 64-bit capability.
That's just stupid. Hopefully, they'll switch to building everything in native AArch64 mode, before it leaves beta.
Strangely, there’s slightly less available RAM in the 64-bit OS than in Raspbian, with Raspbian showing 7.8GB available while the 64-bit OS has 7.6GB available.
When you build in 64-bit mode, your pointers all expand to occupy 8 bytes of memory instead of 4. There might also be additional alignment restrictions and larger stack sizes for each thread in the system. So, it's really not surprising.
BTW, a "pointer" is a data type which holds a memory address. They're used in various data structures, for program abstractions, to hold procedure addresses, etc.
Setting up a RAM disk that stores your most frequently accessed apps in memory is one way a typical user could take advantage of that extra 4GB of memory.
No, because you'd have to copy them into it, every time you reboot. It's better to follow what Eben said, and just let the OS cache your frequently-used apps and files.
The one area where the 8GB Raspberry Pi 4 really excelled was in copying extremely large files. When we ran IOzone, a synthetic test that measures reads and writes, with 4GB of data, the 8GB capacity crushed the 4GB model on read speeds, performing both random and sequential reads that were more than 20 times faster due to the larger Pi’s ability to cache the all the data.
Normally, people test uncached reads and unbuffered writes, to measure the actual storage performance. If you tell people that it's faster at copying files, they're likely to be disappointed since most files one copies won't be already be in cache.