It was originally created as a CDE clone (thus the original name "XForms Common Environment")

In college we had a network of Sun workstations and some of the machines had only 8MB of RAM, IIRC. This was in the 90s.

Then again, the X desktop was really minimal and I would use them mostly to code in C using a terminal.

Windows95 was decent even with 8 MB, on a 66 MHz or 100 MHz 486 CPU.

With either 4 MB or only a 386 CPU, it was definitely crippled, making an upgrade not worthwhile.

I remember buying a bulky external 2MB RAM extension (I think I bought another 2MB) before that for my Amiga 500 running a full desktop OS already on 512k 'Chipmemory' using it mostly to actually as a TempFS to accelerate loading. That was beginning to mid 90s, I guess. But running netbsd on the Amiga meant that you would already at that time need 16MB of RAM and a CPU with an MMU as well as an HDD (my friend across the street did that with his A1200 I think I remember). You would only do it if you wanted more networking beyond BBS I guess.

Correct, 4k is very modern by these standards. But then I'm old, so perhaps it's all about perspective.

Back in the days when computers had 8MB of RAM to handle all that MS-DOS and Windows 3.1 goodness, we were still in the territory of VGA [0], and SVGA [1] territory, and the graphics cards (sorry, integrated graphics on the motherboard?! You're living in the future there, that's years away!), had their own RAM to support those resolutions and colour depths.

Of course, this is all for PCs. By the mid-1990s you could get a SPARCstation 5 [2] with a 24" Sun-branded Sony Trinitron monitor that was rather more capable.

[0] Maxed out at 640 x 480 in 16-colour from an 18-bit colour gamut

[1] The "S" is for Super: 1280 x 1024 with 256 colours!

[2] https://en.wikipedia.org/wiki/SPARCstation_5

This was the main driver of VGA memory size for a time - if you spent money on 2MB card instead of a 1MB, you could have higher resolution or bit depth.

if you had a big enough framebuffer in your display adapter, though, X11 could display more than your main ram could support - the design, when using "classic way", allowed X server to draw directly on framebuffer memory (just like GDI did)

X11 was designed to support bit depths down to 1 bit per pixel.

Presumably every pixel is 32 bits rather than just 8. So the count starts at 33.2MB just for the display.