alt Hacker News> gone are the days of PCIe
This is a wild and very wrong take.
Just about every single consumer computer shipped today uses PCIe. If you were referring to only only the physical PCIe slots, that's wrong too: the vast majority of desktop computers, servers, and workstations shipped in 2025 had physical PCIe slots (the only ones that didn't were Macs and certain mini-PCs).
The 2023 Mac Pro was dead on arrival because Apple doesn't let you use PCIe GPUs in their systems.
Replies
My post Mortem sentiments exactly. The lack of Nvidia GPU support for the M series Mac Pro models kneecapped the platform for professionals. If Apple had included that in those they’d be the defacto professional workstation for many more folks working in AI tech.
Plus modern interconnects like CXL are also layers on top of PCIe, and USB4 supports PCIe tunnelling. PCIe is a big collection of specifications, the physical/link/transaction layers can be mixed and matched and evolved separately.
I don't see it disappearing, at most we'll get PCIe 6/7/etc.
Yup the 4090 and SoundBlaster ZXR in my AM5 7800X3D system would both like to upvote your reply.
> This is a wild and very wrong take.
That's what happens when you quote only part of a statement. Taken in context, it was referring to a very real decline in expansion cards. Now that NICs (for WiFi) and SSDs have been moved into their own compact specialized slots, and Ethernet and audio have been standard integrated onto the motherboard itself for decades, the regular PCIe slots are vestigial. They simply are not widely used anymore for expanding a PC with a variety of peripherals (that era was already mostly over by the transition from 32-bit PCIe to PCIe).
Across all desktop PCs, the most common number of slots filled is one (a single GPU), and the average is surely less than one (systems using zero slots and relying on integrated graphics must greatly outnumber systems using more than one slot).
Even GPUs themselves are a horrible argument in favor of PCIe slots. The form factor is wildly unsuitable for a high-power compute accelerator, because it's ultimately derived from a 1980s form factor that prioritized total PCB area above all else, and made zero provisions for cards needing a heatsink and fan(s).