[p0llard]

> tinker with advanced stuff requiring PCI-E chipsets

What sort of things are you talking about here, out of interest? The only way you're going to wreck hardware is if you're attaching something completely incompatible at the physical level (wrong levels, shorts, etc.); if you're "tinkering" you almost certainly don't have the resources to bring up your own PCIe PHY (there's pretty much no way to do this without custom silicon) so presumably you have some preexisting hardware in mind?

If you just want to play around at the TLP level (for example by using the hard PCIe core in a cheap Xilinx Series-7 chip) then this is all pretty safe: (temporary) catastrophic system instability is one thing (I learned this the hard way when I accidentally had my FPGA spamming MSIs every free cycle), but real hardware damage is pretty tricky.

[dfox]

Gamers and crypto miners do various nasty things to PCIe cards in order to use cheapest cable possible to extend it and actual hardware damage resulting from that seems pretty minimal.

[p0llard]

When I said "incompatible at the physical level" I meant it in the OSI sense of an issue at physical protocol level. An electrical specification mismatch could easily cause hardware damage (but would be very hard to achieve unless you were using your own PHY or literally connecting a power supply to the PCIe lines).

> seems pretty minimal

Is there actually any at all? I can't really see how this could lead to any damage.