Yes, why would anyone want to know about actual practical experiences with these parts… 8 or 16 GHz over copper PCB traces or God forbid cables is no biggy and every manufacturer only uses the best of quality of components for their parts that - from the outside - can’t be just visually checked if they actually perform to spec.
(Luckily humanity has completely evolved from scaming each other…)
PS: USB 3.0 5 Gbit/s was absolutely great starting right from the beginning! No half-baked controller chipsets or bad RF shielding that negatively impacted 2.4 GHz Bluetooth and Wi-Fi components at all…
I think that would depend on how the BIOS is handling PCIe enumeration and the NTBs+bifurcation domains setup in the switch’s firmware; might not be easy to predict how they’ll turn out because BIOS enumeration might become upset with specific combinations and do weird stuff.
I strongly doubt that you can make any useful predictions what’s going to happen downstream of that Microsemi PCIe Switch chipset based on what happens with the default onboard devices on a motherboard.
I’ll also think about a few questions for that manufacturer.
Yep, also b650 runs with pcie gen4 x4, instead of intel dmi. That means it might run on intel platform.
X670E platform has 24 pcie gen 5 lines from cpu directly, 12 pcie gen 4 lines from the chipset, 8 pcie gen 3 lines (configurable as sata) from the chipset.
Call me greedy, but the prospect of even more bandwidth has intrigued me. I’m a Broadcom user as well and my X99 board definitely uses onboard PLX. I never thought I’d see the day when 40 lanes ain’t enough but sometimes I feel as though I could use a little more. All this from a domestic end user. Go figure. Ironically, I also have an X570 platform.
Various ML hardware uses tons of PCIe to send huge vectors of floats around different pieces of neural net running on different cards or even machines, … which drives prices up for PCIe and network gear horrendously. Why do you think Nvidia snagged Mellanox?