I don’t deny this. I’m saying that there is little VALUE in offering PCIe slots that will be constrained at the interconnect if fully populated. At least in my opinion. I suppose there might be some fringe use cases where you’re not using the cards in question simultaneously. You are effectively paying for marketing hype that will never really deliver and both teams are guilty of this. I wouldn’t mind so much if the ratio was less than 2:1 (though 1:1 would be better… admittedly you might at as well incorporate everything into the SoC then? Remember north bridges?).
You’re also making the assumption that I don’t want to pay the price for more lanes. That I’m arguing I should get TR platforms lanes for AM4 prices. Never said that. Not to mention it’s unknown when a Zen2 version will be available?
I guess the TL:DR is It’s just frustrating when you can’t find the mix of features you need w/o buying a bunch that you don’t?
Again… consumer platform, and even relatively high end users won’t be pushing every device connected to the chipset at full speed all the time. Or simultaneously, ever.
Given the typical end user case of 1x GPU, you still typically have a second slot wired for x8 (your discrete GPU will be fine with x8, especially if it is PCIe4 x8 in a new build) that can do anything high bandwidth (capture card, RAID controller, whatever), network adapters, etc. unless 10 GbE (and desktop 10GbE is still very rare just yet) are just fine with chipset bandwidth…
What’s your use case?
I mean i get it - there ARE use cases, but typically by the time you go down that path you’re likely wanting more cores/RAM anyway. And you’re WAY beyond the 99.9% end user case - why should every end user doing none of that be paying for more expensive boards to support stuff they don’t need?
oO;
At this point we’re going in circles and here you arguing my point for me? Why pay for PCIe slots that aren’t going to be used? Let’s just agree to disagree.
On topic - X570 is not a solution for serious FreeNAS builds. In fact an argument can be made that the project favors Intel hardware. Well at least in the NIC side. This is even more pronounced in BSD based projects like pfsense that nearly went Intel only with their upcoming version but decided against it (scraped plans to implement REST api which requires AES-NI). I agree that TR is the way to go, but personally I’d wait for the Zen2 versions. That’s just me I guess. So, for me there isn’t really a AMD solution atm. Which is sad because I really don’t want to go Intel personally.