Y. B550's are basically what will dictate if X570's are even worth considering. If they do the same as B450, with boards having good enough VRM/phases, good looks, and whatnot, then most of the 3600X/3700X people will just get a B550 and be happy with it. (I think even if they shipped with current PCI-E 3.0)
Then again, if those nice mid-X570 boards like the Gigabyte X570 Aorus Pro arrive at around $170~$180, then I think people might make the effort. The VRM/phases on those look really solid, according to Buildzoid.
The Aorus Pro is $250.
This really shouldn't come as a surprise to anyone. PCIe 4.0 requires higher PCB quality (likely through more layers) than 3.0 to ensure signal integrity, as well as redrivers driving up the BOM. Same goes for RAM traces if motherboard makers want to compete in the >4000MT/s RAM space. Then there's the cost of a large-die chipset that includes a PCIe 4.0 switch. X570 will be more expensive than previous AM4 X*70 series across the board - and that's fine, as it brings a lot to the table.
Hopefully when B550 launches it's a selectively stripped-down option that allows for cheaper boards - ideally maintaining rough feature parity but moving down to a PCIe 3.0 switch - even with less lanes that would still be useful, and a lot cheaper to produce. You'd still get PCIe 4.0 through the PCIe_x16_0 slot and the first m.2 slot (unless motherboard makers screw up the traces entirely), but cut pretty much every other element driving up prices of X570 boards.
Meh, I don't care about PCIe 4 anyway. Nothing that has the power to justify being on it is going to be cheap anyway.
Show some more performance numbers, AMD.
I'm thinking with lane splitting and everything, you could connect several NVMe drives directly to the CPU. But we'd need NVMe PCIe 4.0 x2 drives instead of the current 3.0 x4 drives first.
I see
forward compatibility as the main reason to go X570. Even if I am - despite my best intentions when buying my 1600X - planning to upgrade to an X570 setup a scant two years after my last go around, most people buying a motherboard+CPU combo today ought to plan to keep it for 4 years at the very least (heck, I kept my last setup for 8!). (Also, part of the reason I'm upgrading again is that last time around had me quite budget-limited, and this time around I can go balls-to-the-wall, which will be great for longevity.) AMD has given CPU performance a surprise jolt these past few years, but performance increases are going to flatten out very soon. In 3-4 years, PCIe 4.0 SSDs will be ubiquitous, as will GPUs, and a 3700X (or even a 3600!) is still going to be a
good CPU. The main gain of X570 isn't necessarily per-device bandwidth, but total bandwidth and thus the total amount of connectable high-speed devices. It's rather obvious that we'll start seeing PCIe 4.0 x2 SSDs in the coming years, likely at prices somewhere in between current x2 and x4 drives (if not lower as flash prices keep dropping). That makes them a no-brainer for adding storage over time - no real performance sacrifice, but cheaper than x4 drives, and hopefully you'll be able to connect more.
Some X570 boards have Wifi 6, 6 to 8 layer PCBs, 10 Gigabit Ethernet and 16 Phase VRM controllers. They will be more costly than X470 simply because of those and other factors.
There, fixed it for you. Only the increase in PCB layers is universal. 10GbE is definitely staying a premium option given that controller chips alone are $50 or more. Those VRM layouts are flagship boards only. And there have been quite a few boards shown off without WiFi.
Where did you get 11W estimate? AFAIK the only number I've heard was no less than 15W, and given the thermal density of x570 you'll probably need an active cooling system even at the most optimistic TDP.
11W was the number consistently given by both AMD and partner reps at the AMD event - though sources like that are rarely specifically named and sure don't leave a paper trail. It seems somewhat unclear whether this is average power or max TDP - some have stated one, some the other (with 15W being named the max in the cases where 11W is said to be average). In any case, the TDP is supposed to be slightly lower than the EPYC version of the chipset, but not much, and reportedly it won't throttle down significantly when not in use.
Um Intel will have PCI-E 5.0 on their 2020 release.
Server and datacenter only. Likely only the >3000-pin socket designs to begin with too (Xeon Scalable, if I understand their naming correctly), with it possibly trickling down to lower-end Xeon (>2000-pin sockets), HEDT and consumers over the coming years - but there's little reason to expect that to be a quick process. If PCIe 4.0 is driving up motherboard costs, 5.0 will be an utter nightmare.
I am only going to reference WIFI 6. You can buy routers that are 802.11AX but they start at $300 as a minimum and just under $700 here in Canada. The problem is that a Wifi 6 card is a brand new adapteranf cards are usuaslly 1/2 of their compliant router in terms of cost. Wifi 6 should add at least a $50 premium to boards that don't have it.
m.2 WiFi cards are usually in the $20 range, pretty much regardless of the standard and performance. PCIe AICs are much more expensive, simply because that's seen as a retail part and is thus sold at "consumer-facing" prices, while the m.2 cards do the same job (arguably more, as PCIe AICs usually don't have bluetooth) but at off-brand/replacement part prices.
What I find interesting is that they're all calling it Intel AX200, when the AX200 is a CNVi part that only works on Intel's newest chipsets - AFAIK, the regular m.2 PCIe version is called the AX201 (at least according to various regulatory agencies) but has yet to be officially launched. I suppose they might have stealth launched it and ditched the differentiated naming - it's pretty much the same product, after all.
We can always hope.
