Well maybe hot air coming from the same source.
The 16 PCI-e 3.0 lanes, mostly used for PCI-e slots, are connected directly to the CPU, and they are cool.
The 20 "high-speed" PCI-e lanes made available by the chipset are a hoax. They are all connected to the PCH, which communicates with the CPU through DMI 3.0, which has a bandwidth of just 4 x PCI-e 3.0 lanes.
So we have 20 so-called "high-speed" lanes squeezed through a bottleneck of just 4 lanes to the CPU.
Adding to the problem are motherboard manufacturers connecting PCI-e slots and M.2 slots to the chipset's PCH instead of directly to the CPU.
Usually the 3rd and 4th PCI-e slot, and 1 to 3 M.2 PCI-e slots.
Two examples are the Gigabyte GA-Z170X-UD5 TH and the Asrock Fatal1ty Z170 Professional Gaming i7.
Gigabyte GA-Z170X-UD5 TH :
The 3rd PCI-e slot (x4) is connected to the PCH, not to the CPU.
It's also in contention for those 4 lanes with the M.2 connector. You can't use both at the same time. Using the M.2 connector disables the 3rd PCI-e slot.
You also can't use that 3rd PCI-e slot for SLI because the lanes must come from the CPU according to the SLI specs. (Not a great loss at x4 speed.)
But the big one. M.2 using those 4 lanes would max out the DMI 3.0 interface from the PCH to the CPU.
Further the 2 Thunderbolt 3 USB Type-C outputs are good for 40Gb/s each.
But 16 PCI-e 3.0 lanes from the CPU equal roughly 16 x 4Gb/s = 64Gb/s.
20 "high-speed" lanes from the PCH could [b]theoretically[/] equal 20 x 4Gb/s = 80 Gb/s. Hooray! \o/
Oh no! Those 20 lanes are pushed to the CPU through DMI 3.0 which has a bandwidth of just 4 X 4Gb/s = 16Gb/s.
What is going on here?
Also, on November 19th 2015, I asked Gigabyte (my favourite brand for motherboards and graphics cards) at their support website ( esupport.gigabyte.com ) for a recommendation for a good quality cable or adapter to connect one of the Thunderbolt USB Type-C ports to the DisplayPort port of an LG MU3197-B monitor.
Or to put up a pdf file for download with some recommendations (what type, adapter or cable, what quality)
There were a few responses. They claim to have bought their cable for testing from a blogspot.tw subdomain. Yeah right! Gigabyte shops at a subdomain of a free hosting service. Bottom line, I'm still waiting for a response.
I also hinted in that ticket that I was under the impression that they hadn't tested connecting one of the USB Type-C ports to the DisplayPort of a 4K monitor. To which they responded on Dec 1st with :
"Due to we do not know the market in Netherlands, but we will check our responsible team about our test device,
We will come back to you once we get a reply."
Well, I'm still waiting for that definitive response. 3 weeks and counting.
My gut feeling says that they just slapped the Thunderbolt ports on, ran some lab tests, but never connected a monitor to the 2 Thunderbolt ports.
Note to self : Do not become a Guinea Pig. Do not become a Guinea Pig. Do not become a Guinea Pig.
2 weeks ago a very popular YouTube reviewer about finding a cable/adapter for the Thunderbolt 3 over USB Type-C connector (on the new Dell 15" XPS Infinity) at about 10:17 of the video (starting from 9:49 in) : https://youtube.com/watch?v=evjevd8BaDE&t=9m49s
Seems that she also can't source a [b]problem free[/b ] cable or adapter.
(I can send/post a screenshot of the ticket. But they have also been good to me with beautiful hardware.)
Asrock Fatal1ty Z170 Professional Gaming i7 :
This motherboard had many overselling issues.
Singling out just one.
It has 3 M.2 connectors, between the PCI-e slots. All 3 are connected to the PCH which can only reach out to the CPU through 4 lanes. So 12 lanes squeezed through 4 lanes. 3 x 32Gb/s squeezed through a bottleneck of 1 x 32Gb/s. Let's hope that they do not encourage setting up a RAID array spanning 3 M.2 SSDs.
Yep! Also U.2.
I've given up on Skylake. Sticking to my current ancient boxes. Checking out Intel again in 12 months.