It's still early. The consumer market is being targeted not the prosumer market. I think zen+ / Naples / zen fx will hit that market.
There is your problem. The Ci7 6900K is NOT a typical competing Desktop Processor, it is a workstation-prosumer part derived from the Xeon E5 platform with a hefty price tag to match, yet AMD is comparing desktop Ryzen to it. Its solid from a compute and price-performance perspective and makes Ryzen looks good as a CPU, but you're forgetting the I/O difference, which should be obvious, because if you're comparing the AM4 Ryzen against the LGA 2011-3 Ci7 6900K, you're also comparing a desktop platform against a highly expansive Server-first platform. And for that matter, for the price tag of a single Ci7 6900K, you can probably fit a Dual Xeon E5 2620v4 platform with twice the expandability, if that is what you need.You are comparing a desktop processor to what is essentially a workstation-prosumer part which is not an accurate comparison. Ryzen has always been compared against competing desktop processors from Intel such as the 6900K and others. Ryzen was intended as a desktop processor and not as a prosumer part which is why AMD felt that 24 lanes and dual-channel DDR4 is more than sufficient.
These two ugly Ci7 models got the distinction of being the only two Processors in the entire LGA 2011-3 lineup that does not provide the full 40 PCIe Lanes, and forces Motherboards manufacturers to include details about which Slots are limited or can't be used with these handicapped models. Is still the classic Intel market segmentation... There is a reason why I like more the Xeon E5 1620v4 to the Ci7 6800K.For reference the 5820K and 6850K are also 28 lane processors and even with a video card and other devices still would have difficulty saturating even a x4 link for NVME storage. That's also not including that motherboard manufacturers can easily add PLX switches on designs for x16/x16 in SLI or Crossfire designs.
Because the Xeon E5 1620v4 cost the same than Ryzen 7 1700, yet provides the full Broadwell-E feature set minus Multiprocessor (From a feature set perspective, is better than the handicapped Ci7 6800K, if you don't mind losing 2 Cores). If you need tons of RAM or PCIe Bandwidth, the Xeon E5 1620v4 provides at the same price point than the Ryzen 7 1700 the entry level to a more expansive platform than AM4.I don't understand why you are referring to the E5-1620 V4 when the only advantage it holds over the 1700 is larger memory capacity availability and ECC support. Even the R7 1700 would still beat it in multi-threaded and come close to single-threaded. Not to mention that the E5-16 V4 series are not unlocked (unlike the V1/V2/V3 series) whereas the R7 1700 is unlocked and only slightly more expensive. The bottom line is that if you want Zen Opterons and lots of lanes then you need to wait for Naples which will offer quad-channel DDR4 and also 128 lanes. And no the Intel LGA2011-3 platform is not superior when it comes to price/performance. Ryzen costs over 50% less and provides similar if not better and prosumers make a much smaller market segment and share than consumers who use desktops.
AM4 provides LGA 1151 comparable features, which makes sense as they're both desktop platforms. So far, the only HEDT-only feature that you're getting in AM4 is more than 4C in a desktop platform.Exactly. Zen is not targeting prosumers but the desktop market segment. That doesn't mean that AM4 doesn't provide HEDT features or performance but rather it does so more so by price/performance.
I never suggested 16x/16x at any point and already know that usually there are no tangible benefits. I specifically mentioned the 8x/8x/8x/8x/8x arrangement (5 PCIe Slots coming directly from the Processor) for a reason.His main argument is about lanes but even x8/x8 on PCI-Express 3.0 versus a x16/x16 in real-world scenarios doesn't yield real-tangible benefits even in high-intensity 3d applications such as games. AM4 is a unified socket that will be used with APU's, CPU's and other products from AMD (aside from Naples which is LGA based).
I already know the Broadwell-E dies: The Intel Xeon E5 v4 Review: Testing Broadwell-EP With Demanding Server Workloads Intel Broadwell-EP Architecture, Models & PricingThe 6900K is a desktop processor. The 6900K is a 6950X die with two disabled cores. All Broadwell-E processors from the 6800K to the 6950X use the same die. It is not derived from the E5 platform. While it uses the MCC die configuration you can compare the 6950X or 6900K with Xeon E5 V4 processors and see the difference. You can see the die here:
http://hothardware.com/ContentImages/Article/2470/content/small_broadwell-e-die-shot.jpg
The 6900K is a desktop processor just like all the other Broadwell-E parts aside from the 6950X. It just has a higher price range and bracket but it's not a prosumer or workstation part. You can't take a 6950X or a 6900K and put it in a LGA2011-3 socket that uses anything other than X99. I don't see what your point is with that. Consumers don't need quad-channel or 40 lanes. Ryzen provides higher than Skylake dual-channel memory bandwidth and has 24 lanes available (motherboards are offering x16/x8). That's more than enough and most consumers who buy processors are those who fall under the $500 price segment. Prosumers make up a tiny market segment/share. In real-world scenarios more lanes and quad-channel doesn't make a significant difference for consumers and that's what matters.
You missed that PLX was purchased by Avago, and prices increased two or three times: Business side of PLX acquisition: Impediment to NVMe everywhereNf200 and PLX chips don't add too much cost overall to the motherboard. As I remember even with LGA1366 and LGA1155 motherboards it would only add about $20-50 at the maximum with the only disadvantage being a slight performance penalty due to latency (with nf200 for example). And let's be honest would you still buy a E5-1620 V4 just for 40 lanes when you can get a 5820K or a 6800K for less and have two additional cores and the ability to overclock (which you overlooked).