Maybe I didn't make my point across.
There is indeed new naming for Xeons: Platinum, Gold, Silver and Bronze. Silver and Bronze will be indeed KBL-X and SKL-X. The rest are for multi socket.
It seemed at first that they were keeping socket R3 (my mistake of course, but the S Specs are S R3 and I took it for socket R3, as we know is LGA2011-3) but in fact it's socket R4 (LGA2066) for at least Silver and Bronze.
Platinum and Gold are server CPUs, so socket P, LGA2066.
Did you even look at the LGA 2066 wikipedia.org article I looked in. Socket P is
not 2066. That doesn't even make any rational sense. The number are associated with the number of pins. Gold & Platinum all hook up to more than one sockets. So the inter-socket interconnect network demands on those are higher than that of a single socket. Therefore multiple socket solutions need
more pins, no less, than a single socket solution. There are more things that need to get connected so need more connections.
SLK-W ( and therefore SLK-X, Basin Falls ) is LGA 2066. An
increase over the 2011 in part because there is now more I/O. ( yes there is some 2 socket interconnect they can recycle into something else since one socket only, but other I/O growth issues present too. )
https://en.wikipedia.org/wiki/LGA_2066
SLK-SP ( change from -EP , "Purley" ) is LGA 3647.
https://en.wikipedia.org/wiki/LGA_3647
http://www.cpu-world.com/news_2017/2017042701_Upcoming_Xeon_Gold_and_Platinum_CPUs_spotted.html
http://www.anandtech.com/show/11300...ld-and-xeon-platinum-cpus-for-purley-platform
Apple is not likely to use Purley for a Mac Pro. There is a significant amount of stuff built into Purley that is extremely solidly aimed at data center contexts. People's desks are in data centers.
The LCC KBL-X should be identical to the desktop i7/i5-7x00, renamed to 7x40K.Dual channel mem, 4 cores, different socket. And this must be only to maintain the socket across all HEDT options.
KBL-X is
not LCC. It is
not really the next step after SKL-X (despite desktop Kaby Lake following desktop Skylake). At least not in the terms typically used in classic Xeon -EP contexts. The LCC was the range covered by the Xeon E5 16xx series. What KBL-X is more akin in Xeon terms to what has been the E3 12xx series. It isn't LCC at all. It is straight up the baseline design from desktop/laptop product line. Period. That isn't "low core count". There is extremely likely a disconnected iGPU on the die.
It would clean up tons of crap and confusion though if the KBL-X actually did end their HEDT line up ( Core i7 yxxxK and that Intel puts SKL-X into some other category ( maybe core i9 , core W , or something different, because it is different) . It is kind of goofy there are two radically diffient dies in the Core i7 label. I just generates tons of confusing, junk discussions on forums and with customers. )
The rest should be LGA2066 with 4 channel mem and more cores, 12 to be precise.
But this is HEDT. For the mMP Apple might go with the Gold series, or not.
If "Core i7 .... " is part of the HEDT then, as above, it is past time for a new definition.
Will the Gold series be socket LGA 2066 and use X299 (or C6xx) as well or will they be Purley. Or is Purley only for Platinum series 8000?
Gold
and Platinum are Purley. They have been Purley for years. Purley was always >= 2 sockets. Since 2015 era roadmaps:
http://www.itworld.com/article/2985214/hardware/intels-xeon-roadmap-for-2016-leaks.html
One socket workstation has been consistently been carved out as a
separate 'swim lane' for years now.
Two different code names for two different product lines.
I'd think Intel would have WS CPUs still (1600 alike), so it would make sense those to be LGA 2066 as well. Platinum would be LGA 3647 for multi socket platforms. Silver and Bronze for HEDT.
As well? There isn't any else in 2066 but the -W ( workstation and borrowed top end overclocking -X ) line up. It is a one socket only design.
If not, WS manufacturers will either have to use HEDT parts (which is odd)
It isn't odd. They have been doing it for almost a decade now. What was actually more odd is that Intel was partially coupling the Servers to HEDT chipsets . That is what was odd. The C6xx series has traditionally had multiple entries. ( low end and then higher with more SATA and RAID goodies) , but why does a server need 6-8 USB sockets? They don't except for extremely weird corner cases.
The -E/-EP chispets being on a server schedule ( once per socket design) as opposed to the more modern desktop/laptop schedule ( iterating at the same speed as the CPU packages) is partially what has inhibited the top end of the Core i7 product line keeping in step with he rest of the line up. Decoupling off workstations and hooking up to a minor variation of the laptop/desktop PCHs should allow Intel to revise the single socket workstation designs as a pace closer to that of the mainstream desktop line up. With Ryzen 5-7 that is an even more pressing issue for Intel. They don't fully over the SKL-X line up but there is some overlap at the bottom end. ( KBL-X is even more overlap that is a stop gap until Intel can pick up the pace. )
or higher end server parts, adding to the cost and complexity.
But who knows, all about Intel is very murky right know.
The PCH uniformity does make sense though.
PCH uniformity aimed an single user systems makes sense. Same on the data center (multiple remote users ) uniformity also makes sense.
Intel isn't all that murky. There seems to be folks who want to make it murky. Want Apple to pick Purley for Mac pro for some reason that has problems in the big picture. That mismatch is what creates the murky. Similar folks always seem to make the difference between Core i7 x9xxK and i7 x7xxK murky for no good reason also.
please make me eat my hat on that one!