Slashdot Mirror
Intel's Upcoming Coffee Lake CPUs Won't Work With Today's Motherboards (pcworld.com)
Intel's upcoming Coffee Lake CPUs won't work with existing 200-series motherboards that support Kaby Lake, a manufacturer confirmed on Wednesday. In a Twitter post by Asrock last Saturday, the company confirmed the news when asked if "the Z270 Supercarrier [will] get support for the upcoming @intel Coffee Lake CPUs." Their response: "No, Coffee Lake CPU is not compatible with 200-series motherboards." PCWorld reports: According to at least one reliable source outside of Intel, the new Coffee Lake CPU will indeed not be compatible with Z270 boards, even though the chipsets with the upcoming Z370 appear to be the same, PCWorld was told. The source added that there are hopes in the industry that Intel will change its mind on compatibility. Tomshardware.com said it had independently confirmed the news with Asrock officials as well.
Why this matters: The vast majority of new CPU sales are in new systems, and they likely won't be impacted by the incompatibility. However, there's also a very large and very vocal crowd of builders and upgraders who still swap out older, slower CPUs for newer, faster CPUs to maximize their investment. An upgrade-in-place doesn't sell an Intel chipset, but it at least keeps them on the Intel platform. If consumers are forced to dump an existing Z270 motherboard for a newer Z370 to get a six-core Coffee Lake CPU, Intel risks driving them into the arms of AMD and its Ryzen CPUs.
Why this matters: The vast majority of new CPU sales are in new systems, and they likely won't be impacted by the incompatibility. However, there's also a very large and very vocal crowd of builders and upgraders who still swap out older, slower CPUs for newer, faster CPUs to maximize their investment. An upgrade-in-place doesn't sell an Intel chipset, but it at least keeps them on the Intel platform. If consumers are forced to dump an existing Z270 motherboard for a newer Z370 to get a six-core Coffee Lake CPU, Intel risks driving them into the arms of AMD and its Ryzen CPUs.
Intel, for as long as I remember, needlessly changed sockets.
Actually no, this is probably a sign that a new socket is on the way. This is not news, as Intel has been doing this exact pattern for a while now; Intel will keep a single socket compatible for at least two Core generations before replacing it and breaking compatibility. Sandy Bridge, Ivy Bridge (1155). Haswell, Broadwell (1150), Skylake, Kaby Lake (1151).
Outside of the gimmicky super-shredder-killer-fps-man-slayer motherboards, it's not like they have been the most expensive part of a computer build for a long time. Introducing a new video card incompatibility like the transition from PCI -> AGP -> PCI Express would be a whole different story.
I'ts been a long time since AMD has released a competitive product. Intel in a show of appreciation and friendship has decided that the best way to help them along is to assure that unlike the new series of ryzen processors coming out theirs will not be backwards compatible with the hardware you buy. Why else would they restrict the pcie lanes in their top of the line chips by price and lock out features unless they were trying to help AMD along.
once more into the breach
Are you upset that a Ferrari costs 10x the price of a Honda?
Yes.
The cesspool just got a check and balance.
As a professional game developer, I can assure you that videogames are the exact opposite of an "embarrassingly parallel" problem. There's a LOT more to videogames than graphical processing: AI, pathfinding, physics, audio, resource and memory management, animation, world updates / occlusion systems, etc. These are all CPU-intensive subsystems, and many of them are inherently bound to global data (a virtual world simulation), which makes it extremely tricky to split off into independent threads (probably the only exception being audio).
That means that each subsystem in the game must be carefully and painstakingly optimized for threaded performance. It's not possible to trivially split up all these subsystems by thread either. Many of these systems tend to interact with each other and the global world database, and that means the gains tend to be smallish and non-scalable in nature.
Generally speaking, it's an extremely difficult problem, and one which I don't think the industry has really cracked yet. Believe me, if it were trivial to do, we've have done it a long time ago.
Irony: Agile development has too much intertia to be abandoned now.