Slashdot Mirror
Intel's 8th-Gen 'Coffee Lake' Core CPUs Will Be Revealed During the Great American Eclipse (pcworld.com)
Brad Chacos, writing for PCWorld: Intel's response to AMD's disruptive Ryzen processors is soon to get its time in the sun. Well, sort of. On Tuesday, Intel announced plans to livestream the launch of its 8th-generation "Coffee Lake" processors on August 21 -- the same day that the great American solar eclipse casts its shadow across the United States. Intel's throwing shade. Eighth-gen Coffee Lake chips will be built using a revised version of Intel's 14nm process technology for an unprecedented fourth time, following in the footsteps of Broadwell, Skylake, and Kaby Lake architectures. You'll probably also need a new motherboard to use them. But most notably, Intel claims 8th-gen Core chips will be up to 30 percent faster than today's Kaby Lake processors in some applications. Intel chips haven't seen a performance leap like that in years. Beyond that, little is officially known about Coffee Lake, though the churning internet rumor mill thinks that Intel will up the core counts this time around to combat the threat of Ryzen.
This CPU will deliver phenomenal performance but only during solar eclipses.
"Coffee Lake" barely beat out "Man Bear Pig"
Given that just yesterday, Intel announced a product release for September 25th isn't it a little early to talk about what comes after? We haven't had the chance to see how their current compares to AMD's offering which isn't due until later this month.
Guess Intel is afraid of something. Maybe AMD has given them some much needed competition.
Eighth-gen Coffee Lake chips will be built using a revised version of Intel's 14nm process technology for an unprecedented fourth time
Intel is switching to complete 14nm FinFET here. The reason is that some of the competing fabs that stayed with FinFET are now manufacturing 10nm FinFET while Intel spent 3 years now trying to make 10nm Tri-Gates economical and have failed miserably. The Tri-Gate lithography is just too expensive: Too many steps, and the yields too poor.
When Intel beat the world to 14nm it didnt matter so much that Tri-Gate's were not as economical as FinFET because Intel beat the world there by a big margin and didnt have to compete. Now they not only have to complete against 14nm FinFET but also 10nm FinFET and it wont be long until 7nm FinFET is in production by at least the companies that skipped 10nm on purpose (they are not the same fab companies as the ones producing 10nm FinFET's today)
Companies that beat Intel to 10nm (so far):
Samsung, TSMC
Companies that absolutely will beat Intel to 7nm:
Global Foundries, TSMC, Samsung. I predict that this is the order that it will happen in. GloFlo skipped 10nm on purpose to be the first to 7nm.
"His name was James Damore."
Intel's 8th-Gen 'Coffee Lake' Core CPUs Will Be Revealed During the Great American Eclipse
Quick! While nobody's looking!
Taking guns away from the 99% gives the 1% 100% of the power.
Has there ever been a new generation where the promised increase wasn't "up to 30%" ? Because I've never heard any other number used.
Also, am I the only one that finds Intel a wee bit histrionic this last month or so? They've been throwing everything they can (including the chairs) at the wall, but nothing seems to stick so far.
There's part of me that wonders whether Intel knows its offering is going to be underwhelming, and is therefore choosing to release it on a day when a lot of tech/science types will be somewhat distracted.
#DeleteChrome
Just to clarify, 10nm refers to the "process" and not the actual pitch of any of the resulting elements. Each fab implements it differently which results in different densities for the same "process". For example, the TSMC 10nm process has a gate pitch of 66nm and and interconnect of 44nm while Samsung's 68/51 respectively... the end result being that TSMC 10nm chips are denser than Samsung. Intel's is supposed to be 54/36 which is actually the same pitch as Samsung's proposed 8nm process and slightly smaller than TSMC's and Global Foundaries proposed 7nm processes (54/40 and 56/40). Intel's numbers have to be taken with a grain of salt, Samsung and TSMC both initially fudged their 10nm process pitches and Intel may be doing the same. However, the process size is slowly becoming a marketing gimmick.
It isnt really AMD that is making Intel bend over.
Its Samsung and TSMC.
Intel was a generation ahead on process tech with AMD's last iteration.
Now Samsung has sold time to AMD for 14nm production (Ryzen)
and TSMC has sold time to AMD for 10nm production (PlayStation and XBOX console chips)
AMD has always had the better CPU design. What they didnt have was access to matching process tech for about 8 years or so.
AMD is not even close to Intels rival. The other fab companies are, and they are ahead and pulling away now.
"His name was James Damore."
Don't expect much, it's a rework on the same process. Intel has literally stalled out making it to 10nm, they've now redesigned the same cpu on the same process 4 times. The entire Intel advantage of being several process steps ahead of everyone else has evaporated over the course of the last 2 years. Coffee lake will likely be no faster than the generation before it or the 2 more before that.
They will claim it's 30% faster because it has some new instruction that allows it to perform a single task that much faster, but only if you have the CPU and only if your code supports the instruction. Usually it takes a decade before new instructions a prevalent in software.
Covfefe lake processors have only the best cores. Believe me, it's true. And they're fast. They'll drain the process pool for you so quick, you've never seen so quick. No CPU respects threads like Covfefe processors do, it's unpresidented: it grabs them by the stack frame, programs let you do that, you know, when you're a Covfefe multicore CPU. Howe many cores? I'll keep you in suspense. Do Covfefe CPUs support hyperthreading? Yeah, I guess so.
I've fallen off your lawn, and I can't get up.