>Intel recently announced that its 10nm mainstream processors will only arrive by the end of next year, and the continuing delays in the architecture, which was planned for the first time in 2016, could ultimately lead the company to stop manufacturing its major processors.
>According to DigiTimes, the company may end up adopting solutions from other manufacturers, such as TSMC, Global Foundries and Samsung, which are currently working on 7nm chips.
digitimes.com
THANK YOU, BASED INTEL!
Other urls found in this thread:
bitsandchips.it
semiwiki.com
globalfoundries.com
twitter.com
5 ghz pn all 32 cores is coming 2H 2018. Amd is btfo
delet
I bet they got it ready for production like 10 times by now but every time a new vulnerability goes public they go back to the drawing board.
TAIWAN NUMBER ONE!!!
>According to DigiTimes, the company may end up adopting solutions from other manufacturers, such as TSMC, Global Foundries and Samsung, which are currently working on 7nm chip
AAAHAHAHAAHAHAHAAAHAHAHAAHAHAHAAAHAHAHAAHAHAHAAAHAHAHAAHAHAH
INTEL IS FINISHED AND BANKRUPT
Can't they use someone else's process if they pay a fee for it?
The equipment is identical anyways.
I think that would be far more economical than scrapping fabs.
It's not the architecture, it's the manufacturing process.
Intel doesn't know how to work with the new extreme ultraviolet lithography machines, which is very different from the immersion lithography machines they use now.
delid
time to start getting those gov't contracts and blackmailing OEMs and AIBs.
Whatever it takes, right babe?
>The equipment is identical anyways.
no.
You have to order different equipment for EUV masks, orders might be sent late '19 to and so on.
non-EUV sub-10nm gives very low yields and horrible quality chips.
I meant everybody uses the same EUV equipment, so say Samsung process should work in Intel fabs.
Maybe if they started 2 years ago.
everybody buys the EUV masks from the same manufacturer... because only one makes those.
Samsung process is built around mobile.
TSMC is build around big chips like novidia's and amd's GPUs.
GF's is built around Zen and APUs.
Intel CPUs with igps have 2 parts.
one that uses the latest node, e.g. 14+++++++nm which goes for the CPU
and the other one is made on the previous node to avoid losing a chip because of the dead igp part.
I don't know with the latest +++++++ nodes if they moved everything to 14nm, but previous gens on 14nm carried 22nm ipgs, and before that 22nm CPUs carried 28nm igps and so on.
If intel wants to use Samsung's 7nm node or TSMC's node, they have to do one of the following.
1st port their whole 10nm cpu design, to TSMC's or Samsung's 7nm node and use their ASIC libraries
or
2nd port their whole 10nm CPU design to TSMC's or Samsung's 7nm node and use TSMC's or Samsung's ASIC libraries and tools.
What does this mean?
It means that even if intel manages to sell 7nm CPUs early next year(impossibru), and finally manage to get their fucking node to work on either 10nm or 7nm, they will have to port their CPU design to their inhouse node and fix/update their ASIC libraries to work with their node.
This means that intel's anus is gonna bleed money way more than what they lost in their "smartphone" attempt.
I don't think that Intel will manage to get sufficient capacity in any of the 3 major fabs, because Samsung has Qualcomm and their own business, GF has AMD with all over the place ready to sell 7nm chips, TSMC has AMD both on CPU and GPU and ofc noVidia on the GPU side and possibly some shitty SoC that noone will buy.
Moral of the story? Nobody will care about Intel's constipation, when there are traditional contracts in steak. Samsung won't give away its capacity or QC's capacity to favor intel and noone of the others are going to offer AMD's or noVidia's capacity.
con't
con't
Intel faces 3 problems right now.
1st they have to do the greatest deal of their existence in order to get some capacity in one of the busiest era in chip making
2nd they have to get EUV equipment and their order will be shipped last.
3rd they'll have to fix somehow their fabs to work with either 10nm or 7nm before anyone else moves to 5 or 3 nm in 2020-1-3 and they'll have to do so, because they will be forced to outsource their chip making... so yeah, spending gorillions to get to 7nm, and making it possible when everyone moved already to 5nm.
and last but not least, as an FPGA developer, I see that Altera is dying slowly due to the fab problems... Xilinx has already deals with TSMC iirc and does really great.
that being said, everyone else that uses Intel fabs(is this why micron and intel separated their ways?) will have to find a different fab if they want an up-to-date node, and not some 14nm BS.
also check this semiwiki.com
>If intel wants to use Samsung's 7nm node or TSMC's node, they have to do one of the following.
>1st port their whole 10nm cpu design, to TSMC's or Samsung's 7nm node and use their ASIC libraries
>or
>2nd port their whole 10nm CPU design to TSMC's or Samsung's 7nm node and use TSMC's or Samsung's ASIC libraries and tools.
But they'll have to do that anyways if they want to outsource their manufacturing completely, as OP suggest will happen.
I'm just saying closing a fab, firing thousands of people and writing off building worth millions of not billions is a too drastic step and there are better options like this article suggests (selling fab to GloFo or setting up joined venture.
Especially since, as you rightfully pointed out, 3rd party manufacturers have enough orders already so they probably can't squeeze Intel in even if they wanted to.
>OP suggest will happen.
it won't
Yeah, let's make another 14nm optimization and stay behind everyone else in the industry.
Kek imagine qualcomm going past them also. And Samsung hopping in for the drive.
Reminder back in 2014, Globalfoundries did something similar when they licensed Samsung's 14nm process:
globalfoundries.com
I'm sure Intel has enough money to license a process from someone. More importantly, they have lots of IP to horse-trade and arm-twist. But the fact they have to do so, is proof their management has spent the last few years chicken-fucking instead of properly running the business.
This is really good stuff, thanks.
Oh hey it's you. Thanks for the knowledge bomb again. Very informative.
Should be caped and added with the last one.
>is this why micron and intel separated their ways?
Intel and Micron are still working together on 3D Xpoint (Optane / QuantX).
Just the nand joint venture was dissolved (maybe because mission accomplished and they both have an adequate 3D nand product now?)
Actually, nevermind. I hadn't heard the news that their 3D Xpoint venture is splitting up, too.
