David Schor: Intel 10nm in big problems

[LTC8K6]

Hardly.. EUV cuts that advantage Intel has....

Intel has done tons of EUV research and testing, and Intel is apparently the largest purchaser of EUV tooling.

 

[William Gaatjes]

SAQP perhaps?

According to article on Toms hardware it is. I was reading about it.
Difficult stuff.
But it cannot be just that. Gives me an itch. I cannot explain it.

This video makes it look easy but there are so many variables to be controllled.

 

[DisEnchantment]

Intel has done tons of EUV research and testing, and Intel is apparently the largest purchaser of EUV tooling.

According to Digitimes report, Samsung is the largest purchaser of EUV tooling.
Their investment in the last year is equal the rest of the industry combined.

 

[Lee Saxon]

I'm astonished this has happened again. ASTONISHED. Who could've possibly predicted.

 

[wahdangun]

I'm astonished this has happened again. ASTONISHED. Who could've possibly predicted.

Yeah from leading to almost behind everyone,

It's like rabbit vs turtle race

 

[dark zero]

Intel has done tons of EUV research and testing, and Intel is apparently the largest purchaser of EUV tooling.

Indeed, we can't deny that, but they are not fully using it!

 

[oak8292]

TSMC's processes are usually tailored for Mobile SoCs. How high could they clock this time?

Actually they have developed two libraries for 7nm.

"TSMC set another industry record by launching two separate 7nm FinFET tracks: one optimized for mobile applications, the other for high performance computing applications."

http://www.tsmc.com/english/dedicatedFoundry/technology/7nm.htm

I doubt that they developed 7nm HPC without a committed customer. I will guess the 7nm HPC process is for Qualcomm. Pure speculation since they have a server processor design and they are still a fairly large TSMC customer at 7% of revenue in 2017.

How will a 14nm x86 processor stack up against an 7nm ARM processor?

 

[french toast]

I read somewhere they are having to use 6x patterning in some features..ouch!.
Basically they got too ambitious whilst also trying to skip EUV...they got it very wrong.
Looks like global foundries got it just right, designed the features to be just big enough to not need EUV and not need so much cobalt = much easier to manufacture even if it offers slightly worse density.

Also easy port to EUV in late 2019.

 

[beginner99]

I doubt that they developed 7nm HPC without a committed customer. I will guess the 7nm HPC process is for Qualcomm.

And /or Nvidia

 

[DisEnchantment]

Actually they have developed two libraries for 7nm.

"TSMC set another industry record by launching two separate 7nm FinFET tracks: one optimized for mobile applications, the other for high performance computing applications."

http://www.tsmc.com/english/dedicatedFoundry/technology/7nm.htm

I doubt that they developed 7nm HPC without a committed customer. I will guess the 7nm HPC process is for Qualcomm. Pure speculation since they have a server processor design and they are still a fairly large TSMC customer at 7% of revenue in 2017.

How will a 14nm x86 processor stack up against an 7nm ARM processor?

I missed that part. Thanks.

I found this though.
https://www.eetimes.com/document.asp?doc_id=1332965

For its part, TSMC presented an L1 cache compiler now available for its 7-nm process that can operate at data rates up to at least 4.4 GHz. Its 16-nm L1 memories topped out at about 3 GHz. The higher data rate is particularly useful for matching the speed of high-speed CPUs and mobile applications processors.

this one is interesting
https://www.eetimes.com/document.asp?doc_id=1332270

Samsung said it has processed close to 200,000 wafers with EUV lithography technology since 2014 and has recently seen visible improvement with EUV technology, such as achieving 80 percent yield for 256 Mb SRAM.

How bad would it be for logic?


I hope Intel can bounce back soon. We need the competition.
This failure is inevitable when people at the helm badly needed a reality check.
Next year could be stiffer competition.
Samsung's improved EUV opens up other possibilities for improvements.
TSMC's CLN7FF+ also brings considerable improvements.

But think about it we are stuck for so long on this 16/12/14 nm, we need this 7nm, because it opens up many possibilities for new designs.

 

[krumme]

This tight 6x patterning will it ever give acceptable yield? Man. Suddenly euv seems less complicated. Who would have thought that. Some major asumptions about the organizations capabilities was wrong here. It seems those asumptions was more based on historical improvements than physical facts.