Why is intel stuck on 14nm?

[Lord Szechenyi]

Hi

I've always wondered, why exactly is intel really staying on 14nm?

i mean seeing that the performance rivals somewhat closely to AMD, the only reason i can think of is since the barrier is at around 2nm, intel does not want to switch from silicon to something else, and so have used R&D to make better performance, with the same cpu size but only to prepare themselves for when they really are at the 3/2nm barrier.

other than, would there be any explanations? (other than: "Intel engineers dunno how to do smaller than 14nm" because i hardly believe they are not "smart" enough to go lower)

[Mohammad Saleh]

because they do not care about enthusiasts this is just how they were stuck on 4 cores for 10 years. They target the less tech savy market with promises of performance that was not event thinkable a year ago. 

[Guest]

It's been explained to absolute death.

 

https://www.pcgamer.com/what-went-wrong-intel/

[Lord Szechenyi]

1 minute ago, Den-Fi said:

It's been explained to absolute death.

 

https://www.pcgamer.com/what-went-wrong-intel/

im not going on there

[RollinLower]

1 minute ago, Lord Szechenyi said:

im not going on there

to be honest that's more of a problem with your ad blocker than the site. i'm on ABP and i can access it adless just fine.

[jaslion]

2 minutes ago, Lord Szechenyi said:

im not going on there

Odd I get on there just fine. Aaah you use 2 adblockers and one is adblock plus aka the one that is the easiest to detect. Just use ublock origin as your only one it's about as good as it gets on chrome.

[Guest]

2 minutes ago, Lord Szechenyi said:

im not going on there

Beginning to understand why you have this question, lol.

[Skiiwee29]

Site loads and works just fine for me with uBlock Origin. 

[Lord Szechenyi]

9 minutes ago, Den-Fi said:

Beginning to understand why you have this question, lol.

i fail to see the relation.

anyway, 55 ads is a lot for some TLDR

[CommanderAlex]

16 minutes ago, Lord Szechenyi said:

i fail to see the relation.

anyway, 55 ads is a lot for some TLDR

It all ranges from a new lithography size (14nm->10nm), poor yields from the silicon wafers, and also a new way of printing the circuits out on the wafers called EUV (short for Extreme Ultraviolet lithography). Since the new lithography is a lot different than what they were previously using 193nm UV light to print the circuitry into the silicon, this changes how they normally do things as they are delayed in mass producing chips and prioritize fixing the issues that arise. Don't forget that the smaller you go, the increased chances of malfunctions occur since we are on a tightrope with the laws of physics. 

 

One example of going really small is the smaller the lithography size, the shorter the gate's length is in a transistor...now Intel is at the point that electricity will actually jump this gate, causing a short and is unintended and thus, poor yields. Now also don't forget EUV is a whole new process for Intel as they straighten out the issues of this new lithography process. The whole semiconductor industry measures transistor "devices" different so Intel's 14nm is not the same is AMD's 14nm (really TSMC as they are the semiconductor fabrication company that AMD uses as AMD, Apple, and NVIDIA are what we call "Fab-less" as they do not own any semiconductor fabrication plants, rather they contract their work out to companies that specialize in this industry). 

[Lord Szechenyi]

1 hour ago, CommanderAlex said:

It all ranges from a new lithography size (14nm->10nm), poor yields from the silicon wafers, and also a new way of printing the circuits out on the wafers called EUV (short for Extreme Ultraviolet lithography). Since the new lithography is a lot different than what they were previously using 193nm UV light to print the circuitry into the silicon, this changes how they normally do things as they are delayed in mass producing chips and prioritize fixing the issues that arise. Don't forget that the smaller you go, the increased chances of malfunctions occur since we are on a tightrope with the laws of physics. 

 

One example of going really small is the smaller the lithography size, the shorter the gate's length is in a transistor...now Intel is at the point that electricity will actually jump this gate, causing a short and is unintended and thus, poor yields. Now also don't forget EUV is a whole new process for Intel as they straighten out the issues of this new lithography process. The whole semiconductor industry measures transistor "devices" different so Intel's 14nm is not the same is AMD's 14nm (really TSMC as they are the semiconductor fabrication company that AMD uses as AMD, Apple, and NVIDIA are what we call "Fab-less" as they do not own any semiconductor fabrication plants, rather they contract their work out to companies that specialize in this industry). 

Thank You!

now can you please explain to me why intel is struggling with this (which you just explained) but why AMD is Not.

[BTGbullseye]

2 minutes ago, Lord Szechenyi said:

Thank You!

now can you please explain to me why intel is struggling with this (which you just explained) but why AMD is Not.

Because Intel is new to the fabrication method, and TSMC (the fab AMD uses) is not. It's a simple matter of TSMC having a better fab than Intel.

[Lord Szechenyi]

2 minutes ago, BTGbullseye said:

Because Intel is new to the fabrication method, and TSMC (the fab AMD uses) is not. It's a simple matter of TSMC having a better fab than Intel.

thanks.

now new question

Why isn't intel switching to tsmc?

[BTGbullseye]

Just now, Lord Szechenyi said:

thanks.

now new question

Why isn't intel switching to tsmc?

They are. They just haven't gotten there yet. Also, TSMC is already at capacity on their 7nm node because of AMD.

[CommanderAlex]

3 minutes ago, Lord Szechenyi said:

Thank You!

now can you please explain to me why intel is struggling with this (which you just explained) but why AMD is Not.

Well Intel is struggling as they both design and manufacture the chips. They own the fabrication plants (I believe in Arizona, New York, and Israel) and TSMC already has history with this process. AMD used to be a fab company as well but decided to spin-off their fab plants into a company called GlobalFoundries. GlobalFoundries was planning on using EUV but discontinued in like 2017 because EUV is very expensive with all the equipment and processes used. 

1 minute ago, Lord Szechenyi said:

thanks.

now new question

Why isn't intel switching to tsmc?

There has been rumors from Intel's CEO stating that they may start placing orders with TSMC for their processes until they catch up with their own. 

[Lord Szechenyi]

Just now, BTGbullseye said:

They are. They just haven't gotten there yet. Also, TSMC is already at capacity on their 7nm node because of AMD.

i see.

one last question

why isn't amd copying intel?

to be more precise:

intel made the actual dies (or dye idk whats the plural) smaller on the cpu.

if amd does that wouldn't they get a huge performance boost?

[BTGbullseye]

5 hours ago, Lord Szechenyi said:

i see.

one last question

why isn't amd copying intel?

to be more precise:

intel made the actual dies (or dye idk whats the plural) smaller on the cpu.

if amd does that wouldn't they get a huge performance boost?

AMD already uses a chiplet design. Those chiplets are significantly smaller than Intel's chips. AMD can get much better yields per silicon wafer than Intel can because of the combination of smaller and greater number of chiplets. All they have to do is select the ones that have 3 out of 4 cores that pass all tests, and put 2 of them in the lowest end chip. Ones that have all 4 pass, get a pair going into the middle grade chip. Ones that pass 3 out 4 cores above average get 4 chiplets put into the high end chip. And ones that have effectively perfect scores on all 4 cores get 4 put into a top-end chip. (all chips also have the same memory controller chiplet that is made on the 14nm fab for all of the Zen 2 and Zen 3 chips)

 

If AMD switched to doing it like Intel does, they would be nowhere near as good of performance, and way lower yields from the fab.

[Lord Szechenyi]

3 minutes ago, BTGbullseye said:

AMD already uses a chiplet design. Those chiplets are significantly smaller than Intel's chips. AMD can get much better yields per silicon wafer than Intel can because of the combination of smaller and greater number of chiplets. All they have to do is select the ones that have 3 out of 4 cores that pass all tests, and put 2 of them in the lowest end chip. Ones that have all 4 pass, get a pair going into the middle grade chip. Ones that pass 3 out 4 cores above average get 3 chiplets put into the high end chip. And ones that have effectively perfect scores on all 4 cores get 3 put into a top-end chip. (all chips also have the same memory controller chiplet that is made on the 14nm fab for all of the Zen 2 and Zen 3 chips)

 

If AMD switched to doing it like Intel does, they would be nowhere near as good of performance, and way lower yields from the fab.

i see, thank you for the detailed explanation.

have a nice day

[smokeweedeveryday]

7 minutes ago, BTGbullseye said:

AMD already uses a chiplet design. Those chiplets are significantly smaller than Intel's chips. AMD can get much better yields per silicon wafer than Intel can because of the combination of smaller and greater number of chiplets. All they have to do is select the ones that have 3 out of 4 cores that pass all tests, and put 2 of them in the lowest end chip. Ones that have all 4 pass, get a pair going into the middle grade chip. Ones that pass 3 out 4 cores above average get 3 chiplets put into the high end chip. And ones that have effectively perfect scores on all 4 cores get 3 put into a top-end chip. (all chips also have the same memory controller chiplet that is made on the 14nm fab for all of the Zen 2 and Zen 3 chips)

 

If AMD switched to doing it like Intel does, they would be nowhere near as good of performance, and way lower yields from the fab.

fab-ulous