You can't get rich if you pay taxes. The taxes are for the poor to redistribute the wealth to the rich. Just think who imposed taxes historically - kings and lords.
One way to look at it would be that Intel was so far ahead of their competition they may have strategically decided to slow down, maybe even skip a process node or two in order to save on capital investment.
The Roman Empire was so far ahead of everyone else (but the Persians), so they decided to strategically slow down... until they disappeared (by 1453) completely.
The historical assumptions you make are all incorrect. In Intel's context, if the Roman Empire decided to slow down then another empire (e.g. Parthians/Samsung) would over take them - that's not what happened. And the Persians were gone with Alexander to boot. On topic, Intel has capital and will spend it to stay relevant and modernize their processes. But node number means so little now, that their 14 nm+++++ with a high power design is competitive enough still - especially when combined with their processor IP. Intel has real competition, but is still very relevant.
Depends.....Samsung 7nm is out right now?? NO. TSMC 7nm is out for high power SKUs?? NO yet, we'll see what will be AMD 7nm poison, for now only slides. Depends on your manufacturing target, are these fine processes fast as 14nm on high power devices ?? actually NO.
TSMC will be denser. Using the 7nm figures from https://www.semiwi ki.com /forum/content/7926-samsung-vs-tsmc-7nm-update.html , we get 175 million transistors/mm^2 for TSMC 5nm vs 112.5 for Samsung.
Interesting because it’s been stated numerous times that FinFET isn’t suited for 5nm, and that companies were investigating at least three other methods.
@Anton: the first table (showing improvements in power consumption, performance etc.) is labeled as sourced from "manufacturers"; could you please indicate which comparisons are from which manufacturer? Thanks!
"Furthermore, the company promises an increase in logic area efficiency of up to 25%."
Of course if it really be 5nm vs 7nm, the increase would be 100% ( (7/5)^2 - 1 ).
I know, I know, it is all lies, gate lengths (historical measure) are still in 40nm range and regulators are nowhere in sight to stop all this marketing fraud.
Seeing as the foundries' customers are high level engineers with advanced degrees in semiconductors, I really don't think you need to be waiting for regulators to come in and protect them from being fooled. Anyone making decisions about what fab to use is well aware of how big transistors really are.
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
Facebook
Twitter
RSS
21 Comments
Back to Article
AshlayW - Wednesday, April 17, 2019 - link
Poor Intel lol.Dr. Swag - Wednesday, April 17, 2019 - link
Actually, Intel is quite richjospoortvliet - Thursday, April 18, 2019 - link
... and probably doesn't pay much taxes either ;-)Zingam - Saturday, April 20, 2019 - link
You can't get rich if you pay taxes. The taxes are for the poor to redistribute the wealth to the rich.Just think who imposed taxes historically - kings and lords.
FreckledTrout - Thursday, April 18, 2019 - link
Intel is still on 14nm and yet there stock keeps going up. These delays of moving to a new process node don't seem to have impacted them as of yet.surt - Thursday, April 18, 2019 - link
One way to look at it would be that Intel was so far ahead of their competition they may have strategically decided to slow down, maybe even skip a process node or two in order to save on capital investment.Zingam - Saturday, April 20, 2019 - link
The Roman Empire was so far ahead of everyone else (but the Persians), so they decided to strategically slow down... until they disappeared (by 1453) completely.genekellyjr - Saturday, April 20, 2019 - link
The historical assumptions you make are all incorrect. In Intel's context, if the Roman Empire decided to slow down then another empire (e.g. Parthians/Samsung) would over take them - that's not what happened. And the Persians were gone with Alexander to boot. On topic, Intel has capital and will spend it to stay relevant and modernize their processes. But node number means so little now, that their 14 nm+++++ with a high power design is competitive enough still - especially when combined with their processor IP. Intel has real competition, but is still very relevant.Gondalf - Saturday, April 20, 2019 - link
Depends.....Samsung 7nm is out right now?? NO. TSMC 7nm is out for high power SKUs?? NO yet, we'll see what will be AMD 7nm poison, for now only slides.Depends on your manufacturing target, are these fine processes fast as 14nm on high power devices ?? actually NO.
Too much "NO" and a lot of marketing instead.
Someguyperson - Wednesday, April 17, 2019 - link
So, although TSMC beat out Samsung to the 7nm node, Samsung should beat out TSMC when it comes to the 5nm node?FunBunny2 - Wednesday, April 17, 2019 - link
it all depends, it seems, on what ASML et al do. the tool makers are the bottleneck.shabby - Wednesday, April 17, 2019 - link
Tool makers are the bottlenecks?Intel: hold my beer...
ZolaIII - Thursday, April 18, 2019 - link
Random errors in the EUV layers is a brake... Samsun has better IP portfolio for third party costumers and better tool chain.ZolaIII - Thursday, April 18, 2019 - link
Samsung*Felixthecat256 - Thursday, April 18, 2019 - link
Tsmc announced the 5nm process was complete on April 9th. Only time will tell who is better.Wilco1 - Thursday, April 18, 2019 - link
TSMC will be denser. Using the 7nm figures from https://www.semiwi ki.com /forum/content/7926-samsung-vs-tsmc-7nm-update.html , we get 175 million transistors/mm^2 for TSMC 5nm vs 112.5 for Samsung.Sychonut - Wednesday, April 17, 2019 - link
Looking forward to 14+++++++++++.melgross - Thursday, April 18, 2019 - link
Interesting because it’s been stated numerous times that FinFET isn’t suited for 5nm, and that companies were investigating at least three other methods.eastcoast_pete - Thursday, April 18, 2019 - link
@Anton: the first table (showing improvements in power consumption, performance etc.) is labeled as sourced from "manufacturers"; could you please indicate which comparisons are from which manufacturer? Thanks!peevee - Thursday, April 18, 2019 - link
"Furthermore, the company promises an increase in logic area efficiency of up to 25%."Of course if it really be 5nm vs 7nm, the increase would be 100% ( (7/5)^2 - 1 ).
I know, I know, it is all lies, gate lengths (historical measure) are still in 40nm range and regulators are nowhere in sight to stop all this marketing fraud.
saratoga4 - Saturday, April 20, 2019 - link
Seeing as the foundries' customers are high level engineers with advanced degrees in semiconductors, I really don't think you need to be waiting for regulators to come in and protect them from being fooled. Anyone making decisions about what fab to use is well aware of how big transistors really are.