WYP
News Guru
Intel's 10nm and GlobalFoundries' 7nm are "more similar than they are different".
Read more about Globalfoundries 7nm process.
Read more about Globalfoundries 7nm process.
Last edited:
GlobalFoundries' 7nm process is set to make Intel's process leadership evaporate
- Thread starter WYP
- Start date
NeverBackDown
AMD Enthusiast
Intel will still have a lead. It'll be a smaller one but it's still ahead. 7nm GloFlo is still about equal to 10nm Intel. Once Intel hit 7nm they'll get a little bit further ahead.
This makes no sense. Their 10nm isn't even out yet, so when do you think they will get 7nm out?
WYP
News Guru
I haven't even heard Intel say a word about anything beyond 10nm. They have invested in EUV equipment, but it will be a while before they are using to create and sell products.
At this rate, Intel 10nm and Gloflo 7nm will be shipping within a quarter or two of each other. Not much point talking about Intel 7nm when even Intel hasn't spoken about it.
There is more to process technology than size and transistor densities, so there will still be improvements to be made, be it with material design or adjustments to transistor structure.
While FinFET does serve us well now, GAAFET (Gate All Around-FET) will be the next big thing, where the gate will have 360 coverage around the channel region. Right now FinFETs (Or Tri-Gate transistors) only cover it from three sides.
Intel's 10nm Process resembles a Dumpster Fire. They have produced large scale chips on on 2 different nodes and their performance shows zero improvement over their current 14nm+ Manufacturing Node. 10nm+ samples of the BB Sized "U" and "Y" chips used in laptops are expected by the end of 2018. It will be interesting to see what they do. But even if they do yield a marginal increase in performance there is nothing to say such an increase will be prevalent when the production line is redirected to full scale microprocessors.
The problems Intel has been experiencing with its 10nm node has claimed some real victims. Intel's Cannon Lake line of Desktop Microprocessors, which was promised to us in 2016, has been canceled. Intel will be heading straight to Icelake in early 2019.
AMD on the other hand has maintained that the 7nm Version of Epyc code named Rome will be released in the 2nd Half of 2018 with Desktop Processors following on in 2019.
Reports from within AMD and Global Foundries as well as industry experts who follow microprocessor development like Charlie Demerjian have stated the GloFo 7nm "is in Good Shape".
Here is a video of a talk given by Tom Faure explaining the various aspects of the masks used in GloFo 7nm Process.
https://www.youtube.com/watch?v=_UxDGH8uslM
NeverBackDown
AMD Enthusiast
Nothing to do with when. The post makes perfect sense.
Intel will still have a lead whether they are at 10 or 7. Doesn't matter what anybody else has. Everybody classifies their die size backwards so it looks better for investors. Whereas Intel makes sure it is actually the right size they claim it to be.
Once Intel hit 7nm they will have a little bigger margin yet again. It will take a while before anybody else can match them in density and other areas at the same node. Intel may not have as big of a lead but they still are ahead. People need to relax about the impending doom for Intel. It's not like it even matters much. We won't see much improvement in the real world between 10 or 7.
There is considerable discussion with the Electrical Engineering community about whether GAAFET is all its cracked up to be.
http://ieeexplore.ieee.org/stamp/stamp.jsp?arnumber=7890390
Kaapstad
Active member
To be frank I am really getting fed up reading anything about intels future plans as the company have been total a*******s with their very slow uptake of tech.
It took a near bankrupt AMD to release Threadripper and Ryzen to force intel to release anything interesting for the desktop.
It took a near bankrupt AMD to release Threadripper and Ryzen to force intel to release anything interesting for the desktop.
WYP
News Guru
Interesting read. It will be interesting to see how things develop. Samsung currently has GAAFETs on their roadmap, though as you said it remains to be seen how well GAAFETs will perform compared to FinFETs. By definition, we receive diminishing returns by surrounding more of the cover the channel, similar to adding more bits to a single channel in NAND (SLC-MLC-TLC-QLC etc).
Last edited:
here is another interesting read written by Scotten Jones for SemiWiki. It compares and details the 7nm process used by Global Foundries and the 10nm process used by Intel. It would appear the GloFo has erased the deficit it had in relations to Intel and is now running with the Big Dogs in the Semiconductor Foundrie Business.
https://www.semiwiki.com/forum/cont...al&utm_source=twitter.com&utm_campaign=buffer
It's debatable whether Intel even still has the process lead at the moment(Since 14nm they've suddenly changed their decades old policies of releasing technical data and now only release marketing mumbojumbo which gives no clear idea how well their process' are actually performing). Intels 14nm has had a very rocky ride with insiders reporting it's still only recently become economically equivalent to their 22nm parts due to the dismal yields(Which explains why Intel has kept producing and selling 22nm CPUs in emerging markets despite the claims to 14nm's economic benefits). Lets not forget, unlike every previous node, 14nm only got a paper launch with extremely low volume small die mobile parts that wern't actually in shipping hardware till at least 6 months after their "launch".
10nm is already several years behind schedule and by Intels own accounts so far unable to perform equivalently to their already flawed 14nm, with Intel still planning 14nm product launches till the end of 2018(Whiskey lake is a 2018-H2 part based on 14nm). I think it's safe to say unless the other fabs hit some significant walls, Intel has firmly lost the process lead. It'll be hard to say 100% given they now refuse to release any real technical data for their CPUs, but the lack of information alone should speak volumes(Pun not intended).
(I won't release paywalled information given the moral issues of such an action, but SemiAccurate have a few articles on this piece with plenty of info available for free access. So far, SA have never got any information or leak wrong, they have a very reliable array of moles in the industry and a long proven track record. Many of us in the industry or entering the industry use them to get around the press rubbish that usually litters tech sites. Their articles and leaks have often forced Intel to back track on their ridiculous marketing claims.)- https://semiaccurate.com/2017/12/20/state-intels-10nm-process/
10nm is already several years behind schedule and by Intels own accounts so far unable to perform equivalently to their already flawed 14nm, with Intel still planning 14nm product launches till the end of 2018(Whiskey lake is a 2018-H2 part based on 14nm). I think it's safe to say unless the other fabs hit some significant walls, Intel has firmly lost the process lead. It'll be hard to say 100% given they now refuse to release any real technical data for their CPUs, but the lack of information alone should speak volumes(Pun not intended).
(I won't release paywalled information given the moral issues of such an action, but SemiAccurate have a few articles on this piece with plenty of info available for free access. So far, SA have never got any information or leak wrong, they have a very reliable array of moles in the industry and a long proven track record. Many of us in the industry or entering the industry use them to get around the press rubbish that usually litters tech sites. Their articles and leaks have often forced Intel to back track on their ridiculous marketing claims.)- https://semiaccurate.com/2017/12/20/state-intels-10nm-process/
Last edited:
Similar threads
