I'm curious too, can this only be licensed from UK universities or western academic centres?
Chinese semiconductor industry
- Thread starter Hendrik_2000
- Start date
- Status
Eh. Intel 7 is like Intel's 3rd "10 nm" node. I try to differentiate between Intel's early 10 nm from Intel 7. Intel started making the better density than TSMC argument well before they rebranded their latest 10 nm iteration to Intel 7.
To be clear, I don't think Intel 7 is "bad" or that it's somehow worse than 10 nm. I mostly just wanted to address the density vs transistor performance angle, because I've seen it pop up quite often over the years, and I think it's a misleading argument insofar as how we gauge process node performance.
Yeah we'll see. In theory I think their new strategy is right for their situation but they're pivoting like 5 years later than they should have.
I know how the charts work. But when potential clients are looking at different fabs that's not what they're looking at first and foremost anymore. One other thing worth considering is that area was a really big deal when device and circuitboard footprint was a fundamental physical constraint parameter for electronic products. Nowadays the transistors are so small physical footprint differences between say 10 nm and 7 nm don't really matter as much.
Yeah, I mean up to a point it becomes hard to do a 1 to 1 comparison because no one is going to Intel and TSMC with the same chip design and then comparing what comes out the other end.
I appreciate your appreciation
Yeah, sorry if gave the wrong impression. I really just wanted to tackle one very specific thing about the Intel 10 nm vs TSMC 7 nm discussion. I agree that some people here are crapping on Intel quite a bit more than they deserve (though of course everyone craps on them because they went from industry lead to helpless for a half decade at the snap of a neck, and that part of the crapping they've kind of earned).
this is 100% confirmed news, not rumour... this post has been deleted...
@FairAndUnbiased looks to have some business experience here: Do you ever meet a company who invest a ton of money in buying two companies and expanding their manufacturing capacity for nothing? How they are going to sell and recover their investment if everyone prefers to buy ASML scanners even their legacy ones and even if the money comes directly from the government?
One part of their "documents " its says that they developed "20 prototypes" of the "special project 02", so in other to work in tandem they should at least shipped some of those, R&D is different than a final product. ASML created like a million prototype EUV scanners before shipped it final product.
yes
I think the length of time in which the gap could "get bigger" and how much the "longer term" is, entirely depends on when domestic EUV is available...
Short term (2-3 years) up to medium term (3-5 years), I would agree with you.
But after medium term (3-5 years)..... well, it depends on when domestic EUV is ready and how well prepared the various players are to hit the ground running once they have EUV in their hands to rapidly iterate to the most sophisticated node they are able to fab.
Only time will tell.
so I'm not on the business side but from what I understand, business acquisitions are taken very, very carefully and slowly. You never really know what you are going to get, so you need due diligence, and even then, the risk has to be weighed vs other ways to spend the money.
even if you came upon tons of government money, you still have to choose between spending it on your core business vs. an acquired business.
To be quite honest I don’t know how much more room there is for a gap to form. TSMC’s 3nm got delayed. And each shrink after is going to require a radical new transistor design. I think things are only going to get harder and slower for the node shrink game from here on out, and that’s even with high NA coming down the block. The demand pull factor for future nodes is also diminishing. Most fab clients these days have more compute power than they really know what to do with. If it didn’t come at any extra cost they’d all be happy to take more compute power anyways but the cost curve for the smallest nodes are doubling at each new iteration. I think there’s going to be a point where even Apple may not care to pay for the latest and greatest node at any price.
there's also the . at some point simple die shrinks or even new transistor designs may not be enough to deal with the problems of thermal stress and leakage voltage. there will need to be new paradigms such as 3D packaging, , , etc.
Is it really that big of a disadvantage to fall behind in advanced nodes? Last month I built a new PC for my mom. A 12th gen i7 (10nm ESF) to replace her 10 years old 3rd gen i5 (22nm). The only reason for the new PC was that my mom got a new 62 megapixels camera, and she hates having FastRAWViewer in her workflow. So a faster processor was needed to preview RAWs quicker in DXO Photolab. However, the 62 megapixel sensor of the new camera already out-resolves most full frame optics, thus there's no reason ever to upgrade to a higher megapixel camera body (a bigger format camera is physically unusable for my mom), and no reason ever to get a better processor for post-processing.
If Huawei can come up with fully domestic 7nm chips (I know it's years away) I'll switch to Huawei phones. I see zero reason for better processor on my phone now that a phone's power consumption is dominated by screen backlighting. I'll drag my father to the Huawei camp too. He's on the market for a new phone and his biggest requirement is a good microphone to use with karaoke apps.
For business/government computers, my impression is that most of the trendy applications for processing power are pretty parallelizable, and China is so far ahead of everyone else in renewable energy that efficiency for non-mobile applications are desirable but not necessary (except crypto mining, which is banned in China anyway).
I used to think the capacity to cut a country off from the newest and greatest GPUs to trigger a nerd revolt was a very dangerous power in the hands of US government, but gaming GPUs were practically unavailable for two years and no government was overthrown because of that, so now I don't know.
- Status