Intel is a dead company walking if Pat doesn’t step up.
It's not about the amount of engineers, it's about the amount of R&D and basic research China has to essentially redo. Which includes eventually training said engineers with that knowledge.
And the corruption is less about bribes and the likes and more about businesses which don't have any expertise in the whole semiconductor area jumping onto it since it's there where all the government money is. For years a lot of funding went nowhere since the people receiving the funding literally didn't know how to actually invest into advances in semiconductor. That's improving as the knowledge what's required increases, but this whole process is essentially repeating for every single speciality ancillary industry to some degree. That's what makes the whole development very costly for China.
China is now, apparently, 3 years behind the leading TSMC node. The leading node in China today is equivalent of TSMC N5 from 3 years ago. The gap is shrinking.
Where you are seeing claims of China having anything like N5? They are barely able to make something like N7 in quantity - yields are bad, cost is high.
I think they CAN make something equivalent to N5 since they have access to better DUV machines than TSMC had for its all DUV N7, but after that they're stuck. They can't go any further until they're able to successful build their own equipment. They are years away from being able to do so.
They will continue to fall further behind through this decade, and probably into the next, because they have no path to get beyond N5 without matching the effort ASML did - which took the resources of hundreds of companies large and small and around 15 years to manage. Even if China had blueprints and access to a working EUV scanner they'd be little closer to matching it, because a lot of new technology was developed for subunits throughout it - it is probably the most complex device humans have ever mass produced (stuff like CERN's detector is more complex but there is only one of it) You can't copy something if you don't have access to information about how it was made, and what was learned during its development.
I remember reading in article a couple years back about Bai Lan - let it rot. Intel has much bigger fish to fry than worrying about China catching up in some indeterminant number of years.
They’ve got a strong first mover advantage though, that counts for something (assuming they don’t squander it).
ChatGPT is basically synonymous with large language models.
And here is an actual teardown for reference:It's based on teardown of the Huawei Kirin 9010
And here is an actual teardown for reference:
Huawei’s Kirin 9010 Is a Reality Check for China’s Semiconductor Ambitions
Sanctions on China’s chip industry were supposed to slow technological advancement, but are they working? Huawei’s latest smartphone may hold answers.www.ifixit.com
The argument that China will fall behind is - sorry - preposterous.
China catches up, and then gains lead. That's the rule. Industry after industry, product after product.
Lol. Of course they will catch, one can argue how much time they will need to do it. They have far, far more qualified manpower than you can think. And if you think the factor about China's growth is only "cheaper but worse copies" you are deluding yourself.It is easy for China to copy stuff that's not complicated. ASML's EUV scanner took a lot more money (adjusted for inflation) and man hours of effort than the Manhattan Project. Tell me again how China will easily copy it.
You're assuming that because China can make a cheaper EV that they can do the same with anything.
It is easy for China to copy stuff that's not complicated. ASML's EUV scanner took a lot more money (adjusted for inflation) and man hours of effort than the Manhattan Project. Tell me again how China will easily copy it.
You're assuming that because China can make a cheaper EV that they can do the same with anything.
There is not one and only approach to EUV. If ASML engineers were told "start over", knowing what they know, they would not likely build the same machine they sell now.
Here is an article about Japanese scientists who came up with a simpler (and cheaper) approach to EUV:
Japanese scientists develop simplified EUV scanner that can make production of chips considerably cheaper
OIST's simplified EUV litho system uses two mirrors instead of six.www.tomshardware.com
There is not one and only approach to EUV. If ASML engineers were told "start over", knowing what they know, they would not likely build the same machine they sell now.
Here is an article about Japanese scientists who came up with a simpler (and cheaper) approach to EUV:
Japanese scientists develop simplified EUV scanner that can make production of chips considerably cheaper
OIST's simplified EUV litho system uses two mirrors instead of six.www.tomshardware.com
You're in the wrong thread. In another, there's nothing better in the silicon world than Intel's latest and upcoming nodes. Silicon design bureaus will fly to them.The easiest thing is to separate It, If they can't keep up, but then It's likely It will start to lag even more and they would need a lot of customers to pay for R&D and to keep It operational. Would there be that much demand for Intel's process? I am not sure.
How did they give them a softball target exactly? The jump from zen 4 to zen 5 isn't that big, sure. But both of them just jumped 10% with new windows updates. So isn't this more about the difference between intel and and than AMD to AMD? Harder for intel to catch the increased targets of both processors.Must be worse off than we thought.
Intel working with bankers to present board with strategic options
Hopefully their performance is competitive in October... though AMD seemingly gave them a softball target in the consumer space with the Ryzen 9000 series, the story is different in enterprise/DC.
And the real target is Z5X3D. On laptop LNL is a properly solid part, ARL-U/H share similar issues to MTL but is decent enough to not bleed anything.How did they give them a softball target exactly? The jump from zen 4 to zen 5 isn't that big, sure. But both of them just jumped 10% with new windows updates. So isn't this more about the difference between intel and and than AMD to AMD? Harder for intel to catch the increased targets of both processors.
It's difficult to change a first impression. Lots of reviews were "skip this gen" and "meh" in overall assessment. While enthusiasts may be more in tune with recent developments, launch day reviews will linger in the minds of many.How did they give them a softball target exactly? The jump from zen 4 to zen 5 isn't that big, sure. But both of them just jumped 10% with new windows updates. So isn't this more about the difference between intel and and than AMD to AMD? Harder for intel to catch the increased targets of both processors.
Made me remember the Bulldozer days, but even Bulldozer had a saving grace being an "Octacore".It's difficult to change a first impression. Lots of reviews were "skip this gen" and "meh" in overall assessment. While enthusiasts may be more in tune with recent developments, launch day reviews will linger in the minds of many.
Which is why it is also important for Intel to make sure the October new desktop CPU/platform launch is problem-free. Because the Raptor Lake fiasco has cost them mindshare and trust. Trust once broken is not easily regained.
The first impression is that it's not stronger than zen4, but zen 4 already is stronger than intel, especially considering their current degradation woes. So yeah they night nit flock to zen 5 yet but it's just more zen4It's difficult to change a first impression. Lots of reviews were "skip this gen" and "meh" in overall assessment. While enthusiasts may be more in tune with recent developments, launch day reviews will linger in the minds of many.
Which is why it is also important for Intel to make sure the October new desktop CPU/platform launch is problem-free. Because the Raptor Lake fiasco has cost them mindshare and trust. Trust once broken is not easily regained.
If this is really the case, I don't see a positive for the either side. They'll likely end up fading into irrelevance or at best a shell of what it was after the split. The fabrication side still needs a lot of work to be competitive and gain trust, and the design side is significantly behind too. Who would want either of them?Intel is weighing options to split the foundry:
(From Bloomberg)
If this actually happens, then Gelsinger is going to be kicked out. That was pretty much the deal to get the bonus and for him coming in, the IDM 2.0 strategy and getting leadership all over again.Intel is weighing options to split the foundry:
(From Bloomberg)