-
Competitor rules
Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.
Intel and Apple beat AMD to 3nm?
- Thread starter Grim5
- Start date
Soldato
- Joined
- 17 Aug 2003
- Posts
- 20,160
- Location
- Woburn Sand Dunes
Apple have always been first to a new node for years, now. Nobody else has their buying power. Intel on the other hand..
Intel conceding their fabs are crap?
Apple no date given but probably before Intel, apple always get first dibs on the whole process.
Intel Mass production beginning end of 2022 at the earliest, so 2023, unless delayed, everything Intel does is delayed these days.
AMD's leaked roadmap for 3nm Zen 5 is 2023.
https://www.techspot.com/news/89855-amd-3nm-based-zen-5-processors-appear-leaked.html
Apple no date given but probably before Intel, apple always get first dibs on the whole process.
Intel Mass production beginning end of 2022 at the earliest, so 2023, unless delayed, everything Intel does is delayed these days.
AMD's leaked roadmap for 3nm Zen 5 is 2023.
https://www.techspot.com/news/89855-amd-3nm-based-zen-5-processors-appear-leaked.html
Last edited:
Intels's new architectures are modular to the max, every component can be manufactured by a different Fab on a different process node - they just use whomever has the best node at the time of production for the individual components. It's worth noting that it's entire CPU will not be 3nm, that's not how their architecture works - only certain components on the cpu will be
Like Zen 3 and Zen 2 before it. Ironic that Intel were quite derogatory about AMD's "Glue"
They are not going to be using their own 10nm for the main cores are they?
They are not going to be using their own 10nm for the main cores are they?
Won't know until they announce a new product, all we have is their CEO saying they'll use whichever process node provides the best outcome for them - so I assume if that means tsmc 3nm then so be it - however I also know that in 2022 TSMC's 3nm will still be an immature node that won't have yields for big cores and so I assume only small chips can be made on it at that time and so if I had to guess I'd think Intel is using them to make small cores only or even low tdp mobile parts like a 15w cpu
Changing node still won't happen in few months and is like half+ year process.
Instead of normal year+ process...
And performance wise that actually means some compromise in design and AMD competing in more even terms against Intel:
More easy to port to different nodes design can't be fully tailored for node to maximize potential.
Good reasoning would suggest that but for them its about beating the competition at any cost, they have an almost narcissistic mindset to keep the competition down and out.
Intel need their halo products to be 3nm to compete with AMD's because AMD are so far ahead of even Intel's 10nm products its not even funny and Intel are not going to let AMD take even another 5% of Intel's data centre and mobile market share, so even if 3nm yields are so bad it actually costs Intel money, a lot of money to produce halo chips with it they will, its not about the money, Intel have mountains of that to literally burn, its about doing everything they can to stop AMD from growing in to a true competitor, a true competitor is someone who also has mountains of money to burn.
AMD are very good, much better than Nvidia and Intel put together at R&D'ing class leading products cheaply, AMD are in a much better position than they were 5 years ago, they just need to keep doing what they are doing, keep making Intel set fire to their cash, eventually, it may take a decade or more but Intel will burn their reserves out.
Modular to the max probably just means they have concentrating on interconnects.
Both times AMD did really well, the interconnect was a major reason: first HyperTransport, now Infinity Fabric.
So it is natural Intel would go down that route too.
So chiplets connected together.
The new CEO saying they'll use whatever node provides the best outcome for them is radical for Intel, as the various chiplets can come from different places.
However, that doesn't mean that porting designs is going to get much easier. Traditional Intel designs were very very tightly coupled to their nodes. Even if they are about to get much more versatile and generic, the decision to make a part at a given manufacturer's node (so Intel or TSMC) has to made a long long time in advance.
Risk management would imply that they really should design some parts to be dual-sourced right from the start.
Intel do have the volume for this.
Whether the Intel manufacturing will still put internal politics and pride obstacles in the way is the big question.
Also, TSMC are not stupid so they will have to balance taking Intel's money with seeing that Intel manufacturing as a rival who just wants to use TSMC until they get back on their feet.
Hard to say hwo they will approach this: maybe get Intel addicted to TSMC and hope their own manufacturing continues to let Intel down.
And while I can assume that TSMC doesn't consider Intel a long-term partner, Intel have big lobbing power in the USA so maybe TSMC sees building a few fabs in the US will be to their advantage against any American protectionism.
Both times AMD did really well, the interconnect was a major reason: first HyperTransport, now Infinity Fabric.
So it is natural Intel would go down that route too.
So chiplets connected together.
The new CEO saying they'll use whatever node provides the best outcome for them is radical for Intel, as the various chiplets can come from different places.
However, that doesn't mean that porting designs is going to get much easier. Traditional Intel designs were very very tightly coupled to their nodes. Even if they are about to get much more versatile and generic, the decision to make a part at a given manufacturer's node (so Intel or TSMC) has to made a long long time in advance.
Risk management would imply that they really should design some parts to be dual-sourced right from the start.
Intel do have the volume for this.
Whether the Intel manufacturing will still put internal politics and pride obstacles in the way is the big question.
Also, TSMC are not stupid so they will have to balance taking Intel's money with seeing that Intel manufacturing as a rival who just wants to use TSMC until they get back on their feet.
Hard to say hwo they will approach this: maybe get Intel addicted to TSMC and hope their own manufacturing continues to let Intel down.
And while I can assume that TSMC doesn't consider Intel a long-term partner, Intel have big lobbing power in the USA so maybe TSMC sees building a few fabs in the US will be to their advantage against any American protectionism.
TSMC would be smart to move their operation to the US, with China on the march this is probably what they are doing.
If it comes to it TSMC will want to be able to shut down their operation in Tiwan and become a US based company.
Intel. Apple, Nvidia, AMD should all buy significant shares in TSMC to help them do that.
If it comes to it TSMC will want to be able to shut down their operation in Tiwan and become a US based company.
Intel. Apple, Nvidia, AMD should all buy significant shares in TSMC to help them do that.
If you're worried about that just have them buy ASML. Who controls EUV machines will control leading chipset production for the next 5 years, at least until x-ray etching becomes suitable for mass production.