Samsung's win, Intel's pullback, and a shifting chips landscape
Posted by snowfordessert@reddit | hardware | View on Reddit | 14 comments
Posted by snowfordessert@reddit | hardware | View on Reddit | 14 comments
ShoutOfDawn@reddit
intel ability to stumble is truly unmatched, is there any more high-end chips node left now that 18a is canceled?
cjj19970505@reddit
18A is not cancelled. It has internal and external customer and going to be stretched to the end of this decade. The "cancelled" news you saw is that Intel is rumorly no longer pitch it for external customer. If it's true that means, no significant effort will be put to win more customers for this node, which doesn't mean they are not delivering this node to existing external and internal customers.
As far as we know PTL(18A on compute tile) is still planed to be released the end of this year with at least 1 sku
Exist50@reddit
"Existing external customers" seems to be pretty much negligible.
Jellym9s@reddit
Did you know Nvidia actually runs some stuff on 18a? And so does Microsoft. Through Pentagon programs.
Exist50@reddit
Nvidia has test chips. That doesn't count as a customer. And we haven't heard from Microsoft in a while, so very possible they've dropped out. How much they were in for to begin with is also unknown.
Again, Intel has explicitly stated that they have no significant external customers for 18A.
Jellym9s@reddit
In the letter published recently, they have customers working with them on 18A-P, we just don't know who they are. Unless they're lying but they can be sued for it... But definitely don't expect meaningful external volume on 18A, the base node, it was always hopeful but never realistic. 18A is mainly meant for Intel Products to use.
Any node Intel's going to deliver to big customers has to be worked to their specifications, and 18A-P is just about the point when you can start doing that, after 18A is out and customers can see what products on the family perform to.
Then 14A is a whole node that can be designed from start with customers in mind. That one I am hopeful for, but that's also the node where Intel said they won't continue it if they can't get customers.
Exist50@reddit
They don't have customers, it sounds like. They have companies who are maybe interested in the node, but they sure aren't paying for it yet.
Well they've gotten away with an awful lot so far...
This exact argument was made for 18A. Intel 3 and 20A were supposed to be the pipecleaners to get the infrastructure ready for external customers, and 18A was the node that would see mass adoption. Now we're repeating the same claims but for 14A.
And if 18A isn't interesting, why would any of its derivatives be?
Jellym9s@reddit
Usually with foundries, a new node gets customer zero to sign on, then everyone else tries it. So for TSMC that's usually either AMD, Apple, or Nvidia. For Intel's case, it will be Intel. Intel's products coming out on 18A later this year, probably December. Depending on how that turns out, it sets the story for 18A-P.
Exist50@reddit
TSMC sells out all of their initial production. Usually that goes to Apple, but when circumstances don't align, there are plenty of other companies (like AMD) happy to fill their place. That's very much not what's happening with Intel. Even Intel's own product teams try to avoid using Intel nodes where possible, for the same reasons that drive off external customers.
The story for the 18A family is already set. If customers don't trust a node to use it until silicon is already shipping, they're probably not going to use it at all.
Raikaru@reddit
What? When was the last time AMD bought a new node? I swear Qualcomm or Mediatek would be way more likely than AMD
Exist50@reddit
AMD's one of, if not the first N2 customer. They've already announced they'll be using it for Zen 6 / Venice.
DerpSenpai@reddit
4+4.... oof worse than the budget X Elite gen 2 die
RandomFatAmerican420@reddit
The real news of cancellation was Intel saying openly it may cancel 14A, and essentially stop competing at all with making bleeding edge chips.
ProfessorNonsensical@reddit
So basically their customer base is shrinking, bad news.
Most companies looking to compete in the future will not be buying these, the penny pinching foreign companies will but that doesn’t sustain their massive overhead, hence the layoffs.
It is as bad as it looks.