Electronics Weekly Electronics Design & Components Tech News
One is that x86 is not going to get into mobile and the other is that Intel’s foundry business can’t grow without making ARM-based ICs.
x86-for-mobile was a kind of crazy crusade based on the belief that x86 should rightfully inherit the Earth.
Intel, better than anyone, should appreciate that trying to replace an established architecture is a fools errand.
It’s an errand which cost Intel dear – some $10 billion.
For foundry, the ARM licence recognises that Intel’s foundry business can’t grow unless it makes ARM-based ICs – they are too ubiquitous.
The surprising, or maybe not so surprising, thing about Intel’s foundry business is that despite having, arguably, the best process on the planet, Intel is not in the top 10 for foundry.
Intel only claims six foundry clients and one of those, Altera, is owned by Intel, another, Spreadtrum, is obliged to use Intel fab under an investment deal, two other announced customers, Achronix and Netronome have minuscule volumes, and only one, LG could be regarded as a genuine outside customer with substantial volumes.
Intel says it has other customers it can’t name.
Intel foundry is said to be more expensive than other foundries and its process more difficult to design for. One somehow suspects that customer service may also be less than brilliant.
So one thing Intel could do is to make it easier for people making ARM-based chips to use Intel fabs. And this is what the ARM physical IP licence and POP pack aim to do.
Will the strategy help fill Intel’s fabs?
Intel’s long history of failed diversifications says No.
Yes indeed, AnotherDavid, the proof of the pudding will be in the eating and now we’ll see if Intel’s 10nm process delivers any advantage over TSMC’s, Samsung’s and GloFo’s.
Thanks Quicksilver, extremely interesting thoughts too.
I think wafer price is big factor. However, Intel run proprietary and secret process recipes which are tuned to maximise the yield and performance of x86 devices.
These processes will not be optimised for general logic or embedded memory applications and thus yield and performance of these devices will be compromised? In addition I have never heard of an Intel PDK being talked about in public. Since this is the crown jewel of their x86 advantage I would doubt they would share the details with foundry customers?
Until Intel provide an open PDK, (like TSMC et. al.) then I doubt that they will ever be a serious foundry player.
Just my thoughts.
Q.
Its going to be interesting, Intel processes vs TSMC & Samsung processes. There is more chance of a direct comparison than Intel x86 sever chip vs. ARM sever chip.