An old 2012 article I stumbled on. There was no future for Arm and its competitors.
For your amusement Intel re Arm June 2012
- Thread starter blueone
- Start date
Here is my favorite:
Phones/PCs belong on desks not in pockets!
Intel Paul Otellini on working with Apple:
“We ended up not winning it or passing on it, depending on how you want to view it. And the world would have been a lot different if we’d done it. The thing you have to remember is that this was before the iPhone was introduced and no one knew what the iPhone would do. At the end of the day, there was a chip that they were interested in that they wanted to pay a certain price for and not a nickel more and that price was below our forecasted cost. I couldn’t see it. It wasn’t one of these things you can make up on volume. And in hindsight, the forecasted cost was wrong and the volume was 100x what anyone thought.”
And of course Apple ended up partnering with TSMC which disrupted the foundry business with the half node process development methodology which put TSMC solidly in the process technology lead ahead of Intel and Samsung.
Phones/PCs belong on desks not in pockets!
Intel Paul Otellini on working with Apple:
“We ended up not winning it or passing on it, depending on how you want to view it. And the world would have been a lot different if we’d done it. The thing you have to remember is that this was before the iPhone was introduced and no one knew what the iPhone would do. At the end of the day, there was a chip that they were interested in that they wanted to pay a certain price for and not a nickel more and that price was below our forecasted cost. I couldn’t see it. It wasn’t one of these things you can make up on volume. And in hindsight, the forecasted cost was wrong and the volume was 100x what anyone thought.”
And of course Apple ended up partnering with TSMC which disrupted the foundry business with the half node process development methodology which put TSMC solidly in the process technology lead ahead of Intel and Samsung.
Yeah, I think the Apple A4. Apple was openly frustrated with Samsung's design "slowness" on the SoCs they were using, and Apple was convinced they could add unique value in silicon. Good call. Intel had fab leadership at the time, and it was pretty obvious Apple wanted that for handheld devices. Apple was something like 8% of the desktop/mobile computing market back then, and I don't remember talking to anyone in Intel who thought Apple could beat Blackberry, which was maybe 50M units per year in phones. I always heard a lot of internal bravado about Intel fabs producing the "cheapest silicon in the world", and maybe it was at two or three hundred million units per year back then, but I'm pretty sure the iPhone 3 was in the 20M unit range, which was comparable to server CPU sales at the time.
My memory is that it was difficult to see Apple succeeding the way it did back then (I can't say I did), Intel was really fat and happy, and Intel apparently didn't have any senior visionaries like Jobs to see the broad-based handheld computing market potential.*** Jobs had a relationship with Intel that went way back (it's well known that Andy Grove was Jobs' mentor), but I don't think the Intel execs of the day were all that impressed by Apple or Jobs, or perhaps just Jobs' well-known demeanor. It's so far in past that now what really happened will probably remain conjecture.
***Intel's Ultra-Mobility Group of that time was led by Anand Chandrasekher, who left Intel in 2011, went to Qualcomm, and then famously said that Apple's A7 64bit support was "a gimmick".
Last edited:
I've been thinking about it, and I don't know if TSMC's measured development pace was an Apple mandate, as much as it was TSMC doing what made sense for them at the time. Apple's designs for 20nm had to have started at least 3 years prior to product launch (and given how Apple would be used to Samsung they might have needed even more than 3 years to get acclimated to TSMC's ecosystem). 3 years prior to 20nm products was around when the first 32nm products were coming out. It is uncertain, but probable, that this close enough to release Apple would have had to know there were major issues going on with Samsung 32nm. Regardless of if it was 3 or 4 years of design, Apple would have had to have made this switch very quickly after the 32nm debacle to have released a product when they did.
The more interesting angle would be from TSMC's side. It seems TSMC spends around 4 years for node process development. If the original plan was for TSMC 20nm to be finFET based and Apple made them change it to planar, then Apple would have needed to make this request in 2010 at the latest (1 year before 32nm products evan began ramping). Additionally the historical precedent was for TSMC to adopt the newest features one node after intel. Considering the out of nowhere nature of intel 22nm it seems reasonable that this trend would have continued without Apple asking TSMC to be more cautious given that finFETs didn't seem to be on people's radar at the time as a viable option for 20nm nodes (I wasn't there though, so I could certainly be wrong on this point). Without a doubt, the timeline of events does allow for a planar 20nm to have always been the plan, with finFETs being brought/pulled in after 22nm products started shipping/teardowns became available (4 years between 22nm to 16FF should be plenty of time for a fast follower to develop their own finFET FEOL mated to the then in development 20nm BEOL).
Finally statements from most of TSMC's other customers saying they could use 20nm, but chose not to on account of the poor improvements over 28nm does not indicate to me that 20nm was originally intended to be an Apple exclusive request for a yield learning node on the way to 16FF (even if in the end that is what it kind of ended up being on account of the excellent fast follow work done in developing a finFET FEOL as well as the undesirable nature of 20nm).
Regardless of what happened behind the curtain, the results speak for themselves. Even if 20nm didn't really allow TSMC to have any advantage over Samsung or intel for first gen finFETs, this approach likely paved the way for future successes. Without a doubt in my mind this node development mindset helped TSMC not choke on MP like intel clearly did at 14nm and especially 10nm, as well as clearing the way so N5 did to not have to solve the challenges faced by N7+ or 7LPP.