Former Intel Engineer Claims Buggy Skylake Chips Hastened Apple's Switch to Custom Silicon

[chucker23n1]

I know. But you mentioned 14nm, which has to do with fabrication, and not really related to underlying architecture or design/layout. If process node is what's important to Apple, then the key to this transition is TSMC, not ARM or Apple's in-house designs.

While Intel's struggles with 10nm are likely one of the deciding factors to move away from Intel, I doubt it's the deciding one. If it were, Apple could just go to AMD, stay on x86, and use TSMC as the manufacturer regardless.

On top of a newer process node, Apple's chip also have superior single-threaded performance. They also have more control over schedule and feature set. And so forth.

 

[MacBH928]

Only a handful of companies have access to the x86 license. Even fewer have access to the x86-64 license. So short answer: you’d get sued.

So you are telling me that Intel would rather Apple completely abandon them than license an x86 architecture to them?

AMD must be the most powerful, otherwise why would anyone buy AMD instead of Intel?

See how that works?

People buy AMD because its cheaper, I thought that was a given. You want the faster and better, pay for Intel.

...And Apple will be subject to their own roadmap only:
- How many times have people on this forum complained of the lack of new computer models in the last few years, when the underlying problem was that there was no suitable CPUs to put in a new computer model?

Yes but you assume that Apple is going to forever have faster and better CPUs. By going in-house that does not mean their CPUs will be the greatest, they might fall behind just like G5 fell behind.

 

[cmaier]

So you are telling me that Intel would rather Apple completely abandon them than license an x86 architecture to them?

Yes. There are lots of customers who want intel licenses. Intel hasn’t given them out.

And they’d need an AMD license, too. Not just Intel.

 

[anson42]

So you are telling me that Intel would rather Apple completely abandon them than license an x86 architecture to them?

Wait, why would Apple license x86 even if Intel offered it? Apple has an ARM architecture they feel is superior and is consistent with all their devices. At any rate, that's all academic anyways as Apple is going ARM, goodbye to x86 and while there may be some short term pain for some, good riddance.

 

[Gerdi]

So you are telling me that Intel would rather Apple completely abandon them than license an x86 architecture to them?

Apple has not the slightest interest to license such an outdated architecture. Apparently Apple would not touch it with a 10 foot pole, even if it was wrapped in banknotes.

 

[cmaier]

Apple has not the slightest interest to license such an outdated architecture. Apparently Apple would not touch it with a 10 foot pole, even if it was wrapped in banknotes.

I assure you that if AMD and Intel had licensed the x86/AMD64 architectures to Apple for free, they would have given serious consideration to going down that road. It would save them a lot of hassle, still let them come out with new chips each year that spank Intel/AMD (perhaps by a 10-20% less of a margin that they’ll get from ARM), and maybe even do something wild like throw a bunch of ARM cores on an x86 die (or vice versa) to ease the transition if they still decided to run ios apps while keeping mac apps on x86.

But they sure as hell aren’t going to agree to pay for such a license and to agree to also license their own patents to Intel, which surely Intel would demand.

 

[Gerdi]

I assure you that if AMD and Intel had licensed the x86/AMD64 architectures to Apple for free, they would have given serious consideration to going down that road. It would save them a lot of hassle, still let them come out with new chips each year that spank Intel/AMD (perhaps by a 10-20% less of a margin that they’ll get from ARM), and maybe even do something wild like throw a bunch of ARM cores on an x86 die (or vice versa) to ease the transition if they still decided to run ios apps while keeping mac apps on x86.

But they sure as hell aren’t going to agree to pay for such a license and to agree to also license their own patents to Intel, which surely Intel would demand.

I do think the industry needs to move forward from x86 in the bigger scheme of things. I also think Apple would not be able to separate themselfs performance and power wise from the usual x86 crowd as they can now with Aarch64. So there is very little incentive for Apple to go the x86 route - the hassle of maintaining (and developing) x86 designs just for Mac is reason alone to decline such an offer.

 

[Mikael H]

People buy AMD because its cheaper, I thought that was a given. You want the faster and better, pay for Intel.

People buy AMD CPUs because they're cheaper for the same performance, or because they provide multicore performance that Intel can't or won't match at any price point. Intel has been better in a single metric, namely single-core performance, but even that has been shaky the last few months. This will likely change over time as it has a tendency to do, but this is the status right now: You don't necessarily get faster or better with Intel over AMD.

Yes but you assume that Apple is going to forever have faster and better CPUs. By going in-house that does not mean their CPUs will be the greatest, they might fall behind just like G5 fell behind.

I don't assume anything. Technology marches on, and what is a great choice now may well turn out to be an evolutionary dead end in a few years. But by knowing the CPU roadmap and having early internal warning of any upcoming hiccups, Apple can avoid looking bad because of things outside their own control. Remember that they've taken the crap from ignorant and angry users and superficial "analysts" for years without blaming their suppliers. I bet they'll be happy to get rid of that part of potential future PR disasters.

 

[avtella]

So you are telling me that Intel would rather Apple completely abandon them than license an x86 architecture to them?



People buy AMD because its cheaper, I thought that was a given. You want the faster and better, pay for Intel.



Yes but you assume that Apple is going to forever have faster and better CPUs. By going in-house that does not mean their CPUs will be the greatest, they might fall behind just like G5 fell behind.

Just because you pay more does not necessarily mean you get better... If a company has enough clout/reputation they have greater leeway to make mistakes and can keep charging higher prices over a superior product from a competitor or less renown company which has been struggling with its own prior mistakes. This is not just Intel vs AMD but across any business market in general. Theres usually a lag period between perception and reality... Key is to recover between that.

I remember the last time AMD had a lead over Intel over 10 years ago, the Prescott Pentium 4s and Pentium Ds were still selling like hotcakes even though they were terrible in most measures and jokingly referred to as space heaters. One of the reasons for the P4s existence was to win the Ghz race (As consumers think more Mhz is everything), they traded performance per clock for higher Ghz and early P4s were usually beaten by lower clocked P3s in most benchmarks.

Fortunately Intel did switch from the P4s netburst architecture to a much better Core/Core 2 architecture more closely related to the Pentium III / Pentium M and retake the lead while AMD made some bad mistakes of its own.

.....As of the latest gen AMD and Intel have been trading blows even in single core with AMD winning out on more use case scenarios. On top of that at full load the Intel CPUs also draw a lot more power and generate more heat. Until recently Intel had the advantage of superior idle power draw vs Ryzen but if I recall even that advantage is now gone.

 

[adib]

"When your customer starts finding almost as much bugs as you found yourself, you're not leading into the right place."

Back when I was a project manager (mostly for web pages and apps, but still), there were some projects where the customer reported lots and lots of small bugs and inconsistencies within weeks of delivery. Mostly stuff that the team - myself very much included - should have picked up easily and removed before deploying. That was not my proudest moment(s).

One (of many) reasons I switched to teaching a few years back.

"Those who cannot do, teach" ?

 

[Mitthrawnuruodo]

"Those who cannot do, teach" ?

Hehe. I know that expression, and if I were teaching Project Management then it would most certainly have a ring of truth to it.

I'm a bit more confident of my development skills...

 

[chucker23n1]

IBM are still a thing. HP are still a thing in Enterprise hardware. I dread to think how many DL380s we have in our datacenters at work!

Well, IBM is mostly a consulting company now. They're barely a thing in enterprise/cloud hardware. They stopped being a thing in consumer hardware long ago.

As for HP, they're still fairly active (both HP and HPE), but I can't recall the last time something HP did was considered innovative.