AMD's ARM implementation speculation

[Raqia]

????
Financially, Intel is more than twice the size of Qualcomm, and always has been.

And, as I said, Qualcomm gets most of their profits from their patent licensing, which will be drastically reduced in the upcoming 5G era.

Cheers

I think Intel's model is starting to show holes in a more dynamic era of fabless companies like ARMH and QCOM who license designs and pure play fabs who are neck and neck with Intel when it comes to process. Most of their expansion like the one into mobile and IoT have have fizzled (and in fact lost them money), and there's rumors that they're licensing AMD tech for GPUs; so their future is entirely dependent on their bread and butter main businesses of x86 mobile, desktop and server CPUs.

They've held on to these markets and quashed earlier competition by jealously guarding licensing and using rebates + lawsuits; but these are looking more and more vulnerable. I personally haven't upgraded my desktop since 2012 as the i7 3770k in there is still excellent; most of the high impact upgrades have been to the GPU and storage subsystems, and I think the laptop/desktop business is destined to slow down further as performance and qualitative improvements resulting from that performance are harder to come by. Big iron users are changing the emphasis towards performance per watt and flexible integration of client features (like custom hardware for energy efficient convolutional neural net processing) into silicon rather than swallowing whatever the vendor puts on sale year after year; while no one has challenged them in servers for a while, this looks more and more vulnerable to challenges in the future.

They need to spend more on cutting edge fab capacity that they can't quite fill out with just desktop and server cpus (it's been slowing since they cancelled a fab in 2011); so they bought Altera and there's rumors of them wooing Apple for business. Perhaps this will be the first AX w/ an integrated (Intel) modem. However, they need to firing on all cylinders in their core businesses to keep the model going whereas QCOM's model based on licensing is far more flexible and robust. Intel will suffer greatly on the slightest miss in projections, and even though QCOM probably won't get its 3.25% a device FRAND rates in 5G, 3G/4G has hardly reached saturation in many parts of the world and QCOM's royalty on that isn't going away until mid 2020's. (I think the FTC case against them will be dismissed as the dissenting FTC member on that issue is now chair-woman and the Apple case will be weakened and drag out. Note that apple puts the value to the consumer of LTE on each ipad at $130 and puts $200 on its value in the iPhone SE vs iPod so Qualcomm's skim is a relative bargain.)

 

[Deleted member 13524]

Thus, I believe any AMD tablets we'll see will use Zen+ 2c/4t APU (or is there a 4W Zen before that one?) : reuse a chip that's for netbooks, tiny desktops, 12" MacBook-like things on tablets like the Atom tablets that use Windows 10, Android or dual boot.

Mobile Raven Ridge starts at 4W:

They don't speak of any cut-down units in that roadmap (yet). I doubt there will be more than one Raven Ridge physical form and they will only be making 4*Zen + 12*CU chips, but it's possible the 4W variant only activates e.g. 2 CPU cores and 6 CUs. What I mean to say is there doesn't seem to be any chip akin to Intel's 2C+GT2 chips, but the power consumption range of the Y and U series is contemplated within Raven Ridge.

Regardless, I hope AMD manages to take a healthy marketshare from intel's Y and U series, something Stoney and Bristol Ridge failed miserably at.


Regarding the topic at hand, I think it's more probable to see Zen SoCs going down to smartphone power envelopes somewhere in 2019 on 7nm (imagine Win10 phones without the need to emulate win32 apps) than to see AMD trying to compete in the ARM business.
ARM cores, both original and custom, have been evolving at a lightning speed. Cortex A73 has been inside phones on the shelves since last year, and K12's original plans for a Cortex A57 follow-up may have already been surpassed.

 

[Zaphod]

Financially, Intel is more than twice the size of Qualcomm, and always has been.

Actually, back in late 2013 and early 2014 Qualcomm had a larger market cap than Intel.

As for the topic at hand: Are there anyone among the known architecture licensees putting out custom high performance ARM cores unless they also have a mass market product to put them in? I.e. are there any custom arm designs for high performance, high margin products that aren't an iPhone?

Automotive, like what Nvidia seems to be focusing on now? The K12 as announced didn't sound like it was destined for such markets.

 

[Blazkowicz]

Regarding the topic at hand, I think it's more probable to see Zen SoCs going down to smartphone power envelopes somewhere in 2019 on 7nm (imagine Win10 phones without the need to emulate win32 apps) than to see AMD trying to compete in the ARM business.
ARM cores, both original and custom, have been evolving at a lightning speed. Cortex A73 has been inside phones on the shelves since last year, and K12's original plans for a Cortex A57 follow-up may have already been surpassed.

This is what I alluded to, timing is very important in the ARM market.
I'm not even sure if AMD will even make an ARM SoC for servers (A1100 was discontinued), then they'd sit on an ARM K12 core that goes unused.