This article is a spiritual successor to a piece that has just been published. In the prior piece, I noted that Apple (AAPL) had designed a kick-butt system-on-chip with its A7 processor, and that on a core-for-core basis it is very competitive with the very best designs out there from the likes of ARM (ARMH), Intel (INTC), and Qualcomm (QCOM), and on the graphics side of things absolutely decimates its competitors.
Now, this development leads to some speculation that must be taken very seriously for Intel investors: will Apple kick Intel out of the Mac? I believe that the answer is a firm "no" and plan to justify it here.
Apple's A7 - Yes, It's A Great Low Power CPU, But It's Not In The Same League As "Core"
I recently got the following comment from ARM uber-bull (and Intel uber-bear) Rob Tanner that really gets to the heart of the question that I need to answer (and make no mistake, the concern Rob points out is very real and must be answered):
So, first off, while the A7 *is* competitive with Intel's Bay Trail, you need to understand that every application in which the dual core A7 was competitive with the quad core Bay Trail was NOT designed for multiple cores. What you essentially had was a quad core system-on-chip fighting a dual-core system-on-chip in benchmarks that could only use one core. In these benchmarks, the A7 and the Z3770 (product name for the Bay Trail) traded blows, indicating that on a per-core basis, the "Silvermont" core is about as fast as Apple's "Cyclone" core.
Then, if we look at the Mozilla Kraken benchmark, we see a similar "closeness" between the Bay Trail and the A7 (again, this test is pretty much single-core only):
The "Cyclone" and the Bay Trail are again neck-and-neck, although again, remember that this test is largely testing the performance of a SINGLE core and that different browsers may make the results differ. The only thing we can tell here is that the Apple part and the Intel part are similar in performance, and both are significantly faster than the Qualcomm parts.
Now, if we look at Geekbench, as I did in my previous article, you'll see that the Qualcomm Krait 400, the Silvermont, the ARM Cortex A15, and the Apple Cyclone all perform within a hair of each other (the precise pecking order is Cyclone > Silvermont > Cortex A15 > Krait 400). In 64-bit mode, the deeper resources of the ARM64 ISA (and new crypto instructions) allows the Apple part to really pull ahead, although it remains to be seen how well Bay Trail does in a 64-bit environment (the X86 ISA is very register-starved, so there should be a nice bump in going from X86 to X86-64).
Bay Trail CPU Full Tilt Is Much Faster Than A7
So, core-for-core, Bay Trail is about as fast as the A7 in 32-bit mode, but in a situation with all four cores fully-loaded, the Bay Trail pulls ahead by virtue of the fact that it has twice as many cores. I expect that a potential A7X with four "Cyclone" cores (or simply higher clocked ones) could put up a good fight on the CPU side against Bay Trail in a multi-threaded environment, although we'll have to see what the power consumption looks like.
The GPU On The Other Hand...
As far as graphics go, the Bay Trail is actually not as fast as the A7. While Intel's upcoming "Merrifield" will be using the same class of GPU cores licensed from Imagination as Apple is using (and likely can clock them higher due to the process advantage), the Bay Trail is actually pretty skimpy on the GPU side. I expect that Intel will dedicate more resources to its tablet GPUs going forward, but at the very least the GPU in the Bay Trail is right up there with the vast majority of its contemporaries, beaten only by the Qualcomm Snapdragon 800 and Tegra 4 when it is cooled by a fan.
So, What's This Have To Do With The Macs?
So, this leads me to my argument about Intel not getting kicked out of the Macs. Look, Apple's very best CPU design is in the same ballpark as Intel's Atom, which is optimized for low power and low cost. Intel's "big cores" are substantially faster than any of these low power SoCs, and since the company brings to bear its full design resources (and process technology leadership) to these parts, it is unlikely that Apple will want to spend what is likely an exponentially greater amount of money on a much larger/faster Core/SoC to take aim at a stagnant market.
The economics of a phone/tablet chip custom designed by Apple make sense - these chips go into tens of millions of devices per year, and they're not anywhere near as expensive to develop and build as a much higher end chip. The ROI for doing its own mobile chips is there given the volume/growth, but doing higher performance chips (with commensurate higher end graphics, uncore, etc.) likely doesn't bring a solid ROI.
But the real argument is that the Macs will slowly and gradually decline and will simply be a high end niche while the iPad and iPhone will be the dominant devices for the vast majority of consumers. This is true, but it would have been true no matter what. I don't think Apple kicks Intel out of the Macs, but as a % of Apple's total computing revenues, it's clear that Intel will continue to decline over in Apple land by virtue of the broad shift towards iOS devices.