First a disclaimer. most of what I write here is based on performance leaks and benchmarks. AMD confirmed the benchmarks as well as their pricing this morning, but the devil is always in the details, and AMD has a history of bugs, glitches and overpromises. They also saved X86 once upone a time with the 64bit instruction-set and helped assure X86's victory over many competiting chip.

For a while now, Accidental Tech Podcast, Daring Fireball , and the other usual voices and unusual voices have been speculating that Apple would be transitioning from Intel for ARM for Macs. This is due to a lack of performance improvements and restrictions and schedule problems with Intel. There is enough smoke here to conclude there is fire although it may just be in a low power sleep mode fire.

CPU transitions are risky. While Apple has survived CPU migrations twice, Commodore lost relevance with the transition from the 6502 to the Motorola 68k and was desperately trying to pull off a RISC transition when the company went bankrupt. Billions of dollars were lost and continue to be lost by Intel, HP, and Oracle over Intel’s failed transition to the Itanium chip. SGI died trying to pull off a migration to Intel. AMD tried to transition their server workloads to ARM and completely failed. CPU transitions are more likely to kill you than not. These are not insurmountable odds(after all Apple has done this twice) but it's risky and there may be no need to make this transition, even if Intel has lost their mojo.

The reason here is that it appears that x86 processors look to be suddenly competitive again thanks to AMD’s new “Zen” architecture and Ryzen chips . According to AMD these chips will outperform Intel's chips and at 1/2 to 1/3rd the cost. Ryzen’s top of the line R7-1700x offers performance better then any chip you can get in a Mac today instead going at it with the $1000 Intel 6900k. This is enabled by a massive improvement in instructions per clock (IPC) compared to the previous generation of AMD chips and significantly better multi-core tech. In cinebench for example, the Ryzen is almost 40% faster then the Intel competitor.

Assuming AMD isn't fibbing, what does this mean for Apple? Apple’s x86 transition corresponded with the rise of the Conroe. Apple has never gotten any of the advantages of competition in the CPU market. If the numbers and benchmarks are right, not only has AMD become competitive, Intel will be on their back foot a while. Ryzen might be much better in the exact scenarios that Apple hardware sells well in - multi-threaded computation and audio and visual processing. Ryzen is a clean room design, which not only delivers vastly better price/performance, but also theoretically has a lot of head-room to compete with anything Intel comes up with in the short term. That is unless intel drastically cuts their prices (which is still great for Apple).

So what about Apple lveraging Ryzen as well as Kaby Lake? There are some problems, but I would argue vastly less then going down the ARM path.

The first (and most obvious) is that AMD is initially targeting the enthusiast space for these chips. There are no laptop capable chips yet - that comes this summer with the ryzen APUs.. AMD is promising a similar quantum leap in the graphics space with their VEGA architecture for graphics, but it’s unknown if this will deliver the performance that the systems need.

So while we wait on laptops, what about the Pro, mini and iMac? Thanks to x86 compatability, most issues are fairly simple. One big exception is Thunderbolt 3. None of the newly released chipsets support Thunderbolt 3. AMD or Apple would have to license Thunderbolt 3 from Intel. The good news is that AMD and Intel have a good cross-technology license already place. These Thunderbolt 3 connections require PCI-e lanes. In general while the Ryzen CPU’s have more lanes then the Intel competitors (24 versus 16), the chipsets on the top of line models have less lanes - the X3720 has a total of 32 lanes compared to the 16 lanes on the MacBook Pro Quads and 12 lanes on the MacBook, but short of the 40 that the Kaby Lake + desktop chipset offers. I don’t think this is a issue in the laptops, iMac or a hypothetical iMac Pro, but may be a issue in a future successor to the trashcan Mac Pro. It’s obviously suitable for a Mac Mini if apple ever decides to build one again. In addition, Ryzen also supports ECC ram and high speed DDR4 memory. All pluses.

So what about the MacBook pro. The Macbook Pro's value prop has always been a insane amount of CPU to throw at problems and make them disappear. The Intel chips that are in the pro are laughably out of date now, and the mainline intel chips don't support features that Apple users demand - like ECC ram. AMD may have a answer for this, this summer AMD will release a 32-core beast called Naples.. If AMD applies similar pricing to this vis a vis Xeon, the value proposition may be overwhelming. Can you imagine a mac with 32 CPUs? If you really want to get back to the world where the Mac Pro is unquestionably the most powerful machine you can buy in the PC verse, that pretty much would do it.

One last interesting note is that AMD is apparently also taking a page out of ARM's book, and licensing their CPU designs. Want a Apple designed chip with all of the Apple IP pre-baked in, without forcing a migration? That may be a option.

Assuming the benchmarks are true this sure looks like a better alternative for Apple then the ARM path. Increased competition in the x86 space will hopefully bring back a renaissance in CPU design, when both AMD and Intel benefited from the introduction of 64 bit technology and high-speed low cost multi-core chips, which pretty-much killed all of the x86 competitors - the Sparcs, the Itaniums, PA-RISC, etc. Even if Apple didn’t embrace Zen, and Intel found a way to right the ship, I suspect ARM on the desktop would be sooner or later be killed by a resurgent X86 platform. If Zen is what it appears to be, the last place Apple should go is ARM on the desktop.

1 Comment