So, The Transition Kit that was presented Monday. No custom desktop CPU, no Thunderbolt ports, but 16GB of ram and 512GB SSD. Hey that’s still 10 GB of ram more than what comes standard on the iPad Pro with the same CPU!

But let’s talk realistically about this device. First of all, its not meant to represent anything that would be going into immediate production. It’s essentially a bodge where Apple redesigned an iPad Pro logic board to fit inside a Mac Mini case. Port wise, It will have 2x USB-C (3.1) ports, 2x USB-A (3.0) ports, and an HDMI 2.0 port. Networking, it has 802.11ac Wifi, Bluetooth 5.0 and Gigabit Ethernet. Essentially, it really is an iPad Pro stuck into a Mac Mini case. The HDMI port, Gigabit ethernet, and possibly the other USB ports are basically just your bog standard USB-C hub that provide the same exact ports that you can buy on Amazon for about $40. Only difference is that it is built into the logic board.

But this actually follows Apple tradition regarding these machines. Intel’s Developer Transition Kit for Intel was a laughably undersized logic board stuck inside a PowerMac G5 chassis.

That thing had more empty space than actual components. The only ports that were included were audio in and out, 4x USB 2.0 ports, 1 FireWire 400 port, and an Ethernet connector. The actual PowerMac G5 that would normally be in that case would have FireWire 800, two S/PDIF (Toslink) ports, and I think it was even missing Bluetooth and WiFi. Also, a 3.6Ghz Pentium 4, standard 1GB of DDR2 Ram in four slots (two occupied) had hardly any resemblance to the first production Mac Pro that came out. What we’re seeing with the Apple Developer Transition Kit (ARM), is pretty much the same. What we get now, and what will be actually released will only bear a superficial resemblance to each other.

The main purpose of this new machine is simple. Get developers to port their software as fast as possible. That’s why the kit costs $500 (half the price of the Intel transition kit), but as usual in this case, you’re going to have to give it back to Apple. It’s basically a stepping stone in order to get a large enough software base on the market by the time the actual consumer products. But one thing we can compare is performance of the last computer using the G5/G4 processors, and the first using Intel’s Core Duo.

	Speedmark 4.5	Cinema 4D XL 9.5.21	iMovie 6.0.1 aged filter	iTunes 6.0.3	Unreal Tournament 2004	Zip Archive
iMac G5 2.1GHz	194	2:09	1:10	1:22	40.7	2:54
	+10%	+45%	+11%	-3%	+21%	+12%
iMac Core Duo 2.0GHz	214	1:11	1:02	1:25	51.6	2:33
	> Better	< Better	< Better	< Better	> Better	< Better

	Speedmark 4.5	Cinema 4D XL 9.5.21	iMovie 6.0.1 aged filter	iTunes 6.0.3	Unreal Tournament 2004	Zip Archive
PowerBook G4 1.67GHz	134	3:57	1:50	2:00	22.9	3:29
	+20%	+70%	+24%	+23%	+125%	+10%
MacBook Pro 2.0GHz	161	1:11	1:08	1:33	51.6	3:08
	> Better	< Better	< Better	< Better	> Better	< Better

Benchmarks via MacWorld: https://www.macworld.com/article/1049771/speedmark45.html

So, in general, the performance boost in the first Intel desktops was about 16% on average, depending on the application. Graphics boosts can be counted more as better GPUs in the new Intel models, rather than just CPU performance. As for the much higher speed boosts for the laptops (45% on average among the benchmark scores)? Well, that can be considered the reason why Apple decided to go to Intel. IBM was not able to make a PowerPC 970 (G5) with TDP low enough to be placed in a laptop. At the time, event the Motorola/Freescale PowerPC 7447a had issues with heat generation, where users could burn their thighs if using the laptop on their lap for an extended time. The PowerPC based processors reached a dead end regarding heat generation and performance. While currently, Intel has diminishing returns on performance improvements in subsequent generations of their x86 processors.

What can we expect with the ARM transition? Well, at this point, not much. The current Apple ARM CPUs are designed for ultra low power applications, where heat production and battery life matter more than raw power. But due to the difference in architecture between mobile phones and desktop/laptop computers, most likely Apple with push a fork of the A series into those areas. The first thing they’d need to add, is PCI-Express support in order to get ThunderBolt 3 ports. This is not optional, but a given for production machines due to Apple’s reliance on Thunderbolt displays. That might mean that Apple may support third party GPUs in their future ARM based desktop machines via a PCI-Express bus, and maybe even laptops. AMD’s AMDGPU-PRO Driver doesn’t seem to have an available Linux ARM driver, so there might be less of a chance that Radeon GPUs will be available in any of the first series of ARM Macs. Though, NVidia does provide Linux drivers for the GeForce Series for ARM platforms (aarch64). This could give a chance for NVidia GPUs to return to the Apple’s computers over the next couple of years.

However, even with the possibility for 3rd party GPUs, Apple most likely will be pushing their own integrated GPUs used in the A series silicon. Along with built in co-processors dedicated to AI and such, the compute functionality of 3rd party GPUs might not provide a sufficient boost in performance for the cost of including them in a production design. Most likely, we can probably see Apple’s own integrated GPU in most, if not all of their laptops (MacBook Air & MacBook Pro), and the Mac Mini. A good possibility also for the iMac. The Pro version of the iMac and the Mac Pro might include third party GPUs, but most likely this will come at a premium regarding price. But for that, we probably are at least two years away from that.

As for the next part in the series? Virtualization and Windows maybe?

Previous Posts in this series:
https://irrelevanthacks.com/2020/06/25/apple-and-arm-part-1/