Slashdot Mirror
Why Apple and Microsoft Are Using Last Year's Skylake Processors In Their New Computers (gizmodo.com)
Apple released new MacBook Pros yesterday that feature Intel's year-old Skylake microarchitcure, as opposed to the newer Kaby Lake architecture. Two days earlier, Microsoft did the same thing when it released the Surface Studio. Given the improvements Kaby Lake processors have over Skylake processors, one would think they would be included in the latest and greatest products from Microsoft and Apple. Gizmodo explains why that's not the case: In the case of the new 15-inch MacBook the answer is simple. "The Kaby Lake chip doesn't exist yet," an Apple rep told Gizmodo. Kaby Lake is being rolled out relatively slowly, and it's only available in a few forms and wattages. The 15-inch MacBook Pro uses a quad-core processor that has no Kaby Lake equivalent currently. That particular laptop really does have the fastest processor available. The same goes for the Microsoft Surface Studio and updated Surface Book -- both also use a quad-core Skylake processor with no Kaby Lake counterpart. But the Studio and Surface Book are also using much older video cards from the Nvidia 900 series. Nvidia has much faster and less power-hungry chips (the 1000 series) available based on the Pascal architecture. Microsoft's reasoning for going with older video cards is nearly identical to Apple's for going with a slower processor in its 13-inch MacBook Pro: the Nvidia 1000 series came out too late. The major intimation was that Kaby Lake and Pascal came so late in the design process that it would have delayed the final products if they'd chosen to use them. New technology, no matter how amazing an upgrade it might be, still requires considerable testing before it can be shipped to consumers. One minor bug, particularly in a system as engineered as the Surface Studio or MacBook Pro, can turn catastrophic if engineers aren't careful. In the case of Microsoft, it's frustrating, because that old GPU is significantly slower than the Pascal GPUs available. It's a little less frustrating in Apple's case, largely because of the old processor microarchitecture that Apple elected to shove into its new 13-inch MacBook Pro. Apple went with a new Skylake dual core processor that draws a lot of power -- more so than any Kaby Lake processor available. It then uses all that extra power to ramp up the speeds of the processor. Which means it is capable of pulling off speeds that can actually match those of the fastest Kaby Lake processor out there. The only downside to this decision is battery life.
From what I can tell (I don't think Apple has given us the chip numbers), it goes like this:
(remember that "i7" and "i5" don't have meanings- they are just marketing garble, and don't, for instance specify the difference between hyperthreading and non-hyperthreading, or two and four cores: all of these chips have hyperthreading)
https://en.wikipedia.org/wiki/...
(I could have messed up something in transcription)
13" Cheap Model, with TDP 15W:
base: 2 core i5-6360U @ 2.0GHz single core boost to 3.1GHz with Iris 540, (listed as unreleased on wikipedia)
high end: 2 core i7-6660G @ 2.4GHz single core boost to 3.4GHz with Iris 540 (listed as unreleased on wikipedia)
13" Spensy Model, with TDP 28W:
base: 2 core i5-6267U @ 2.9GHz single core boost to 3.3GHz with Iris 550 (listed an unreleased on wikipedia)
midline: 2 core i5-6287U @ 3.1GHz single core boost to 3.5GHz with Iris 550 (listed as unreleased on wikipedia)
high end: 2 core i7-6567U @ 3.3GHz single core boost to 3.6GHz with Iris 550 (listed as unreleased on wikipedia)
In this case, all of the high end models have Iris Pro 550, and all of the low end models have Iris Pro 540. Intel's actual highest listed Iris Pro models are Iris Pro 580, but all of those are on chips that are either pretty expensive, have a higher TDP, or both.
Meanwhile, the 15" laptops all have Radeon graphics cards in them. These have chips that offer more processing power, but less graphics power (with the obvious assumption that the Radeon graphics will be used for that purpose).
15" models all have TDP 35W chips.
15" 256 GB model base: 4 core i7-6700HQ @ 2.6GHz single core boost to 3.5GHz with HD 530 (listed as Sep 1 on wikipedia)
15" 512 GB model base: 4 core i7-6820HQ @ 2.7GHz single core boost up to 3.6GHz with HD 530 (listed as Sep 1 on wikipedia)
Both model high end: 4 core i7-6920HQ @ 2.9GHz single core boost up to 3.8GHz with HD 530 (listed as Sep 1 on wikipedia)
These models are all generally more capable than similar models released earlier. It is likely that Intel and Apple actually reached an agreement via branding and capability on these: it is likely not a coincidence that Intel happened to have highly compatible i5/i7 branding for each step of Apple's needs, for instance.
Regardless, I've seen folks pointing out that Apple really IS using the best Intel chips available on social media, including doing it myself some, as people were all 'muh kaybee layke?' over the last day. These chips are a mix of hyperthreaded 2 core chips with Iris 540 or Iris 550 (on the 13 inch) and hyperthreaded 4 core chips with the lesser HD 530 (on the 15 inch). Meanwhile, the only Kabylake that looks like it could be show up to this party at all is the 7500U, a 15W chip with 2 cores, going from 2.7GHz base to 3.5GHz single threaded boost with HD 620. This chip could maybe have sat in over the cheap model (it costs more though), would require just that one model to be designed and tested around Kabylake stuff, and wouldn't have the Iris graphics (and doesn't have a graphics card). Intel certainly doesn't have the Kabylakes needed to fit their intended build case.
Sorry, but only person who has absolutely no clue about how a hardware product is being developed (and how long does it actually take!) can say nonsense like this.
A new product like the Surface computer or Mac Book is in development for more than a year, often even 2-3 years. And in the latter stages you need actually a stable and working system so that things like drivers can be developed, OS adapted, demo units produced, CE/FCC testing done, etc.
So if a new CPU/chipset combo shows up in the last 9-12 months of the cycle, it is simply too late - it would delay the release of the product by at least that much. This is *not* about just swapping a motherboard/CPU/GPU - the board for the chips needs to be actually *developed* first, before you can even start thinking about integrating it into a product.
The risk mitigation is also important, but that comes into play only after everything above is sorted out already. If there is nothing new to put in your product, you have no "unknown" to fear in the first place.