Apple has been upping the ante on their silicon design capabilities faster than anyone anticipated, with impressive results. The end point was known, the steps to get there were understood, but the exact cadence of progress was only conjecture until a few recent steps made things much clearer.
As we stated almost two years ago, Apple is pretty hell-bent on making their own mainstream CPU, top to bottom. Fair enough, but given the magnitude of that task, how on earth were they going to get there? Pundits are fond of pointing out hires, and pointing out how a single person is going to change everything, a laughable notion. What no one seems to do is put all the pieces together to see the big picture.
We won’t rehash all the people, teams, and technologies that Apple has bought over the past few years, but suffice it to say that there is a long list. There were been few outward signs of progress, but that changed with some recent CPU releases. Even with things changing, the progress was, err, progressive, and so once again, few noticed the magnitude when taken as a whole.
If you recall, the first iPhones had a Samsung SoCs in them. They weren’t exactly bleeding edge, and Apple did little more than pick a few pieces and order up the count of various portions, picked clocks, and added the feature here and there. It was not exactly a tour de force of their silicon might, but few companies can do even that much. Then came the next generation, and lo and behold, there were a few Samsung cores, Imagination GPUs, and the rest came from Cupertino. Apple did large swathes of the uncore themselves, at the time an unexpected move. Progress, massive progress, but everyone cares about the cores and GPU, so no one seemed to notice that Apple did the trickiest parts of the SoC in house.
Then all was quiet. The next generation or two, depending on how you count, was just more cores, more GPU, and nothing really recognizable from the outside. World class chips, world class performance, and pretty amazing power numbers somehow didn’t get people’s attention. We blame the state of the press for not seeing what was in front of them, but the changes were there, and the tech was impressive. Because it didn’t involve cores or shader count, no one gave them a second look.
Then came the A6 line, the most recent chips. No A9 core this time, it was a somewhat unexpected A15 ISA core. More unexpected was that it wasn’t a vanilla ARM licensed A15 core either, it was a full blown custom core. No one other than Anand seemed to notice through the reality distortion field that Tim Cook seems to have finally got tuned back up, but Apple made their own core, top to bottom. Headline count? Woefully low. Pointless stories about the finish of the case? 7-8 orders of magnitude higher. If that isn’t progress, I don’t know what is.
Ok, I do know what is, and it was done quietly and in the shadows with few outside of the new hires getting wind. In fact, the simple news that these people are now at Apple are not known outside of a trusted few, SemiAccurate can’t find a single reference to them anywhere, public or private. And now they fill a massive gap in Apple’s silicon capabilities.
What did Apple just grab? Would you believe a full blown GPU design team, one of the best in the world? If so, let me tell you the tale of one of the ex-ATI, ex-AMD GPU teams that are now calling Cupertino home. About 1-2 months ago, a team at AMD responsible for some of the most difficult portions of the GPU, up and quit. Radio silence followed, and we did look for where they went. Teams leaving, if not simultaneously, over a very short time span usually say something, and that part we heard.
Nothing turned up any clues, resumes, calls, web searches, all zip. Where they were before was abundantly clear, where they are now was not. It was time to get serious about the searching, and that could only mean one thing, hitting the coffee shops for the elusive brown-splotched white robed mole. He confirmed our suspicions that they had been snapped up en masse and were were now toiling under a fruity banner, singing the company song, and all that happy stuff that in no way resembles armageddon cult behavior. Armageddon cults wear colorful robes, Apple uses white ones on campus. With this bit of info in hand, it was easy to get confirmation from the glassy-eyed chanting moles of central Cupertino, the ex-AMDers were indeed at Apple.
In the weeks that have passed since, a few more clues have popped up if you know where to look, and the story is now complete, Apple is doing an internal GPU. They got the right people, have almost all the other parts in place, and now this should complete the puzzle. The company can do uncores with the best in class, cores with the best in class, and now GPUs will likely do the same in a few generations. Other than radios, Apple can now do an entire SoC top to bottom. How long before they buy a fab?S|A
Latest posts by Charlie Demerjian (see all)
- Nvidia’s Tegra roadmap slip is now official - May 21, 2013
- Sean Pelletier leaves Nvidia for AMD - May 20, 2013
- Nvidia GPU roadmaps aren’t an improvement with Volta - May 20, 2013
- Redhat put Calxeda based ARM servers in to production - May 15, 2013
- Why would Nvidia release an unlocked GTX680 to GTX770 BIOS? - May 14, 2013
Follow these categories: analysis