Accidental Tech Podcast586: Do a Barrel Roll!
https://atp.fm/586
Good morning! We have a huge show ahead of us, and it's all about iPad. This is the biggest show about iPad since its introduction. But first…
mingistech 
Parkscope Joe :RCID:@atpfm @marcoarment skippy would be proud
Diego Petrucci@ParkscopeJoe @atpfm @marcoarment It’s actually Peppy that tells the player to perform a barrel roll. The Toad you're thinking of is named Slippy. (you were close) I'm a big Star Fox nerd, sorry. :)
Alex Padilla@atpfm Perfect, I’ve got 3 hours of driving to do this afternoon/evening
Starxxon@atpfm My understanding is that it's the OLED screen that enabled Apple to make the iPad pro thinner, so the battery should be about the same size?
Potch@atpfm re: data center stuff: given the dire supply constraints (and corresponding prices) for commercial AI hardware right now, it wouldn’t surprise me if apple, trying to launch LLM-powered features this year, found racking their own chips to be surprisingly cost effective to meet demand from ios-scale customer traffic, if not the only way to get that scale *at all*
erkka peitso@atpfm the question about do people need iPad Pro level power: it’s future proofing for us normal people, as we tend to use our tech a lot longer than you guys expect. I’m still using the original 12,9” Pro from 2015 for my work annotation, photoediting and illustration (I’m not a pro-level) graphic designer) purposes and it still does the job. The M4 iPad Pro is the first iPad that makes me want to upgrade and I’d expect it to serve me the next (not 3 or 5 but) 10 years.
Erlend@atpfm PSA, @caseylisscom: You can remap Caps Lock to Escape on iPads, in the settings. (Don't remember exactly where - but I assume under Keyboard.)
You gotta find a new way to rage post, though.
vince propane
Anthony AllisonUnderstandable that @atpfm may not be super familiar with the full-sized truck market, but there are more $100k trucks on the road than you might think. Now, those aren't your typical F-150, but neither is the cyber truck.
Mike Stanley ☁️ 
Marcos Huerta@atpfm @siracusa Re: GPU on Apple Silicon and inference.
Now, Apple may have their own super-power access to the Neural Engine directly, but for anything else the only way to use it is CoreML, and CoreML decides on its own how to distribute across GPU, CPU, and ANE.
For example, in the CoreML enabled build of Whisper.cpp, most of the inference is done on the GPU. I think most inference would benefit from running on both the GPU and ANE.
So if Apple was building some big Apple Silicon based data center to do AI inference (with or without new chips)I think lots of GPU cores would help.
Wade@atpfm The “ML” CPU extensions were AMX (“Apple Matrix Extensions”, I assume), starting with the M1. They’re basically custom matrix instructions, as kind of an extension of NEON. They were never officially documented and I wouldn’t be surprised if Apple rejects any apps (for their App Stores) which use them directly. AFAIK officially the only way to use them is through Apple frameworks.
I suspect the M4 adopts Arm’s official version of this, SME if not also SME2, which are kind of extensions of the SVE2 SIMD capabilities. Which also means the M4 supports SVE2, a first (I think) for Apple’s CPUs. Those two or three combined will provide some *big* boosts to some workloads - assuming Apple implements all the good bits (it’s optionals all the way down these days, with Arm’s ISAs).
That also implies that the M4 is Apple’s first ARMv9 CPU (SME is ARMv9 only). ARMv9 is a relatively insignificant upgrade on ARMv8, compared to what ARMv8 was over ARMv7. Most stuff in ARMv9 is optional and available for ARMv8 too. Still, It’ll be interesting to see what other new ARM ISA extensions the M4 supports (e.g. TME, Transactional Memory Extensions, could be a big perf boon for many serious apps). (3/3)
R2DHueApple doesn’t consider AMX a ”discrete” coprocessor or processor like its GPUs or NPU but considers AMX part of the CPU — which is confusing considering people have identified a distinct “block” with 4 specialized ALUs appearing apart from the CPU cluster on the floorplan.
Nevertheless, AMX is CPU controlled logic that’s like a CPUs FPU or NEON hw.
It’ll be interesting to see if AMX becomes obsolete if Apple fully embraces ARM’s SME/2 in ARMv9.# in future SOCs e.g. M5…
@atpfm Apple’s in a relatively unique position of making high(ish)-end CPUs that ship to their customers and that can also be used in their datacenters. It’s a logistical nightmare to have lots of SKUs out in the wild, for developers and Apple, but in a datacenter you don’t have to care. You can take all sorts of mangled CPUs, that you could never sell publicly, and put them to good use still. CPUs that would otherwise be literally binned (in the “trashed” sense).
@atpfm Designing a typical server isn’t as expensive or time consuming as you probably think. A million dollars or two, and 3 to 6 months. Assuming you have a reference motherboard design to start from (a Mac Pro’s would do). You can do it in parallel with the silicon, and have machines operational in the datacenter literally the day the first CPU samples arrive from Taiwan. And those machines are cheap. Apple making their own servers is easy and not a big leap.
Marc Puricelli@atpfm Excellent episode as always. Thank you John for mentioning the biggest iPadOS issue in my opinion: letting tasks run in the background!
Odin@atpfm Glad to hear it pointed out that the iPad Pro continues to have a single port, I guess the arguement is that there are two if it’s in the keyboard case, but I really think they could offer more
@atpfm I like to imagine Apple just way over manufactured stickers years ago like the first HomePod and are finally starting to get low on stock.
Rob Howard@atpfm I suspect most Apple stickers end up stuck on Dell laptops.
Aaron Dippner@atpfm @siracusa re: Apple AI server chips, I wouldn't mind them bringing the data center home (however unlikely). Let me self-host an appliance that does the work on my dime, let them handle the relaying and integration. Heck, bring back Xserve or the Server app so my existing devices can be nodes (I can dream! 😆)
Steve Wilkinson (cgWerks)@atpfm Why is Elon a horrible person?