tenthousandthings

zenwaves said:

So, as a 'civilian' consumer - who has owned almost every Apple product since the Mac Plus - I feel fine criticizing them when they underperform (AI, ML, or whatever broken tech runs Siri), or make bad decisions (see developer relations).

Apple is so far behind in this space, it's no wonder they had to punt with a Google deal.

Just today, I did a voice search within the tvOS Apple Music app. The most unambiguous name. There is only one.... and yet instead of a "it just works" 'intuitive' result, I get wildly incorrect search results. Sure, there's a box showing exactly what I said in quotes, but it requires me to use the finicky touchpad on the remote to complete a few more clicks. THEY CAN"T SEARCH  THEIR OWN CONTENT LIBRARY!

For what it’s worth, I couldn’t reproduce that in tvOS 17.4. I get the expected search results.

tht said:

tenthousandthings said:

I may have missed this in the article, but as I understand it, the USB/Thunderbolt controller is built into the SoC silicon. For M3, that is a USB4/Thunderbolt 4 controller. So Thunderbolt 5 seems very unlikely for M3.

More likely M4 Pro/Max in June 2025 with Thunderbolt 5, and new displays when the Studio and Pro launch. 

The Mac Studio and Mac mini is the last Macs to still get their M3 updates. Not many M3 models left before the M4 cycle starts.

If they want to, they can put TB5 into an M3 Ultra Mac Studio by connecting a TB5 controller onto 4 lanes of PCIe 4. Intel's TB controllers typically have 2 ports on one TB bus, and if the same for TB5, a Mac Studio with M3 Ultra could have 2 TB5 ports. Enough to drive two 6K monitors at 60 Hz or one 6K XDR at 120 Hz. That's like $10k in monitors alone, assuming Apple updates the XDR. Or two 8K monitors at 60 Hz might be doable in the 120 gbit/s mode, need to do the math.

Narrow market. 240 to 360 Hz 4K gamers. Apple's market of 5K to 6K buyers who want 120 Hz.

I wasn’t aware of this, but the M2 Mini is an example of Apple being able to adjust the USB4/Thunderbolt controller in the SoC. The M2 Air SoC, for example, only supports Thunderbolt 3, but the M2 Mini SoC supports Thunderbolt 4. 

I may have missed this in the article, but as I understand it, the USB/Thunderbolt controller is built into the SoC silicon. For M3, that is a USB4/Thunderbolt 4 controller. So Thunderbolt 5 seems very unlikely for M3.

More likely M4 Pro/Max in June 2025 with Thunderbolt 5, and new displays when the Studio and Pro launch. 

I read the article, looking for something to support the headline’s claim that Young “tries excusing his return to Spotify by saying Apple Music is now as bad” — but nothing in the article supports that assertion. Young explained why he was returning, his “excuse” being that Rogan’s disinformation is now available everywhere, so his rationale of not wanting his music on the same platform is no longer valid.

I get that Apple isn’t the same as Spotify in terms of supporting Rogan’s disinformation, but Young didn’t say anything about that, at least according to this article.

I gather Young thinks Apple is “now as bad” as Spotify in terms of technical sound quality, his personal white whale, but that has nothing to do with why he left Spotify, or why he returned. The headline seems to conflate the two issues (disinformation and sound).

blastdoor said:

tenthousandthings said:

blastdoor said:

tenthousandthings said:

If the past is any indication, then, yes, M4 will be based on A18, and it (M4) will launch in June 2025. That's the A15/M2 release pattern. TSMC N3P will be the process node for this, with volume production in "2H 2024" and so it is obviously in risk production right now, with Apple leading the way.

It is too soon to say this, but I will anyway: the silicon production cadence is set now. The only real unknown (in my view) is whether or not every Mac will get every generation of Apple M silicon, every 18 months (not all at once, but spread out within that period). I think probably yes (the M2 iMac was a casualty of the global pandemic), but we'll see.

Here's my predictions (within the current Pro/Max/Ultra structure, that is, leaving out the possibility of some kind of "Extreme" Mac Pro configuration):

A14 Bionic (October 2020) [TSMC 5nm gen1 "N5"]

M1 (November 2020) [TSMC N5]

M1 Pro/Max (October 2021) 

M1 Ultra (March 2022) 

A15 Bionic (September 2021) [TSMC 5nm gen2 "N5P"]

M2 (June 2022) [TSMC N5P]

M2 Pro/Max (January 2023) 

M2 Ultra (June 2023) 

A16 Bionic (September 2022) [TSMC 5nm gen4 "N4P"]

A17 Pro (September 2023) [TSMC 3nm gen1 "N3"]

M3, M3 Pro/Max (October 2023) [TSMC N3]

M3 Ultra (June 2024) 

A18, A18 Pro (September 2024) [TSMC 3nm gen3 "N3P"]

M4, M4 Pro/Max (June 2025) [TSMC N3P]

M4 Ultra (January 2026) 

A19, A19 Pro (September 2025) [TSMC 2nm gen1 "N2"]

A20, A20 Pro (September 2026) [TSMC 2nm gen2 "N2P"]

M5, M5 Pro/Max (October 2026) [TSMC N2P]

M5 Ultra (June 2027) 

That's a helpful breakdown, but isn't it called N3E, not N3P?

No, in my view, N3P is next. Apple will skip N3E.

N3E is in volume production now, but Apple isn’t using it, as far as we know. N3P is a small adjustment to N3E, and Apple is in a unique position to leapfrog onto it, instead of using N3E like everyone else.

I read N3P enters volume production in the second half of 2024. If true, that seems too late to support an iPhone launch. But I suppose there's always some uncertainty about these timelines, so maybe Apple will manage to get early access. 

I wonder, though, if Apple will use N3E this year, N3P next year, and N2 might only come in 2026. As you note, Apple went three years on different spins of TSMC's 5nm process. The same thing might happen with 3nm. 

According to https://www.anandtech.com/show/21241/tsmc-2nm-update-two-fabs-in-construction-one-awaiting-government-approval, volume production on N2 isn't planned until 2H of 2025 which seems too late to support a 2025 iPhone launch. 

Well, if that were true ("2H" for volume production means no iPhone launch on that node in September of that year) then A16 Bionic would not be on N4P, and A17 Pro would not be on N3 (also known as N3B, though TSMC appears to have shifted back to calling it N3 now that N3E is established)... TSMC used "2H" for both of those, it looks like standard practice for them now with regard to estimates for the nodes where iPhone will be the first consumer product to launch, and they're using the same protocol for N2 ("2H 2025") and N2P ("2H 2026"). Although that timeline depends on the introduction of GAA transistor architecture (N2) and backside power delivery (N2P) both continuing to go smoothly, so I wouldn't bet the farm on either of them. They are both major steps forward, there could be delays.