As such: a second Neural Engine has been found in the M1 Max chips. Apple mentions only one such neural engine on the chip (and therefore two in the M1 Ultra), but on social media the discovery of a second neural engine in the architecture is in progress. Apple didn’t mention it at all, but you can tell even from the marketing images.
If it’s there, it’s for something.
2022🔎 first evidence of 2 Apple NeuralEngine found on M1Max die photography @VadimYuryev @HotHardware @hishnash @timothy_lkh_ @reneritchie @tatech15 @Locuza_ @onrefaitlemac @iweeknews @iCave_Dave @leolaporte @ZONEofTECH @Jon4Lakers @jon_prosser @TeamNitish2 @SnazzyQ @ValentinMoeller pic.twitter.com/zY5nfhLHwK
— Fred the Frenchy (@Frederic_Orange) January 2, 2022
The first piece of evidence comes from SoC specialist Frederic Orange, who identifies the location of the two neural motors. This second Neural Engine would be disabled, and could be enabled via firmware update on Macs that feature the chip.
The source itself admits: it raises many questions. Why isn’t Apple taking advantage of this second neural engine? Better two than one, and better four than two in the case of the M1 Ultra, which is nothing more than two M1 Max chips put together. Apple could use a balancer and interpret these two engines as one with twice as many cores, among other things.
It is possible that this heart is reserved for very specific functions, perhaps in this future Mac Pro which will probably equip an improved version of the M1 Ultra. We can only see its usefulness when Apple decides to activate it and reveal its presence. The mere fact that he is already there implies that sooner or later he will somehow. Nobody puts a second neural engine on a chip where every nanometer counts just for trim.
