I've only watched the first video so far. Sorta watching the second.
Focusing on the CPU, I think @leman and others are entirely correct that this is almost exactly what one would expect from a 12 P-core M2 CPU that uses, I can't quite remember the technical x86 term, variable clock speed to achieve its scaling (dynamic voltage something? DFVS? I'm too tired to look it up) and has no E-cores. I would've liked to seen comparisons with AMD here, as @exoticspice1 pointed out, especially in x86 laptops they, not Intel are the ones to beat in terms or performance and performance/W even if Intel is still the larger chipmaker.
I am torn. I do wish they had been bought by someone other than Qualcomm if nothing else because going after x86 in the server space is where x86 makes much of its profitability. On the other hand releasing quite a good 1st generation product for consumers could also spur growth of ARM into the PC space as well. This can even have a small knock-on positive effects for Apple consumers (thinking of Wine/CrossOver only having to emulate Windows rather than also go through Rosetta and greater optimizations for ARM overall) even if for Apple this could slow growth against the Windows ecosystem which is one reason why Microsoft was so hot-under-the-collar to get a good ARM consumer chip.
An interesting aspect of this is that we get a window into what an Apple chip would like without their E-cores. It turns out scaling clock speed just by itself is pretty damn good if you have a good P-core design! This is probably one reason why AMD has been slower to adopt E-cores in its consumer designs than Intel was.
However, Qualcomm compared to Apple is likely operating at lower margins here. This is nothing new in the PC space relative to Apple but a significant disadvantage of relying on P-cores is that Qualcomm has to spend a lot more silicon to achieve their multicore performance at their target wattage. It’s great per mm^2 if you’re building an 80W monster (and looking at some Windows laptops I’m not convinced that MaxTech is right here and even then he was only skeptical about laptops not going there), but Apple having such good E-cores is a big advantage in the M3 and M3 Pro range.The GPU performance basically confirms our earlier analysis of the Adreno GPU performance: it’s fine, but the hype was unwarranted. I know people were not thrilled about the hype around Nuvia/Oryon but matching the M2 in their first outing is better than just fine, it’s damn good, especially for the PC space. Maybe I’m doing the GPU a disservice, but … at least in raster performance there’s nothing terribly special here at first glance. And we know Qualcomm’s compute performance has historically been terrible, but maybe they’ll focus on those drivers now that they have a PC chip.
The Qualcomm NPU is interesting, obviously a big focus for them. It sounds like Apple will also be looking to make theirs much bigger in the next generation as well and it should be interesting to see what else Apple does with matrix units, ie the GPU, if anything.
I guess he just always assumed it was a TBDR design?
Last edited:
