[whatisthis9]

Apple could easily corner the US smartphone market and a big part of the LLM hardware market in a few years.

US smartphone market? It's growing every year even with huge margins, lowering the price would accelerate this pace.

LLM hardware market? Like really, what is the incremental cost for 8gb to 16gb to 32gb to 64gb to 128gb for their macbooks? The could probably shave a few grand off the 128gb and still make a decent profit. Now you have 128gb unified memory macbook going for almost the same price as a rtx 4090 with a computer (and with only 24gb vram). That means a huge shift in hobbyist developers who will turn all their focus to ML with macbooks (e.g. MXL vs Nvidia/cuda). and if they do this with new macbooks, suddenly the market for older slower macbooks collapses, leading to much higher demand and even more hobbyists albeit with slower Cpus/less ram.

It sucks but apple laptop cpus and mobile cpus dominate anything else out there. I wish there were more competition. And their margins are massive.

[Alpha3031]

> It sucks but apple laptop cpus and mobile cpus dominate anything else out there. I wish there were more competition. And their margins are massive.

Ryzen 7940HS and 7945HX are fairly close to the M3 16 core, if they had actually been able to build them on the same node, and if AMD were willing to shell out for the same die size, it's likely that they could have achieved or exceeded the uplift.

[whatisthis9]

> if they had actually been able to build them on the same node, and if AMD were willing to shell out for the same die size, it's likely that they could have achieved or exceeded the uplift.

I mean yeah, probably, but they haven't been able to despite apple coming out of nowhere to dominate the laptop cpu market in performance only a few years ago with their first release m1. If AMD were to shell out more, and be able to build them on the same node, that would probably entail significant R&D costs and costs for die size and possibly raise the price to the m3 max or m3 ultra (whenever that's released).

>Ryzen 7940HS and 7945HX are fairly close to the M3 16 core,

https://nanoreview.net/en/cpu-compare/apple-m3-vs-amd-ryzen-... fair, i didn't know about these cpus, and it seems they absolutely destroy the base m3 in multicore performance. However, it looks like even the base m3 still far exceeds the 7945HX in igpu performance, power efficiency (probably because 3nm vs 5nm) and slightly beats the 7945HX in single core performance. And the Ryzen 9 is the best AMD has to offer right now, the M3 is the base model (apple still has the pro, the max and the soon to be released ultra). The price of 7945HX laptops also already seems a lot higher than base m3 macbooks (and again I think any ram upcharge is basically pure profit for Apple).

Now i haven't even mentioned AI cores. Even M1 series from years ago still seem far ahead of most laptop cpus today it seems for hobbyists running LLMS. I'm not sure if that's something the 7945HX competes vs Apple with?

[Alpha3031]

Yeah, 7945HX laptops are going to use a dedicated GPU like the 7900M or 4090 because the integrated graphics on there is mostly... vestigial. 2 CU basically means 2 cores (from the last generation) in Apple terms. Rumours are AMD might shove a (laptop) 4070-ish GPU with 40 CU into the next refresh (if I had to guess the numbering, 9950HS) but I don't think they would see a point in going higher than that (if they even try that much), because the higher GPU performance segments are going to be served by their dedicated GPUs paired with the HX CPUs that have 2 CU.

AMD doesn't really have the volume to outbid Apple on the leading nodes, so what they do get is going to keep going to server/HPC. That stuff is more money for them anyway. Adding more cores, both CPU and GPU and shrinking to TSMC's latest node, it isn't easy but it's technically simple enough that they could do it in a few months if they thought there was enough of a market for it. There just isn't, unfortunately.