Consumer targeted arm chips power most phones and tablets. Many of them are more than a handful for i5 and i3 class Intel chips and draw a lot less power. When you are talking about a $599 price point, you aren't talking about top-shelf Intel, anyway.
> But there is some sense to trying to lay the groundwork for future ARM SoCs running Windows
Now that ARM has grown up to be a viable alternative to Intel, this makes a ton of sense.
Outside m1/m2 there isn't any chip that can be compared to i5 intel 13 gen especially for PC space [that is in the same price bracket]. (maybe some arm chips can get close to i3 12100f but you can get it for 90 usd and get good single core performance so im not sure if in this price point is even any arm alternative)
Intel & AMD’s consumer and/or server market share.
I can’t imagine going back to an Intel after an M1. The battery life is better. It’s dead silent. It doesn’t get warm. It’s like a totally different kind of object.
I know PC people (those who want Windows) aren’t interested in an M1 Mac. That’s fine.
But I see PC laptop reviews with 4 or 6 or maybe 8 hours of battery life. They get hot but the fan isn’t “too loud”. And I know the performance isn’t the same.
And I just wish reviewers would call it out. They’re not on the same level. I’m sure fanboys would complain about the comparison in every review, but why shouldn’t Windows users have something much better? It’s been proven possible. Hold AMD/Intel/Qualcomm to the fire more.
If Qualcomm could get a chip with reasonable performance at a reasonable price that just doesn’t get hot and waste all its battery playing space heater, I bet they could really get a hold on the laptop market.
The vast majority of PCs are running 128 bit wide memory, with workstation CPUs like the threadripper (and pro) being the exception, but a VERY small fraction of the market.
The M1 has 128 bit wide 67GB/sec peak (that you'll never see) bandwidth, like most PCs. Upgrade to the Pro and you get 200GB/sec. Max will take you to 400GB/sec, and Ultra takes you to 800GB/sec.
On the Intel (i3, i5, i7, i9) or AMD (ryzen r3, r5, r7, r9) you get ... the same memory bandwidth. Check the 8 core vs 16 core scaling numbers and for most benchmarks you'll see poor scaling. Sure you can increase GPU performance by adding GPUs, which reduces (but not removes) the need for extra memory bandwidth. Sadly iGPUs (outside the XboxX and PS5) largely stink and are only good enough for non-GPU intensive workloads. Apple on the other hand does scale GPU performance, granted not to the levels that AMD and Nvidia do.
So why can't anyone in the PC space do more memory bandwidth and a decent iGPU, especially when for years the GPUs were in short supply and had exorbitant prices. I think it does come down to OS support, volume (which could be problematic if current GPU customers avoid you), and potentially reducing profits for AMD (who would have sold an expensive external GPU). Not to mention that fast/wide ram requires soldering chips on board or increasing size/cost with large banks of ram. Even servers with 8 memory channels (minimum 8 dimms) only get you to the M1 pro level (1/2 of the m1 max and 1/4th of the m1 ultra).
Apple can say we have X% of the market today, and all new customers will be on our new platform with 2 years, so the driver, OS, iGPU, memory bandwidth, etc will be amortized over substantial volumes. Additionally Apple gets a larger fraction of the revenue, since they aren't paying Nvidia or AMD for a GPU. Who is going to push a MBP or Apple studio competitor that could ship the same volumes?
The best of market ARM designs don't really compete head-to-head with Xeon right now, but there are still a ton of server applications where they make already make sense. As a simple example companies like Google and Facebook have hundreds of thousands of servers that are doing things like running memcached or running some application like D/GFS where the server is mostly just doing a lot of I/O and doesn't necessarily need really beefy single-threaded CPU performance.
Longer term obviously if there are ARM or RISC-V CPUs that can compete head-to-head with Xeon in terms of features and single threaded performance then that opens up pretty much the entire enterprise/server market.
