Friday, July 9th 2021
Qualcomm Wants to Build an M1-Like Processor for PCs
Qualcomm is trying to get into the PC space with their mobile Snapdragon chips, which offer great battery and decent performance. However, so far only Apple managed to get the right formula for developing custom low-power, high-performance chips. It is exactly Apple's M1 processor in question that Qualcomm intends to mimic. According to the recent interview with Qualcomm's new CEO Cristiano Amon, we are informed that Qualcomm plans to produce laptop chips that would directly compete with Apple's. That means that, despite the ecosystem differences of Apple M1 (macOS) and Qualcomm Snapdragon (Windows-on-Arm), the company wants to deliver equal if not better performance and great battery life.
With the recent acquisition of Nuvia, Qualcomm has a team of very talented engineers to back up its claims. The company also recently hired some of the developers behind Apple's M1 chip. The company notes that it will be using only the best solutions for its upcoming SoC, which will include a 5G modem. Mr. Amon has also noted the following:
Sources:
Reuters, via ArsTechnica
With the recent acquisition of Nuvia, Qualcomm has a team of very talented engineers to back up its claims. The company also recently hired some of the developers behind Apple's M1 chip. The company notes that it will be using only the best solutions for its upcoming SoC, which will include a 5G modem. Mr. Amon has also noted the following:
We needed to have the leading performance for a battery-powered device. If Arm, which we've had a relationship with for years, eventually develops a CPU that's better than what we can build ourselves, then we always have the option to license from Arm.
37 Comments on Qualcomm Wants to Build an M1-Like Processor for PCs
He quite anonymously announced so, did he? :-P
Interesting. So they want to build something like that? :D
Thanks for pointing that out!
= Battery
Batterd cod and chips G.
8cx is this already?! Too.
the bleed...
the deed...
the greed....
the need...
the steed....
Apple had success in that even before M1, so it wasn't literally out of nowhere as they had years of experience with their custom iPhones architetures. So, I think even if Qualcomm is to be successful in their endeavor, it will take many years.
Semi-custom but still, they have some experience in the area.
Apple chip are still based on arm so are also semi-custom
The problem is Windows software is hard to move off x86/x86_64 (thanks, closed source software), the capabilities of the SoCs aren't that relevant.
*not to mention their server SoCs: en.wikipedia.org/wiki/Qualcomm_Centriq
People that want a Mac know why they want a Mac, it’s more than just M1, it’s also the macOS experience and the Apple ecosystem like iMessages, Final Cut, Airplay and so on. Qualcomm’s chips, no matter how powerful they become, will not enable these users to use macOS.
Qualcomm will be relegated to running windows, and such, its more direct competition will be AMD and Intel x86. And those two companies will be firing on all cylinders. Especially now that we know that Intel and also AMD will be pursuing hybrid architectures. Seems like Qualcomm’s focus should be on supercharged X86, not M1. And as we know, the snapdragon experience on windows currently sucks.
Qualcomm has framed this argument as M1 competition. So the questions are: which segment of the market will manufacturers target with these Qualcomm chips, and who will potentially be interested in purchasing a device with a chip that directly competes with M1?
Given the existing high cost of snapdragon laptops (that are very easily outperformed by x86 and m1), we’re not talking about the smartphone or tablet markets, we’re talking about the laptop market.
Yes there are chromebooks, but most people who buy chromebooks don’t do so for M1-level performance. Also, most people spending thousands of dollars on laptops aren’t doing so to run Linux as their primary OS.
This market segment is primarily running either windows or macOS, i.e., x86 and m1. This is the segment of the market that this new Qualcomm chip will compete in. And x86 is the primary competitor, not m1. Tiger Lake and Ryzen 5000 already perform quite well and yet the Mac is flourishing. Qualcomm is not going to lure many Mac users away from the Mac. Instead, it will have to lure would-be x86 windows users away from x86. And that is an uphill battle, given the rumors of what’s coming from Intel and AMD.
And even if Qualcomm does gain a foothold in the windows laptop market, it’s not clear to me why its chip is being framed as a super-competitor to M1 when I’d argue most Mac users aren’t interested in purchasing a windows laptop in the first place.
Perhaps I’m missing something, so I welcome others’ perspectives.
If you can run the Linux kernel (and you can run it on everything from ARM phones to servers), running apps on top of that is easy-peasy.
But I don't see what that has to do with macOS.
For a list of places Linux will go, see: en.wikipedia.org/wiki/List_of_Linux-supported_computer_architectures
www.theverge.com/2021/1/21/22242107/linux-apple-m1-mac-port-corellium-ubuntu-details
So while Linux runs on ARM, existing support is for the standard cores. The M1 is custom and is undocumented, so while Linux kernel 5.13 does support booting m1, the problem remains that the m1’s GPU as far as I remember is still unsupported. So good luck booting a linux GUI like Gnome or KDE with m1. A ton of reverse engineering needed to take place to even port Linux to m1.
I’m linking a quite fascinating read of the project to port Linux to m1 here: asahilinux.org/2021/03/progress-report-january-february-2021/
If you look closely under the chip it says Intel Inside.
So to do x86 to ARM you have to decode the Micro ops of the x86 instruction and convert them into arm instruction.
The opposite is a bit harder. You have to look at all the instruction, figure out what they want to achieve and regroup all theses simple instruction into the corresponding x86 instruction.
So x86 emulation on ARM is way easier than the opposite. One is a decoding work (and now in Rosetta 2, it can generate new binary at the installation in arm, so no real time transcoding), while the other is almost a compiler work.
But back to subject, for low power system, having the option to not have a heavy front end that decode cisc instruction into risc is a great power saving. less power mean less heat. Also these use die space that arm don't have to bother with. (arm have a front end but it's much simpler than a x86 front end).
Anyway, if Qualcomm is serious, they could make a very good arm process for Windows and Linux. Both OS could simply decode and convert x86 to arm. It's just a matter of wanting it.
In the end, the best scenario for everyone would be to have outsider getting good product that people can use that use low power and have high performance. Since x86 is limited to a duopoly, it will have to come from either ARM or RISC-V.