3 ms·

I don't know the history of non-laptop mobile CPUs. I assume they have pretty much almost all always been ARM. Please correct me if I'm misinformed. That hasn't

I don't know the history of non-laptop mobile CPUs. I assume they have pretty much almost all always been ARM. Please correct me if I'm misinformed. That hasn't factored into desktop/laptop considerations in the past. Currently, it seems to only factor into Apple's plans for unifying their OS X and iOS stack. ARM is growing in servers, but AMD is as well[0]. It's not clear though how either smartphone or server architecture will affect desktop/laptop purchasing for consumers en masse. Of course if we're talking about 2025-2030, I'm sure any predictions I make are a roll of the dice, at best. But right now I don't think there's enough momentum of any players to have absolute certainly about 2025 and beyond. There is a lot of inertia with x86 in desktop/laptop, and so far Apple's Macbook Air/Pro and Mac Mini are the only high performing options on ARM. I like AMD but I'll be happy to see any technological progress that makes significant improvements to our quality of life. Anecdote City: Just five years ago I didn't consider laptops viable for gaming, and now I do most of my gaming on one. The Macbooks with M1 seem like they are capable of some level of gaming, but not "max setting" 1080p gaming, so it's not yet an option for someone like me to switch. But when I'm working, I do most of that on a powerful but very quiet desktop. The M1 chip would not improve my quality of life on my desktop because the efficiency of the chip won't really change anything for me. There's no compelling reason to swap out of my custom built machine to a Mac Mini. Anecdotes are very personal. And so are computers. For many consumers where an M1-based machine work, the benefits are all but lost on them anyway, for a variety of reasons. They don't make decisions based on CPU efficiency - just what they are used to and what features they need and want. If a feature is really life-changing for that particular person they might be OK with change, e.g. switching out operating systems. For a developer, it's either easy to think about switching because you know everything is cross-platform, or it's perhaps impossible to switch because you use exclusive software. [0] https://www.extremetech.com/computing/318217-amd-arm-both-increased-their-server-market-share-in-q3-2020 https://www.extremetech.com/computing/318217-amd-arm-both-in...

ghaff 6y ago
> non-laptop mobile CPUs. I assume they have pretty much almost all always been ARM. All kinds of things over the years. For example, Qualcomm's Snapdragon line. And Intel was definitely pushing to get into mobile at one point. They made a big deal about processor compatibility from mobile up through the server. I still remember at one IDF, they made a big deal about how you wanted to run Intel for mobile (this was pre-iPhone) so that Flash would run the same everywhere. I agree that it's hard to make predictions more than a few years out and certainly x86 has a lot of inertia. On the other hand, there's a lot more abstraction than there used to be and we know there's going to be a lot heterogeneity anyway (GPU, DPU, TPU, FPGA, SIMD instructions, etc.) given the slowing down of CMOS process scaling. So I don't think it's too big a stretch to imagine that we'll see a more varied processor landscape. (I expect ARM to gain share although I don't expect it to dominate on servers--though I have colleagues who do expect that to happen.)

ghaff 6y ago

> non-laptop mobile CPUs. I assume they have pretty much almost all always been ARM. All kinds of things over the years. For example, Qualcomm's Snapdragon line. And Intel was definitely pushing to get into mobile at one point. They made a big deal about processor compatibility from mobile up through the server. I still remember at one IDF, they made a big deal about how you wanted to run Intel for mobile (this was pre-iPhone) so that Flash would run the same everywhere. I agree that it's hard to make predictions more than a few years out and certainly x86 has a lot of inertia. On the other hand, there's a lot more abstraction than there used to be and we know there's going to be a lot heterogeneity anyway (GPU, DPU, TPU, FPGA, SIMD instructions, etc.) given the slowing down of CMOS process scaling. So I don't think it's too big a stretch to imagine that we'll see a more varied processor landscape. (I expect ARM to gain share although I don't expect it to dominate on servers--though I have colleagues who do expect that to happen.)