Is the performance gap so huge? Power efficiency yes, absolutely, but for peak performance last I saw the last AMD vs M3 benchmarks were a slightly slower single core, and a little faster in multicore. Doesn't seem as world changing as described.
I feel like somehow my big Linux desktop with a Ryzen 7950X and 64 GB of ram feels less "snappy" than my M2 Macbook Air running Asahi when doing lightweight tasks, despite the big Ryzen being obviously much better at compilation and stuff. I'm not sure why and my guess was the RAM latency. But maybe I misconfigured something in my Arch Linux...
It has a Ryzen 9900X, 64GB of DDR5, AMD Radeon RX6600XT, 2x2TB Hanye NVMe, ROG B650 ATX motherboard and 850W power supply.
I bought the system mostly to increase the single core performance from the Ryzen 5 3600 I had before. As well as to get rid of all the little 256GB SSD disks I had in the previous one.
If I go look up an arbitrary 850 watt supply (seasonic focus gold), it looks like it wastes about 10% plus 5 watts anywhere from 0-60% load. So extra capacity doesn't hurt that one.
Yes. No other laptop can sustain peak performance as long as the M-series Macs. The only thing that competes is a dedicated desktop with a big cooler and fan.
Mac laptops feel faster, even if the synthetic benchmarks say otherwise.
I don’t agree. Compile times are definitely and very noticeably quicker on my Intel gaming laptop (that’s actually a few years old now) vs my M3 MBP.
That said, I’ve never once felt that the M3 MBPs are sluggish. They are definitely super quick machines. And the fact that I can go a full day without charging even when using moderately heavy workloads is truly jaw droppingly impressive.
I’d definitely take the power performance over that small little extra saved in compile times any day of the week. So Apple have made some really smart judgements there.
In guhidalg's defense, they did say that the "Mac laptops feel faster" (emphasis mine) not that they are faster. There's a trick here with Macs, which is that their user interfaces for the OS and many programs are tightly integrated with the hardware which makes the UI faster--that's the "feel faster"--it's a software, not hardware thing. In cases where the software is equivalent (i.e. cross-platform compilers like GCC/Clang/Cargo) you're going to see little difference, but your OS experience is definitely snappier on Macs.
The arm architecture is also optimized for UI-like tasks, quick to start and stop processes on one of many cores with differing power constraints, whereas x86 is more for workstation-type sustained workloads
M3 vs other high end intel chips on code compilation generally has the higher clock speed always winning. Only with the M4 is starting to hit clock speeds nearer to high end intel chips . We are 2 generations out to probably 5ghz sustained on Apple chips.
I bought an M1 Max with 64gb of ram at release. I'm still not sure what will get me to replace it other than it simply breaking. Maybe an M5 will finally make me want to buy something new. I'm debating getting a cheaper Air and maybe a base Ultra now that I do most of my heavy work at a desk.
I’ve gone entire work days with my Pro on battery because I didn’t notice I hadn’t plugged it in. All my docker containers, IDE etc plugged into my external monitor. It was a good 9hrs before I noticed.
Macs are easy to beat depending on what trade offs you want to make though.
Also, most laptops will run at significantly worse performance when not plugged in. Macs are much more consistent both thermally and when not plugged in.
The power efficiency gap equates to a fan noise gap, and the fan noise/heat of powerful Windows laptops is much more annoying than merely having poor battery life.
I ran some bioinformatics pipelines on an AMD pangolin notebook. Its was faster than the apple M2 (I think it was an M2 or M3) notebook my work neighbor had. My machine had more RAM, but still for workloads that use the extra cores it made a difference.
The performance alone says nothing. What about the battery life, size, weight, temperature, fan noise, and quality of the touchpad? These are important trade offs in any laptop.
When we’re both plugged in and my process finishes 10 minutes faster, it says something. Also the gp post specifically was about performance and specifically not efficiency.
I could get 7+ hours from my work Linux laptops battery and I don’t really care for macOS. The OS quality matters more than the touch pad to me. I’ve come the appreciate a Mat screen. But im glad there is choice.
Everyone values different things and has different requirements, so I’m glad your laptop works out for you. I’m just cautioning against an emphasis on performance: even if the laptop is plugged in, other design considerations will dictate and limit the raw performance.
Yes. You need to go to server class chips (eg. threadripper) before beating the raw multi-core performance of a top-spec M4 Max in a Macbook pro, and the battery life is still crazy good!
What gave me pause was when my base-spec M1 Air handily beat my admittedly old server (Xeon E5-2650v2) on a single-core compute-bound task [0] (generating a ton of UUIDs). I know Ivy Bridge is 12 years old at this point, but I didn’t expect it to be 2x slower.
EDIT: Also, I know the code in the link is awful; the point is doing a 1:1 comparison between the architectures.
It's a laptop, same performance with higher power efficiency means same performance with a much longer mobile uptime, which makes the Macbooks tiers above their competition.
And for data centers, same performance at better power efficiency means hundreds of thousands of dollars saved in power.
My M1 Max MBP is a bit slower feeling than a new Lenovo Thinkpad P16s (Ryzen 7 Pro 7840U) I have, when using VS Code in Linux. But a M3 Max MBP I have blows it away. If you run Linux, and care about MacOS apps or need non-dev stuff like Outlook, then a Linux AMD laptop can be a really cheap, fast option. Unfortunately the manufacturers don't want to load them out... like my AMD chip supports 128GB, but no laptop manuf. will lay down more than 64GB.
