Hacker News Reader

> I love the idea of the Framework 13" machines with Ubuntu. Almost same weight as a 13" Air and with strong upgradeability. Disadvantages (to me) are the battery life on Linux is significantly less than an Apple device (although hard to find exact numbers

As one battery life datapoint, I have a Framework with the previous gen AMD board and the 61wh battery, and in Linux Mint with no special configuration I get about 7 or 8 hours in normal use (with wifi+BT+average backlight, just doing normal browsing/file editing, not maxing out CPU rebuilding a massive project for hours). That's totally fine for my needs, it's effectively a full workday or an intercontinental airport + flight without power. I'm very happy with it and the upgradeability has been great (I upgraded mainboard, and now having the old mainboard running as a home server).

Lacking power? Below buying the very best Intel or AMD CPU which needs a new mainboard and RAM, I seem to never get any PC upgrades in the region of even an M3.

For mobile, this kind of performance is insane. I usually am happy if it is not a netbook CPU, since my 4th gen i5 dualcore+HT is still up to anything I want to do with it.

That’s something that I’m puzzled about. If we take a look at both PC and Mac Geekbench results which use the exact same Intel CPU as a baseline, Macs wipe the floor with any Intel or AMD professor. Am I reading it wrong or?

[1] https://browser.geekbench.com/processor-benchmarks

[2] https://browser.geekbench.com/mac-benchmarks

Geekbench is biased towards Apple Silicon for whatever reason. Take a look at any other benchmarks, preferably ones that test real world workloads.

This is a claim that has never been backed up by anything verifiable.

Could you expand upon it?

I am not suggesting that Geekbench is intentionally favoring Apple Silicon. The code the benchmark runs just happens to run better on Apple Silicon. And Geekbench also poorly scales with higher core counts so multicore scores are almost useless.

IMO with benchmarks it's never a good idea to rely on a single score. You should compare many different scores. Preferably with benchmarks that test workloads you will actually make use of.

But even then, what benchmarks contradict it? You’re claiming an inherent bias, but other benchmarks also run just as well on Apple Silicon when normalized for core count. Cinebench, Specs etc…

What do you recommend?

That sounds about right. Of course depends on the workload, but e.g. when compiling Rust code, having a Apple M-series chip makes a huge difference. That alone would make it hard for me to consider switching to any non-Apple laptop.

From personal experience, I'd also say that there is a noticeable bump in performance between M Max and M Pro chips (= running same workload as colleagues on identical specs apart from the chip), that isn't really apparent in the benchmarks here.

No, the bench results for Apple laptops are accurate (since there is close to no variance in hw config). It is the x64 processors that appear worse than they actually perform.

Might be an issue with the test methodology for x64 processors, eg. cooling and RAM speeds, SSD speeds etc. You should run geekbench on your own machine. C++ and C project compilation benchmarks I ran more or less match what I saw with Geekbench (in terms of %diff).

Basically, what I got is as follows (fully spec'd M4 MBA vs Ryzen 5900X, WSL2):

* for multi-core, my M4 Air is only 25% slower than my Ryzen 5900x desktop

* for single-core, my M4 Air is 0-20% faster than my 5900x

Compared to the M1 Air, it is almost exactly 2x as fast (Apple's claims are true in this regard), has slightly brighter display, and much better speakers. The "2x as fast" is sometimes actually "4x+ as fast" (for some workloads) because you have to take thermal throttling into account.

I just went through this process and ended up getting the HP Omnibook Ultra Flip with a core 7 258v, running Ubuntu 24.10. Performance is excellent, battery life is the best I've ever had on an Intel running Linux. If you truly need more RAM than 32 GB it won't be an option though.

The air or an mb pro seems so nice until I remember the sting of dealing with the dev environment and docker on a Mac compared to Linux. No amount of battery life or marginal jump in performance (which gets lost through needed virtualization) will make up for that for me.

It does seem quite nice. Is the pen input decent? If the build quality is OK the pricing is quite good, in France with 32Go/1To is 650€ cheaper than an equivalent MacBook Air. I have not looked for a proper CPU comparison but I don't that it matters that much for a device of this type, and as you said, with Apple Silicon you lose a lot of the performance once you venture away from optimized software (which is still quite restrictive).

Considering the (relative) failure of the ARM push for Windows PCs I doubt ARM is going anywhere in the consumer market (that probably means it will stay niche in the server market as well but we'll see).

> even with the new Ryzen AI Max processors and the DDR5 memory, speed is much lower compared to the M-series soldered on a chip

Are you referring to memory bandwidth specifically? Yes, that difference is not worth it. Even if you fully load the CPU and GPU on Apple Silicon you will not come close to using all that bandwidth [0]. You're better off comparing real-world benchmarks instead.

[0] https://www.anandtech.com/show/17024/apple-m1-max-performanc...