Hacker Newsnew | past | comments | ask | show | jobs | submitlogin

No, what’s worse is Apple, or is it random self-styled influencers, spreading the false narrative that Apple Silicon magically requires less RAM and people actually believing that.


No, but Apple tends to have faster RAM (way more bandwidth), faster SSDs, and macOS tends to have a better memory compression algorithm.


> macOS tends to have a better memory compression algorithm

Compared to what? Linux? Windows?

Are there any published benchmarks anywhere, otherwise this just proves my point above.

> Apple tends to have faster RAM (way more bandwidth), faster SSDs

Yes, sure, due to its integrated nature, but that does not reduce the RAM requirement. My 8GB M1 MBA, which is used as a home browsing-only laptop, is almost always in yellow on memory pressure once we have a few tabs open.


But that’s what I said: My post starts with “no”. The other points in the “but” are all still true.

The RAM or SSD are specifically not faster because of its “integrated nature”. The RAM is faster because Apple engineered an actually wide bus + multiple channels for memory access. The fatter Macs have the equivalent of up to 8-channel memory, which not even server CPUs of the competition provide. The SSD gets easily 7 GB/s reads in my testing. Both Windows and Mac have a working memory compression algorithm and a sister post claims that it works better due to hardware acceleration, which I’m inclined to believe. Memory compression on Linux is a mess as you cannot keep compressed memory pages in physical RAM, instead forcing you to partition the memory and manually handling how much and what to put in compressed zram.

Low memory situations are thus handled better than on other operating systems. Memory and memory pressure are neglected concepts in the competition, both the hardware providers and operating system providers.


> The fatter Macs have the equivalent of up to 8-channel memory, which not even server CPUs of the competition provide.

8 channels has been standard in Epyc chips since they released in 2017, the latest Epyc Genoa does 12 channels. They're also not splitting that memory bandwidth with a bandwidth hungry GPU.


The memory compressor is not higher compressing than Windows, but it is hardware accelerated, so it's much faster.


Again, any published benchmarks that can be referred to?


No, but I just told you.


So we should just believe an anonymous stranger on the Internet.

Got it!


Who's anonymous? My account is under my real name.

Anyway, sometimes people know what they're talking about.


SSD performance scales somewhat linearly with capacity, and modern 1TB drives are capable of 6+GB/s reads. The 256GB M2 does a measly 1.4GB/s, and from what I've seen the 1TB is also lackluster at ~5GB/s.


https://m.youtube.com/watch?v=hmWPd7uEYEY

I am sick of people making claims without quoting any numbers about real-life performance.




Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: