Hacker News Reader

Comment by aurareturn | Hacker News Reader

Many Lenovos and Dells also started with 8GB in 20223.

Yes, Cook is not a product innovator. But I think market forces had more to do with it:

1. Cloud computing moved heavy workloads to data centers so RAM usage is lower locally

2. RAM $/GB has slowed down considerably starting at around the time Cook took over. [0]

3. Businesses/people focused more on mobile software and hardware performance over the last 15 years. More people are doing business on their phones.

[0]https://aiimpacts.org/wp-content/uploads/2020/04/memorytrend...

View on HN

Consumers are not offshoring their chrome tabs to the cloud

If Chrome knows that users leave open an insane amount of tabs which kills performance why wouldn’t they have a built in mechanism to suspend the content on the tabs until it’s necessary? Why does there need to be 3rd party ad on to do the same. Why doesn’t the Chrome team write unbloated code? The binaries just keep getting bigger and bigger. While consumer hardware typically stays about the same with less incremental upgrades.

Ehh... Chrome does exactly that without plugins.

They do have a way to suspend inactive tabs though. The feature is literally called Memory Saver [0].

[0] https://blog.google/products/chrome/new-chrome-features-to-s...

> Why doesn’t the Chrome team write unbloated code?

They're busy rewriting internet protocols so they benefit Google more.

> If Chrome knows that users leave open an insane amount of tabs which kills performance why wouldn’t they have a built in mechanism to suspend the content on the tabs until it’s necessary?

It's built into Chrome now but it's useless. Because it's cheaper to glue together a few memory leaks ^H^H^H javascript frameworks than write something that has constant ram usage for the same working set.

What does tab suspending help with when a google sheet with not much on it starts at 3-400 Mb ram usage and ends up at 1.5 - 2 Gb in a few weeks if you don't close the tab?

The main issue is that memory reporting tools don't handle paging well. You cannot tell if the ram is actually being used. You can only tell that some app has X amount in the allocation table which is often more likely backed by files on storage than actively used RAM.

When you have lots of chrome tabs open the ram isn't being actively used. You may be using a tool that shows total memory allocated. I'd actually recommend looking for a tool that shows page table misses (points where the OS has to load from disk into RAM). Huge numbers of page table misses and actual noticable performance drops are more indicative of any actual problem than a simple allocation which can easily mislead just by having large files mapped.

They don't handle paging well, they don't handle memory-mapped files well, they don't handle shared libraries well... Memory reporting tools are generally crap for non-experts, and pretty poor even for experts.

I had 50+ tabs opened on my M1 8/256GB as a dev while running an IDE and a few local servers.

It was fine.

I hardly think that the general mass has problems with Chrome on a 8/256 Apple Silicon Mac.

If we are doing anecdotes, a single tab Google doc with Grammarly running can easily surpass 4Gb of ram, especially if the doc is over about 30 pages. Use cases like that are not particularly uncommon.

But did it affect the overall user experience?

I was a heavy Google Docs user myself on my 8/256 M1 Air.

It's not actually using 4gb of ram. Pretty much every memory reporting tool cannot handle paging and will report total allocated. You can create a one line c program today that just calls mmap(file) on a large 4gb media file and your os will claim 4gb used. Actually measuring real physical ram usage is hard. I'd be looking for page table misses (and the likely noticable resulting performance issues) as a better indication of any potential issues.

Yep one thing in this discussion is that people have no idea how to read memory usage. A 4GB set of videos that are memory mapped will use a few megabytes of ram as the current portion of video is paged in. It will show 4GB of memory used in pretty much every tool. Chrome suffers from this as browsers work with lots of media that may or may not need access as the page is scrolled around. Using the mechanism all OSs have to deal with this (memory mapping) will report GB of ram used. But it will still be responsive! It's not actually using that much RAM!

I see so many comments along the lines of 'chrome reports xGB of ram usage when I open tabs. You can't get by with 8GB of ram!!!!' yet the computer will be running smoothly. Just as my old iPad with 256MB of ram runs smoothly with those same tabs open in chrome.

You might need more than 8GB but please don't try to claim people can't do the basics with only 8GB. I have an 8gb MacBook Air as a lounge PC. A PC used for the basics when I couldn't be bothered firing up the big gaming rig. Unsurprisingly the 8gb laptop is better than my still perfectly capable 256mb ram iPad running the same apps although the iPad was still decent at the role in any case.

The claims that 8GB is completely useless and can't even do web browsing because Chrome shows 8gb ram usage is so obviously stupid and false. Many are probably reading such claims on a portable device running the same browser with less RAM. Yet media and even tech sites are posting this crap repeatedly.

It was paging a lot I guess, any metrics about it?

What do you know about consumers and their computer usage?

Local RAM may see a comeback with local/private LLMs.

Agreed. If RAM demand picks back up due to local LLMs/GenAI, Apple can increase the base to 16GB and still have plenty of people who will upgrade to 32GB+.

The situation right now is that Apple wants the 8 to 16GB upgrade money. But if they increase the base to 16GB, there aren't enough 16GB to 32GB upgrades to cover the deficit. This is due to the reasons I posted previously.

I'm currently on a 16GB MBP M1. For my next upgrade, I'm considering an M4/M5 Max with 128GB+ to run local models. Without local GenAI, I have little interest in more than 32GB of RAM. 16GB is actually enough for me.

SSD speeds have also increased. Prior swap space was incredibly slow, where as today it’s slower than ram but not always human noticeable for many tasks.

But ssds are not repairable so it’s going to decrease life of your primary storage.

I’d be fine with my MacBooks 8gb of ram if there was a m.2 slot for storage

I maintain a fleet of 100 PC/Macs. I have never needed to replace an SSD. I have replaced plenty of non SSD drives.

Even the 8gb machines will age out before the drives do in most cases.

An M1 could be the first machine that could have mainline support and work for the most part for over 10 years. Plenty have people have done that in the past but it wasn't a pleasant experience.

I have an m1 MacBook Air and it’s fine for most people.

I’m not most people, I’m here! I would prefer a replaceable ssd.

So I modify my usage of the machine because I rather not kill the ssd. Rather than use it as I see fit and repair wear items.

But the keyboard is also a weakness but not essential like boot storage

Swapping constantly kills SSDs though, this is not a good argument. The reality is that power users who should buy an upgraded model like to complain about the base one and the ones petty enough to deliberately go with the 8GB report quick SSD wear because they are overloading it with processes

> Swapping constantly kills SSDs though

This is the conventional wisdom around memory management and swapping yet in my experience a dead SSD in a Macbook is pretty uncommon. They seem to die of about 10 other things first, well into their useful lifespan, including physical wear and tear, worn out ports, etc.

The situation of base 8GB macbooks wearing down their SSD in the hands of power users is well documented but it's extremely specific.

Normal users with 8GB and power users with >= 16GB indeed have no SSD problems because they swap very infrequently

> The situation of base 8GB macbooks wearing down their SSD in the hands of power users is well documented but it's extremely specific.

I can't really find it documented much at all. Conducting a web search reveals that 90% of the user documented concern around this has nothing to do with actual failures but "I'm worried about my SSD failing" or "Will your SSD fail?" discussion.

If it is rare, extremely specific, etc. then I think that bolsters the case that it doesn't make sense to increase base RAM out of that concern alone.

Power users with 8GB macbooks being worried about wearing down their SSD is very well documented, but if you have information on how often it's actually a problem I'd love to see it.

SSD speed as measured by bandwidth.

The latency numbers have barely moved in a decade. And swap in operations are 100% a latency test, best reflected by qd=1 random 4k read (well 16k on a mac, but the point remains since the IOP numbers will be roughtly the same with a 4x larger bandwidth at 16k). It's rare to see an SSD that can even do 100MB/sec qd1 random read, and the numbers largely seem to be going backward as the number of bits being packed into each cell increases (ex MLC->QLC) on many products.

Bandwidth has always been easy, I had spinning RAID arrays that were faster than your average SSD 15 years ago if the only measure is bandwidth. This is why the xpoint disks were a big deal. 2-3x faster QD1 iops numbers.

I think it’s less about $/GB (though that always matters) and more about power usage. As much as people love having 16GB of RAM, the extra memory will get used (fill RAM then swap right?) and active LPDDR5 memory uses quite a bit of power I think.

Low Power DDR5 [0][1] uses 20% less power than LPDDR4X.

[0] https://semiconductor.samsung.com/dram/lpddr/lpddr5/

[1] https://www.jedec.org/news/pressreleases/jedec-updates-stand...

My 18 Gb mbpro is currently at 80% battery and estimates 12 hours left. Isn't that enough? ;)

Reply on news.ycombinator.com