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

> I disagree, and I always upgrade the RAM even though I don’t like paying more. Why? Because it puts pressure on developers to keep the bloat under control.

Shouldn't you be doing the opposite then? Keeping the baseline amount so you can know what it's like for people without a large budget and stop patronizing the applications without acceptable footprints in that circumstance?

> In that world you’d have all the software growing in memory use without bound.

We already live in that world. In the 90s you could run Netscape Navigator on a machine with 8 megabytes of memory. I've seen individual browser tabs use more gigabytes than that.

And not all of this is Electron bloat. The Stable Diffusion XL model is ~13GB. In general the quality is going to be proportional to the model size. So for the thing to get better, people need machines with more memory. And 8GB is already too small.



Shouldn't you be doing the opposite then? Keeping the baseline amount so you can know what it's like for people without a large budget and stop patronizing the applications without acceptable footprints in that circumstance?

I'm not a developer of native Mac apps. If I were, I would definitely have a baseline machine for testing.

The Stable Diffusion XL model is ~13GB

That's not a baseline consumer application. See my other reply (re: grandma and little Billy). If you're developing a native Mac app for grandma and little Billy, Apple probably doesn't want you shipping a 13GB model with it. This is an example of the point I'm trying to make: find a way to compress the model so the end user doesn't have to deal with that kind of bloat (or host it in the cloud).


> I'm not a developer of native Mac apps. If I were, I would definitely have a baseline machine for testing.

You expect every developer to buy a second Mac? They're all doing what you're doing and paying more for the machine with more memory because other applications need it, and then their application runs fine on that machine so they don't even notice the problem for the people with 8GB.

> That's not a baseline consumer application.

It will be before any of these new machines go out of support.


> You expect every developer to buy a second Mac?

It's very easy to check the memory usage of your project (it's kinda in your face on XCode). If you do a few minutes test or let the app stay open for a few hours and usage has ballooned to a few GBs, that usually means you have a leak somewhere to fix.


But the developer of exactly the app using all your memory is writing it in Electron from Windows and has never used XCode.


And is there any wonder Electron has such a terrible reputation?

The thing that really bothers me is that if you look at what regular home and office users were doing on their computers in the 90's it's almost identical to what those users are doing right now, except in the 90's they had orders of magnitude slower computers with orders of magnitude less memory. Yet in many cases those 90's computers were MORE RESPONSIVE than what they have today.

All of those countless billions of investment in technology hasn't done a damn thing for the productivity of Sally the office worker or Billy the 6th grader. Arguably, it's made their lives worse (viz. social media's deleterious effects on mental health). Now everyone's pushing the heck out of AI and all I see is high schoolers using ChatGPT to cheat themselves out of an education. They can't read (critically), they can't write, they can't even spell!

So in light of all that, why should we be pushing more and more computing power (and memory, the original issue) on regular users who aren't getting any benefit (broadly, to their way of life) out of it?

Gosh, now I sound like a luddite!


> And is there any wonder Electron has such a terrible reputation?

Certainly not, but you can't fix it by putting less RAM in the machines of people with budget constraints. The developers will just pay for more themselves and then not care about those people because people who can't afford RAM generally aren't lucrative customers.

And it's also worth considering what actually causes this.

Developers want their code to work on every platform. They don't want to write different code for each platform. But each platform wants them to have to, because that makes it more likely there will be software that only works on their platform, or that doesn't work on some new competing platform. So they refuse to develop or implement cross-platform standards.

Then someone else has to do it, but that's rather a lot of work, and it turns out the easiest way to do it is to piggyback on the work already done for browsers to make them work on every platform. That's Electron. It's terribly inefficient but it saves the developer a lot of porting work, so it's widely used.

If Apple doesn't like this, they should provide cross-platform native APIs for developing applications.


If Apple doesn't like this, they should provide cross-platform native APIs for developing applications.

It’s not in Apple’s interest to do that. It would cost a lot of money to develop and only benefit the competition. It would also slow down Apple’s own ability to innovate on the APIs until the competitors catch up.

Or are you saying Apple should develop the APIs for Windows and Linux as well? Why would they do that?


> Or are you saying Apple should develop the APIs for Windows and Linux as well?

What they should do is provide open source implementations of their APIs for Windows and Linux, i.e. make them standards.

> Why would they do that?

Because then people would use them instead of using Electron.


There’s a lot of profiling tools. There’s even one in the browser’s inspector. It’s not something exclusive to XCode. You can even use Task Manager/Activity Monitor/System Monitor in a pinch.


>We already live in that world.

Nah, the treadmill stalled out years ago.


The CPU treadmill did because Moore's Law petered out. RAM keeps getting cheaper and applications keep using more of it.


RAM usage has barely budged for years. Taking the '90s as baseline again: In 1991, you could get a brand new PowerBook 100 with 2 MB of RAM. In 2001, you'd get a new PowerBook G4 with 128 MB of RAM, a 64-fold increase. But in 2013, a MacBook Air came standard with 4GB, and we're looking at only 8GB in 2024.


You're taking as your example the thing people are complaining about. 8GB of DDR5 is ~$24 retail, and that's the amount Apple is putting in their $1000 laptop. PCs of the same price typically have 32-64GB.

You're also using a time period that includes COVID and when the DRAM manufacturers got busted for price fixing.


>PCs of the same price typically have 32-64GB.

That's not a run of the mill consumer set up.

>You're also using a time period that includes COVID and when the DRAM manufacturers got busted for price fixing

No, I'm actually including two separate time periods where they were busted for price fixing. They were doing it in the late '90s too.


> That's not a run of the mill consumer set up.

It kind of is, e.g. Steam hardware survey has more than 80% of people with at least 16GB of RAM and almost a third with 32GB, and that's a measure of installed base rather than new computers.

You can now find 64GB in <$550 laptops:

https://www.newegg.com/p/1TS-000D-110E6?Item=9SIA7ABJ459240

> No, I'm actually including two separate time periods where they were busted for price fixing. They were doing it in the late '90s too.

They started doing it at the end of the '90s. By then PowerBooks were already coming with 64MB and PowerMacs with 128MB:

https://everymac.com/systems/apple/powerbook_g3/specs/powerb...

https://everymac.com/systems/apple/powermac_g3/specs/powerma...

In the second case it started only 3 years into your measurement period instead of 7, and then right after that was COVID. It's only now that the prices are starting to resume their historical downward trend and they're still slightly above where they were when the price fixing started in 2016.

But that explains why it's not a factor of 64 during this period. It's still the case that 8GB of DDR5 is ~$24. What reason is there to not include $100 worth on a $1000 machine? Or, if some excuse for that could be generated, why isn't there a $1100 machine with four times as much?


> But in 2013, a MacBook Air came standard with 4GB, and we're looking at only 8GB in 2024.

You're saying "RAM usage", but your evidence is "RAM provisioning", which is the entire basis for the criticism. The reality is that RAM usage has increased significantly, but Apple has been stuck by the addressable memory limits they've baked into their architecture.

As another point of comparison, a 2013 iPhone 5c (technically not released until the last quarter of 2013, but we're being generous here) had 1GB of RAM and 8GB of storage, though you could upgrade that to 32GB. A modern iPhone has 6GB of RAM, and comes with storage from 128GB to 1TB.

Maybe the bytes are growing more on iOS? ;-)


>As another point of comparison, a 2013 iPhone 5c (technically not released until the last quarter of 2013, but we're being generous here) had 1GB of RAM and 8GB of storage, though you could upgrade that to 32GB. A modern iPhone has 6GB of RAM

Fine, take that as your reference. It still shows that growth in memory was 1000% faster not too long ago.


No, it doesn't. It shows that Apple's RAM offerings have grown at widely different rates depending on their product. Apple has been stingy on memory compared to the broader industry. The biggest EC2 instance you could get in 2013 had 244GB of memory, but today you can get one with over 24TB of memory.

...and that still proves nothing about the growth of RAM usage in that time period, because the size of the offerings in individual products are largely independent of the increases in RAM usage.




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

Search: