Why the New Macs Are So Short of Memory | Hacker News Reader

Why the New Macs Are So Short of Memory

forbes.com

31 pointsfortran7719 comments

This should include a (2020) in the title. The situation for the new iMacs is even worse with 8gb ram max. It can't be because the additional GPU Power necessary to properly drive the 4.5k display limited available space on the die for the chip as the ram is not on chip but on SoC ? Is it that the demand for high capacity ram modules is outstripping supplies and Apple wants to conserve what it has for a MBP 14/16' launch ?

They are configurable to 16GB. The Air can drive a 6K external monitor already at 60hz (in addition to the built in display), and by now multiple benchmarks have shown that performance is basically the same for 8 vs 16GB on the majority of workloads, except if you’re rendering multiple 4K video streams at once.

While this is true for the benchmarks, the UX experience may not be the same between an 8GB and a 16GB depending on your use case.

My use case: Mail, Calendar, Messages, Signal, Safari running at all times. Safari running an adblocker and with 1-3 tabs open. Firefox open as needed, adblocker, no more than one tab. Lightroom as desired.

8GB MacBook Air: slowdowns when typing. Slowdowns when switching apps. Returned.

16 GB MacBook Air: zero slowdowns, snappier than my old i9 MacBook Pro 16.

Could I have adjusted how I used my Mac and gotten a better UX? Probably, but I have been working that way for 5+ years.

That's sad. 8GB should be more than enough for that workload. Sometimes I feel like we are regressing.

I would expect the workload as described to be fine on M1 Macs too (perhaps excepting Lightroom). If I had to guess, something in the background was adding memory pressure. For instance, I’ve seen the Spotlight background processes (mds, mds_stores) take over 16 GB. And those definitely work hard during the first run.

Keep in mind the new M1 macs have pretty decent CPUs (wins handily over the Intel I9 on several benchmarks) and a quite fast SSD.

I believe it's sustains around 5GB/sec to the SSD and I've seen a fair amount of hand waving about custom hardware support for compression/decompression and some OSX magic for compressed swapping.

So 5-10 years ago you tended to have spinnings disks, or SSDs connected over SATA (a bandwidth bottleneck), and SSDs not designed for nearly the IOPs available today.

So while swapping could be really annoying 5-10 years ago, it's not necessarily so today. Switching tabs or apps is not necessarily going to be annoying if some swapping happens.

Numerous reviews of things that usually required a ton of ram (like editing multiple streams of high resolution video) have shown that 16GB ram M1s do quite well compared to previous generation machines with more ram.

With all that said, I'm anxious to see future apple products with more ram and M2/M1x or whatever they call the follow on chip.

The new iMacs are not 8 GB RAM max. They can be configured with 16 GB

Why in general has RAM been so limited in computers for such a long time? A decade ago we were expecting new kinds of RAM to replace hard disks entirely, using new computer architectures that did not use disk drives at all. Instead most general use computers have been stuck with 8GB or 16GB of RAM for like a decade. A few go up to 128GB. It seems like RAM has been artificially limited as a sales feature (price tied to RAM size as a feature customers easily understand similar to CPU speed), which has limited greater use of it. Also similar to the way the size of SD cards that can be handled in devices are never future compatible (got a new 1TB card? Good luck using it in your older phone!)

Perhaps there are national security reasons for this? Like a computer that has 10TB of RAM could simulate nuclear bomb reactions to a degree of fidelity and with a speed that would give adversaries new capabilities? I don't know, but I know it seems like an artificial limitation. Possibly there are hardware limitations (like more RAM uses more power, but RAM could be turned on and off and speed managed just like CPU's are, depending on dynamic need).

CPU's are currently limited by the I/O speed of data, so moving all data to RAM could bring huge performance benefits, yet current computers are still very much limited by RAM sizes and RAM-CPU integration.

If Apple can get away with decent performance at 16G why would they ship more in the first generation? A few people won't upgrade yet because they need more memory, but a lot of people will upgrade again when they release the 32G models in a year or two.

they're giving you the bare minimum product at the highest possible price as it it increases profit margins.

So far every M1 Mac has had the same entry price point (except IIRC the MacBook, which was priced lower), with comparable or higher specs. They very seldom increase (or decrease) margins on a given product line, and I seriously doubt this is an exception.

They tend to target a feature set, margin (generally ~30% on Macs), then price point—usually in that order. Given the pricing so far, my impression is they’re generally taking reduced costs as an opportunity to improve the entry level feature set.

> they're giving you the bare minimum product at the highest possible price as it it increases profit margins.

You can buy a Mac mini with M1 with 8GB of RAM for around 800€, and the same Mac mini M1 with 16GB of RAM for around 1030€. That's hardly "the highest possible price"

With flash memory performance as good as they're getting now, swapping is not the bottleneck it was with spinning disks. The pain of being memory-constrained is not so painful anymore in many use cases.

Flash storage is still a lot slower than RAM and also has limited write lifetime

The article is pretty light on the "why", in fact it doesn't approach the question at all, and it's wrong on the basics. Unified memory is hardly new, the competing Intel mobile architecture with IGPs has had unified memory for 17 years.

Agreed.

The short answer is Apple's using a SoC with LPDDR4x (thanks jeffbee) inside the package, and they are already using the highest density memory chips (available when the M1 started shipping). This helps save power, increase bandwidth, decreases latency, and allows for more channels (2 channels per DDR5 chip).

Not sure if Apple's going to just switch to a higher density chip (when they become available), add more memory channels, or maybe just make all the higher end iMac, mac pro, and similar machines dual socket which would double the max ram. Maybe even both, double the density and double the sockets. Clearly the heat/power/space for the M1 is substantially lower than the intel chips they replaced.

LPDDR4X I think, but yeah. The reason the M1 comes with 8 or 16GB is the same reason that almost every Intel Tiger Lake laptop comes with 8 or 16: those are the practical sizes.

Ack, sorry, you are exactly right, @ 4266 MT/sec

>Unified memory is hardly new

Yes..... but

>the competing Intel mobile architecture with IGPs has had unified memory for 17 years.

Intel CPU and GPU are using Unified Memory in literal sense but not unified memory in technical terms or how Apple uses it. Or to be more precise, Intel CPU and GPU on Windows does not share the same memory address space. They are partitioned in the memory. Which is not Unified Memory really meant. But I am sure the way things are going this is going to spin out of hand and the term will be redefined. ( Hopefully Not )

Reply on news.ycombinator.com