8 GB of Unified Memory
Starting at $1,599, the 14-inch M3 MacBook Pro comes with 8GB of unified memory. That makes it $300 more expensive than the $1,299 starting price of the now-discontinued M2 13-inch MacBook Pro with 8GB. Users can opt for 16GB or 24GB at checkout, but these configuration options cost an extra $200 and $400 at purchase, respectively, and cannot be upgraded at a later date because of Apple’s unified memory architecture.
[…]
In a recent interview with Chinese ML engineer and content creator Lin YilYi, Apple’s VP of worldwide product marketing Bob Borchers has directly responded to this criticism.
Comparing our memory to other system’s memory actually isn’t equivalent, because of the fact that we have such an efficient use of memory, and we use memory compression, and we have a unified memory architecture.
Actually, 8GB on an M3 MacBook Pro is probably analogous to 16GB on other systems. We just happen to be able to use it much more efficiently. And so what I would say is I would have people come in and try what they want to do on their systems, and they will I think see incredible performance.
You could make the case that the performance of certain tasks with 8 GB of RAM is good, but his statement goes way beyond that, and I don’t think the reasoning is sound.
It’s not clear to me what “efficient” is meant to refer to here. One could argue that macOS is less efficient with memory since it no longer runs in 32-bit mode. Maybe it’s a reference to Dynamic Caching, but that doesn’t seem like it would have much effect on memory use for common apps.
Memory compression has been available in macOS for a long time. I’ve used 8 GB and 16 GB systems with and without it and have no doubt that real RAM is better. Citing memory compression also doesn’t make sense because it’s a constant. The old M2 MacBook Pro also had memory compression. So does Windows.
The unified memory architecture does not seem like a technology to help stretch RAM. Rather, it means that some of that 8 GB will be used like VRAM and not available to apps or to the system.
William Gallagher (Hacker News):
A core claim of Apple’s is that this improved design means Mac need less RAM than they did.
The 8 GB M3 MacBook Pro is more expensive than the 16 GB M2 MacBook Pro was, and I’m not aware of any change in the M3 that would make up for that.
The thing we should be mad about are the prices. They’re charging $200 or more for each step-up in RAM. I understand that their RAM is integrated and special, but an 8GB stick of SD RAM for a PC is like $30. $100 might be understandable, but $200 is obscene.
The storage is even worse. Even the M3 MAX defaults to 1TB of storage. To upgrade to 4TB is $1000. A Samsung 990 Pro M2 SSD with 4TB of storage is under $300. I understand the apple storage is different, and that justifies some markup, but over a 300% markup is absurd.
It should probably not be a controversial opinion that, in late 2023 (and surely through most of 2024), one should not sell a pricey “Pro” computer with only 8GB of RAM. And yet here we are.
[…]
Apple has a long history of providing less RAM than it should for the price of its laptops and overcharging to get more, but it’s reached ridiculous proportions. The cheapest standard configuration with more than 8GB of RAM is 2 grand! The cheapest MacBook Pro you can configure with more than 8GB is $1,800!
[…]
Not that Windows laptops and Macs are directly comparable, but comparably-priced Microsoft Surface, HP Envy, Alienware, Dell XP, and Lenovo Thinkpad laptops all have 16GB of RAM or more, standard. You can spend all day mired in laptop configurations (and I have) but the bottom line is this: 16GB is standard at prices over $1,000 even in laptops with premium displays and other high-end features.
Perhaps unsurprisingly, Yuryev saw significant performance improvements across the board using the 16GB machine under both middling and heavier workloads. The 8GB model suffered double-digit losses in Cinebench benchmarks, and took several minutes longer to complete photo-merging jobs in Photoshop as well as media exports in Final Cut and Adobe Lightroom Classic.
These tests were conducted as single operations with nothing else running, but also repeated with browser tabs, YouTube videos, spreadsheets, emails, and the like, open in the background to simulate typical real-world multi-tasking scenarios. As expected, the performance gap between the two machines widened further as the 8GB increasingly relied on its SSD swap file, while all-round responsiveness took a hit. Yuryev even reported crashes on the 8GB model during Blender rendering and a Final Cut export.
How about comparing the 8 GB M3 to the 16 GB M2?
Previously: