Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

macduke

macrumors G5
Original poster
Jun 27, 2007
13,495
20,609
I was looking at the system requirements for Topaz AI because I haven't upgraded to the newer versions and was curious if they have a build now for Apple Silicon as I have a 14" M3 Max 16/40 64GB on order. I'm looking to upgrade some of my professional tools to avoid the performance impact of Rosetta 2 wherever possible.

Given the discussion recently in the forums about Apple saying 8GB = 16GB, I found the RAM system requirements for Topaz Photo AI to be interesting.

System requirements for M-series chips: 8GB

Screenshot 2023-11-15 at 5.19.42 PM.png



System requirements for Intel Mac: 16GB

Screenshot 2023-11-15 at 5.19.38 PM.png


System requirements for Windows: 12GB (16GB or more recommended)

Screenshot 2023-11-15 at 5.20.25 PM.png


So it seems like there is some real world evidence for this argument, although I feel like 12GB is probably closer to the real answer, and I thought as much when that story came out. Memory compression goes a long way, and having unified memory can really help out apps like this that can be so GPU intensive doing AI stuff.

While I wouldn't recommend 8GB even to my grandfather, it's still interesting that there is something real and tangible to this whole story, as evidenced above. 8GB still isn't enough for a professional user if they ever plan to do more than one thing at a time. And while it will work, evidence has shown that more memory helps, especially in AI applications for loading the model into memory, not to mention other applications like video rendering, etc. For normal users, it will significantly shorten the usable lifespan of their computer with regard to future OS updates. You can't tell me this 8GB system won't be crawling when it gets macOS 20 in six years! And a lot of regular people keep their computer that long or longer. Heck, I'm a professional and I kept my maxed out 2012 retina MBP for nearly 7 years. Didn't keep my 2019 that long because of the transition and massive improvements to heat and battery life, but I plan to keep my upcoming MBP M3 Max for about that long!

Anyway, just thought that was interesting. Peace.
 
A lot of folks think Apple are just being cheap, and I don’t disagree, it’s not like it costs that much more in parts costs to equip these machines with 16GB instead of 8GB, especially given what they charge for even the cheapest of their machines.

But there is technical merit to the “more efficient” claim for two reasons. One is ARM processors run a RISC instruction set instead of CISC. The number of functions the CPU can perform are reduced but the ones that are present are considerably faster.

And the second is code bloat. Not everyone here will remember this but when Apple switched to Intel they carried over the 512MB of memory base model machines came with, and it simply wasn’t enough for Intel versions of OSX. I had both an iBook G4 and a brand new Core Duo Mac Mini at the time and the Mini suffered from slowdowns and beach balls a lot more frequently than my trusty Book did.

In order to save multiple gigabytes of hard drive space there was a utility around at the time of OSX Leopard that would strip the Intel code from binaries if you had a PowerPC machine, which made them up to 2/3rd smaller in my experience. PPC was also a RISC based design, so the code was smaller.

Long story short Apple had to take a much less optimized approach to coding OSX during the Intel phase to accommodate CISC execution. Moving to their own silicon meant they could once again highly optimize the code base because, like in the AIM days, the CPUs are made only for the Macintosh. They never had that with Intel, they just had to take whatever Intel had on the shelf and make do.

I have an 8GB M2 Air and I regularly push it to play games, record and master audio, and occasionally archive and encode video. Haven’t yet run into any memory issues.
 
Last edited:
These requirements prove very little. If another software company puts equal requirements for both then you'll say that's proof that they are equal?
Anyway from my daily usage of M1 Mac, Intel Mac and Intel PC I can say that the RAM usage difference is not significantly different. That has nothing to do with speed.
I would say that indeed M1 is a bit more efficient than Intel in RAM usage with the same apps open, but by a couple hundreds MB to 1GB at best, depending on the apps you use, definitely not 8GB.
 
  • Like
Reactions: Chuckeee
One is ARM processors run a RISC instruction set instead of CISC. The number of functions the CPU can perform are reduced but the ones that are present are considerably faster.
...but, crudely speaking, you may need several RISC instructions to achieve the same thing as a single CISC instruction. I don't think you can really generalise about the size of the resulting binaries.

Anyhow, RAM size and processor speed are pretty independent - if anything a faster processor (and, possibly, one with more cores) is going to be able to shift more data and will benefit from having more data available in RAM.

In order to save multiple gigabytes of hard drive space there was a utility around at the time of OSX Leopard that would strip the Intel code from binaries if you had a PowerPC machine, which made them up to 2/3rd smaller in my experience.
...it's not clear if that changed the RAM requirements or just saved disc space (I doubt that the Intel code ever got loaded on a PPC machine, or vice versa). Anyhow, we're back in the same boat now as many apps are again shipping as "universal binaries" containing both ARM64 and x86-64 code.

Anyway, in an age when we're slinging around 220ppi high-DPI images, 128k sound samples, 4k video, frame buffers for 5k screens, huge AI training data sets etc. storing the executable code is only one aspect of memory usage.

There's also the question pf what you mean by "enough RAM". Mostly the effect of running low on RAM will be that things slow down. That's offset by Apple Silicon Macs being all-round faster than Intel Macs - dramatically so if the media/neural engines come into play. Even if you run out of RAM and start swapping to SSDs, Apple Silicon Macs have super-fast SSDs so that might not be the catastrophe that it would have been on Intel...

BUT- even the fast SSD on AS Macs is still an order of magnitude slower than RAM, so if you run low on RAM your processor is being throttled by lack of data - so the question is, are you happy with your shiny new Mac being faster than your old Intel Mac, or do you want it to run to its full potential?

Apple Silicon is almost certainly more efficient in its RAM usage but "8GB=16GB" as a general claim is just nonsense. If your application needs 10GB of data in memory on Intel it will need 10GB of data in memory on Apple Silicon. Also, for most pro apps the amount of memory needed depends hugely on what you are using it for, so while it is quite plausible that a particular app which needed 16GB of RAM to even load and run the demo on Intel will get by with 8GB on Apple Silicon that doesn't mean that your workload won't be throttled by lack of RAM.

The problem is that Apple are too tight with their base RAM specs and too greedy with their upgrade prices: 16GB should be a no-brainer for doing anything beyond "perdonal productivity" and an 8 to 16GB upgrade shouldn't cost anything like $200. You just have to decide whether you're prepared to pay the Apple tax.
 
If that software company puts 16GB RAM as requirement on Mac, he is basically killing most of his Mac audience. So it is in his best interest to list 8GB RAM as a minimum requirement on Mac.

To proof or disprove Apple is claim is super easy. Take a 16GB dataset and load it in the memory on:
1. MacOS
2. PC + Windows
3. PC + Linux based OS
4. PC + Unix based OS

And then check if the RAM usage on Mac OS is indeed half of a Windows, Linux and Unix based PC's.
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.