So Confused: Core Chips versus Core GPU versus RAM versus ????

[CameraCarl]

I need to replace my 2017 5K 27 inch Retina iMac because it is getting too slow to cope with all the processing needed to do generative AI on Lightroom Classic. I've been starting to shop for a new MacBook Pro because it doesn't appear that Apple will be making anything other than a 24in iMac any time soon.

But I'm so confused. There are M2s and M3s; pros and not-pros; 8 core chips, 11 core chips, 12 core chips; then 10, 14, 18 and 19 core GPUs; and then there is RAM in every size; etc. etc. I am retired and can't afford to spend a lot of money on a computer, considering that I will need also to buy a high resolution monitor.

So, if I want to future proof my purchase as much as possible, and want to be able to process 45 megapixel images and use all the AI features in Lightroom Classic without wanting to wait many minutes for the images to process, what computer features/specifications are critical that I should shop for?

For example, would a 14 inch M2 Pro, 12 core chip, 16GB RAM, 19 core GPU be fast enough? Or should I opt for 24GB of RAM but scale back to a 8 core chip with 10 core GPU for about the same money?

 

[Abbott Schindler]

First, do you need a portable? Since you're using an iMac now, I'd seriously look at a Studio Max or Ultra (I have a M2 Ultra) and an external display. Your current keyboard, mouse, trackpad, Intuos, etc. will all work with a Studio or Mini. This future proofs you in that rather than "upgrading" (i.e. "replacing") the computer + monitor, you can replace either component as needed/desired. I just saw a 27" BenQ MA270U monitor announced that look like a really good match for either Mac (I have a pair of 27" NEC monitors connected to my Studio).

I went with a Studio for its faster ports (all TB4 vs TB4 & USB3.2 on the Mini) and built-in fan.

Given that no current Macs are storage/memory upgradeable, I'd suggest going with the most you can afford. Benchmarks are available for all machines. My experience matches what I read: I have a 16" M2 MacBook Pro Max and a M2 Studio Ultra. Benchmarks suggested that although the Studio has twice the compute & memory components as the MBP, it would have "only" about 1.8-19.X the MBP performance...and this is what I see.

Unless you're doing video editing, I think you'd be fine with 64GB memory in either machine (I have 32 in the MBP, 64 in the Studio and never come near running out of memory with my 45 mpxl EOS R5 Raw files).

The amount of storage to configure depends on you. For a MBP you might want 2-4 TB if you want to store a lot of data on the machine. I have 1 TB and keep my multi-TB photo library on an external TB4 SSD, and backup to a USB3 SSD. This allows me to move my library among the 2 machines for editing or whatever.

One thing I like about the Studio/Mini idea is that it's easy to store ALL of your data on external drives. That's important for me, as IF the machine dies or needs service, I don't need to send my data in with it. That's not as easy to do with a MBP, as external storage isn't quite as convenient.

You'll probably also want to get a TB4 dock, btw. I've got a Satechi; CalDigit is also very well regarded.

Good luck with your decision.

 

[Greybeard2023]

For example, would a 14 inch M2 Pro, 12 core chip, 16GB RAM, 19 core GPU be fast enough?

Thats exactly what I have - certainly no problems or delays with 40MP images

 

[MarshallG]

Just to be clear: Both memory (RAM) and internal storage are non-upgradeable in all Mac models sold today. While you can get very fast Thunderbolt SSD’s, they aren’t quite as fast as the built-in storage. In my workflow, I dump my memory card to the internal storage, do my RAW editing, and then move the folder to my external SSD. I can edit from the external drive just fine, but this method is a little faster for browsing the thumbnails.

I have an M1 MacBook Pro with 16 GB of RAM and 512 GB of storage, and zero complaints.

I use DxO PhotoLab and I haven’t tried generative AI with Lightroom, so I don’t know what that would take.

There’s a guy in this forum who absolutely insists on having over 10,000 RAW images in a folder, so he absolutely needs 32 GB of RAM and an ultra-fast SSD, because heaven above forbid he create a folder and break things down into 500 or so images per folder. My point is, no, you don’t need to be a genius to figure out a “need” for extreme CPU and Memory requirements, but whether you truly need them is another story. I tried searching Adobe for their system requirements for AI in Lightroom, and they haven’t published any!

It’s safe to say that a system with 16 GB of RAM and any M2 will work just fine, but if you’re doing professional work for hours and you want everything as fast as possible, perhaps using a desktop box with 32 GB of RAM will make you happier.

 

[Thom Hogan]

But I'm so confused.

Generations: We currently have three generations of Apple Silicon available on Macs: M1 (the original), M2, and M3. Late this fall we'll get M4. Apple has been using the reduction of semiconductor process size to drive each new generation. In general, they've been balancing performance versus power usage.

Types: We have the base silicon (say M3, the baby bear). We also have expanded versions (larger semiconductors with more transistors) that comprise M3 Pro (Goldilocks) and M3 Max (daddy bear). With more territory you can fit in more RAM and more cores.

CPU Cores: The main ARM-based computation is done either by a Performance core (speed) or an Efficiency core (power conservation). Some tasks you need performed quickly should be done by a Performance core (macOS and apps parcel this out as needed). Otherwise, tasks on a laptop that don't need a lot of computation (e.g text editing) should be using Efficiency cores. The number of CPU cores Apple lists for a chip are split, often somewhere near half and half between types.

GPU Cores: Graphic Processing Units evolved as special purpose processors, typically ones that, duh, deal with graphics.

NPU Cores: Neural Processing Units evolved to handle specific needs of AI, ML, and other similar tasks. They aren't there to crunch numbers, they're there to do brain-like things with data (e.g. neural network).

The good news is that all of the Apple Silicon has a handful (or more) of each type of core. The more cores it has for a job, the faster it will process the job (assuming the software engineers were paying attention). For instance, video compression is usually—but not always, since Apple Silicon now also has dedicated compression cores!—handled by GPUs. The more of those you have the faster the compression is done.

All that said, much of the time your computer is basically just using a Performance CPU core for most of what you're doing. Moreover, even the simplest Apple Silicon chip has more than enough "power" to handle most things most users do. If you're going to do regressive analysis on a big data set, compress or decompress 8K video, or some other "big" task, well, then should be considering one of the higher end chips.

Along with cores of various sorts, Apple also puts all memory used by the computer on the same Apple Silicon chip, which is why you have to be careful about how much memory you buy on day one: you cannot increase it.

The reason Apple did this is performance (and efficiency). By having all memory on the single processing chip all those cores and other things built into the chip have direct and extraordinarily fast access to the memory.

That said, pretty much no matter how much memory you buy, you're going to eventually run a task that needs more. No worries. Apple has basically taken the approach my team did with a product called RAM Doubler: (1) free up memory blocks that aren't being used at all; (2) compress blocks that haven't been accessed in awhile; and (3) if you still need more "memory" the system swaps out to the SSD. All this is why I say buy memory before cores, make sure you buy plenty of SSD (typically at least 1TB).

So, if I want to future proof my purchase as much as possible,

This is the part where you'll find many of us Mac experts disagreeing. We're right now in a very fast iteration cycle that's producing strong gains each iteration. With Apple adding things like Apple Intelligence and who knows what next, there's a pretty strong feeling among some that trying to buy a 10-year futureproof machine these days is a bit of fools folly. Others think there's no real issue coming. My advice has been to not get too deep into cores, make sure you have a lot of memory, and don't skimp on an SSD. Those latter two things make a Mac more futureproof at the moment.

and want to be able to process 45 megapixel images and use all the AI features in Lightroom Classic without wanting to wait many minutes for the images to process, what computer features/specifications are critical that I should shop for?

That's a trickier question than you think, as Adobe tends to be pretty inefficient in initial offerings and it's only over time that they figure out how to make their features into more hot rods than sluggish rods. However, that said, on an M1Max 64GB MacBook 14" I've never felt like I'm waiting for Adobe. But even on my M1 16GB MacBook Air I don't really see any real issues with speed, as long as the images I'm working on are on the internal SSD (which is far faster than any external SSD at the moment).

For example, would a 14 inch M2 Pro, 12 core chip, 16GB RAM, 19 core GPU be fast enough?

I'd say yes. But make sure you're not skimping on the SSD.

Or should I opt for 24GB of RAM but scale back to a 8 core chip with 10 core GPU for about the same money?

For a photographer doing image processing: memory before cores.

 

[CameraCarl]

Thanks for all the detailed answers. Thom, you have given me a lot to think about -- and absorb. I could understand it so much easier in the olden days where RAM was king and everything else (except storage, perhaps) was secondary.

 

[Philip Eihuyar]

Thanks for all the detailed answers. Thom, you have given me a lot to think about -- and absorb. I could understand it so much easier in the olden days where RAM was king and everything else (except storage, perhaps) was secondary.

Thom pretty much covered it however, I thought to include a real world example regarding memory use - My own.

First, I checked on price difference between 18GB and 36GB RAM. $400.

Currently, I have open 4 browsers, Affinity publisher, Affinity Photo, Word and not much else. Already though, my memory use is 22GB RAM use. I have 32 GB RAM in my intel 2019 MBP so there is no memory pressure or swapping. If this sound like your typical use, it might be something to think about and choose for the extra $400 the 36GB option.

 

[Mike Boreham]

Thanks for all the detailed answers. Thom, you have given me a lot to think about -- and absorb. I could understand it so much easier in the olden days where RAM was king and everything else (except storage, perhaps) was secondary.

Thom pretty much covered it however, I thought to include a real world example regarding memory use - My own.

First, I checked on price difference between 18GB and 36GB RAM. $400.

Currently, I have open 4 browsers, Affinity publisher, Affinity Photo, Word and not much else. Already though, my memory use is 22GB RAM use. I have 32 GB RAM in my intel 2019 MBP so there is no memory pressure or swapping. If this sound like your typical use, it might be something to think about and choose for the extra $400 the 36GB option.

And what was the Memory Pressure ? This is what matters. Macs will use memory they find available so someone with 96GB RAM might see 60GB used which doesn't mean everyone needs more than 60GB.

 

[MarshallG]

But I'm so confused.

Generations: We currently have three generations of Apple Silicon available on Macs: M1 (the original), M2, and M3. Late this fall we'll get M4. Apple has been using the reduction of semiconductor process size to drive each new generation. In general, they've been balancing performance versus power usage.

Types: We have the base silicon (say M3, the baby bear). We also have expanded versions (larger semiconductors with more transistors) that comprise M3 Pro (Goldilocks) and M3 Max (daddy bear). With more territory you can fit in more RAM and more cores.

CPU Cores: The main ARM-based computation is done either by a Performance core (speed) or an Efficiency core (power conservation). Some tasks you need performed quickly should be done by a Performance core (macOS and apps parcel this out as needed). Otherwise, tasks on a laptop that don't need a lot of computation (e.g text editing) should be using Efficiency cores. The number of CPU cores Apple lists for a chip are split, often somewhere near half and half between types.

GPU Cores: Graphic Processing Units evolved as special purpose processors, typically ones that, duh, deal with graphics.

NPU Cores: Neural Processing Units evolved to handle specific needs of AI, ML, and other similar tasks. They aren't there to crunch numbers, they're there to do brain-like things with data (e.g. neural network).

The good news is that all of the Apple Silicon has a handful (or more) of each type of core. The more cores it has for a job, the faster it will process the job (assuming the software engineers were paying attention). For instance, video compression is usually—but not always, since Apple Silicon now also has dedicated compression cores!—handled by GPUs. The more of those you have the faster the compression is done.

All that said, much of the time your computer is basically just using a Performance CPU core for most of what you're doing. Moreover, even the simplest Apple Silicon chip has more than enough "power" to handle most things most users do. If you're going to do regressive analysis on a big data set, compress or decompress 8K video, or some other "big" task, well, then should be considering one of the higher end chips.

Along with cores of various sorts, Apple also puts all memory used by the computer on the same Apple Silicon chip, which is why you have to be careful about how much memory you buy on day one: you cannot increase it.

The reason Apple did this is performance (and efficiency). By having all memory on the single processing chip all those cores and other things built into the chip have direct and extraordinarily fast access to the memory.

That said, pretty much no matter how much memory you buy, you're going to eventually run a task that needs more. No worries. Apple has basically taken the approach my team did with a product called RAM Doubler: (1) free up memory blocks that aren't being used at all; (2) compress blocks that haven't been accessed in awhile; and (3) if you still need more "memory" the system swaps out to the SSD. All this is why I say buy memory before cores, make sure you buy plenty of SSD (typically at least 1TB).

So, if I want to future proof my purchase as much as possible,

This is the part where you'll find many of us Mac experts disagreeing. We're right now in a very fast iteration cycle that's producing strong gains each iteration. With Apple adding things like Apple Intelligence and who knows what next, there's a pretty strong feeling among some that trying to buy a 10-year futureproof machine these days is a bit of fools folly. Others think there's no real issue coming. My advice has been to not get too deep into cores, make sure you have a lot of memory, and don't skimp on an SSD. Those latter two things make a Mac more futureproof at the moment.

and want to be able to process 45 megapixel images and use all the AI features in Lightroom Classic without wanting to wait many minutes for the images to process, what computer features/specifications are critical that I should shop for?

That's a trickier question than you think, as Adobe tends to be pretty inefficient in initial offerings and it's only over time that they figure out how to make their features into more hot rods than sluggish rods. However, that said, on an M1Max 64GB MacBook 14" I've never felt like I'm waiting for Adobe. But even on my M1 16GB MacBook Air I don't really see any real issues with speed, as long as the images I'm working on are on the internal SSD (which is far faster than any external SSD at the moment).

For example, would a 14 inch M2 Pro, 12 core chip, 16GB RAM, 19 core GPU be fast enough?

I'd say yes. But make sure you're not skimping on the SSD.

Or should I opt for 24GB of RAM but scale back to a 8 core chip with 10 core GPU for about the same money?

For a photographer doing image processing: memory before cores.

 

[Philip Eihuyar]

Thanks for all the detailed answers. Thom, you have given me a lot to think about -- and absorb. I could understand it so much easier in the olden days where RAM was king and everything else (except storage, perhaps) was secondary.

Thom pretty much covered it however, I thought to include a real world example regarding memory use - My own.

First, I checked on price difference between 18GB and 36GB RAM. $400.

Currently, I have open 4 browsers, Affinity publisher, Affinity Photo, Word and not much else. Already though, my memory use is 22GB RAM use. I have 32 GB RAM in my intel 2019 MBP so there is no memory pressure or swapping. If this sound like your typical use, it might be something to think about and choose for the extra $400 the 36GB option.

And what was the Memory Pressure ? This is what matters. Macs will use memory they find available so someone with 96GB RAM might see 60GB used which doesn't mean everyone needs more than 60GB.

I'm not sure how my intel mac uses memory and the 'M' chip memory is a mystery to me.I I assume my recommendation probably holds.

 

[more cow bell]

I likely know a lot less about the “how much memory is needed/best for LRc on a Mac with Apple silicon” question than the folks who have already chimed in, so I won’t attempt to address that, but I will share my experience which will hopefully be helpful to you.

Late last year I switched from a 27 inch 2017 5K iMac to a 14-inch MBP last year and I’m thrilled. My iMac was well configured (4.2 GHz i7, 40 GB RAM, Radeon Pro 580 GPU and 512 GB internal SSD with a 4 TB external hard drive for raw file storage) so performance was usually not a big problem (though it’s fans got very loud when importing a lot of raws into LRc or working on a photo with multiple masks). I also had a super lightweight but severely underpowered 2015 MacBook (very different animal than a modern MacBook Pro), which was ok for email and light web surfing, but effectively useless for LRc. My main motivation for switching was to simplify my life and have just one machine that had the power of my desktop with the mobility of a laptop. To simplify even further, I decided to see if I could also afford a machine with enough internal storage so I wouldn’t need an external storage device for my 1.5 TB (and growing) collection of photo, video and music files.

My recollection is that when I was looking, an M3 based 14-inch MBP with 4 TB of internal storage and 48 or more GB of RAM was $5k+, which was more than I was willing to spend. So, I started looking at Apple’s refurbished website, and saw they had a 14-inch MBP with M1 Max CPU, 64 GB of RAM and a 4 TB SSD going for around $3,300, still not cheap but much closer to what I was willing to spend. Out of curiosity I decided to also check OWC (i.e. macsales.com) for used units, and I stumbled across an Apple factory refurbished MBP with an M1 Max CPU, 64 GB of RAM and 4 TB of storage for $2,900, which struck me as a good deal so I pounced. I’ve gone back to macsales.com a few times since, and sometimes they have Apple factory refurbs but most times they don’t, so perhaps my timing was lucky.

I’m certain that MBPs with M2 Max or Max M3 CPUs perform better than my M1 Max machine, but I have not been disappointed at all. I didn’t do a whole lot of formal benchmarking before selling my iMac, but anecdotally, everything on the MBP is at least as fast as my iMac and many things are noticeably faster. For example I don’t get the sluggish UI issues I used to see on the iMac when editing a photo with multiple masks. A big plus is the fans rarely kick in, and when they do they are not nearly as loud as the iMac. I haven’t used LRc’s Generative AI that much, but the handful of tests I just did took 5-7 seconds.

A few other random pros to MBP vs. iMac:

• Because I can use the MBP while sitting on the couch “watching” TV I’m now culling old photos that have been clogging my LRc catalog for years.
• Because it has enough power and is portable, I can start to cull and process photos while traveling, instead of waiting until I get home to my desktop.
• The screens on Apple silicon based 14 and 16 inch MBPs are HDR capable, while my iMac screen was not. I didn’t fully grasp what this meant until I saw how great videos and photos can look on the MBP screen. The HDR topic is its own complex kettle of fish, but if you’re interested this thread has a lot of background info https://www.dpreview.com/forums/post/67384702

Best of luck with your decision!

 

[Zeee]

I'm not an expert but when I was asking about a new Mac I was told get the fastest processor you can afford. So the cores, etc would be better answered by the people who are more knowledgable.

Also I learned the hard way that AI eats VRAM. I got 8 GB VRAM on my 2019 iMac intel and I'm glad I did as I can still use it today and it works quite well.

The new M chip does not have separate VRAM. It is unified memory so I would get as much RAM as I could.

Just for an example my 27" 2019 iMac Intel with 64GB RAM and 8GB VRAM takes 35 seconds to run Adobe Denoise AI. My lowly 2021 MacBook Air M1 with 16GB RAM takes 25 seconds. You are coming in at the right time. My Air used to take 85 seconds but Apple fixed the bug that prevented the use of the Neural Engine cores. So at the release of LrC 13.3 (I think) this has been corrected. This goes for ACR as well.

I also get minimum storage. My iMac is 500GB SSD and the Air is 250GB as it is for travel. External drives are cheap so I store my files there. The money I save on the SSD goes toward processing power and RAM.

 

[graybalanced]

I kinda don't want to say this because a lot of the posters in this thread are actually very knowledgable, but a lot of the advice posted is a little off the mark and over complicated.

For the purpose you said, Lightroom Classic with 45 megapixels and wanting to use AI:

For a laptop, start at the M3 Pro with 12 CPU cores and 18 GPU cores. (The lowest M3 Pro lacks a second fan and has other compromises. ) The base 18GB memory is OK, but if you can afford it, move that one level up to 36GB. Any other upgrades from there are only if they don't blow your budget. I would have lowered the specs if the images were less than 45 megapixels.

For a desktop, start with the base Mac Studio. It comes with lots of CPU and GPU cores, more than the laptops. And it comes with 32GB already.

The rest of this is if you want the reasoning behind that recommendation.

But I'm so confused. There are M2s and M3s; pros and not-pros; 8 core chips, 11 core chips, 12 core chips; then 10, 14, 18 and 19 core GPUs; and then there is RAM in every size; etc. etc.

A very important thing is how the chip levels are stacked. The single core performance of each generation is about the same. Any speed increase comes from more cores. But that only happens with software that can take advantage of more cores. You say you use Lightroom Classic, the good thing is it can use more CPU cores. (Less true with Photoshop.)

Example: In single core performance, M1/M1 Pro/M1 Max/M1 Ultra are about the same speed. To get more single-core performance you go up to M2 or M3. To get more multi-core performance on any model, you buy more CPU cores. The Pro/Max/Ultra levels are mostly about adding on more CPU and GPU cores.

OK now about Pro and not-Pro: The not-Pros are the MacBook Air. They don't have a fan and are less capable overall. The Airs are OK for Lightroom occasionally but not if you are working all day because then the heat builds up so you need the fans to maintain top speed all day.

If an app uses GPU acceleration then you must have enough Unified Memory for optimum GPU performance, and that is on top of the memory for system and apps. On old Intel Macs you might say, I need 16GB for OS and apps, and an 8GB graphics card. On Apple Silicon Macs, that means you need that same total amount in one single unified pool, so you say you need 24GB or 32GB Unified Memory.

Absolutely critical is the detail you mentioned about Lightroom generative AI on 45MP images. Gen AI is mostly GPU, so absolutely do not cheap out on the GPU cores. This is even more critical if you want to use AI Denoise often, that feature is almost all GPU (with help from the Apple Neural Engine). Doubling the GPU cores will halve AI Denoise processing time according to many tests and reports. Other GPU-accelerated features will not necessarily see that degree of improvement per core.

Not shorting the GPU means don't choose an Air, only Pro models because they can have many more GPU cores.

For your use case, 45MP images, a reasonable build would be M2 Pro/Max or M3 Pro/Max with 36GB Unified Memory. (16-18GB is getting to be iffy for supplying OS, apps, and GPU with enough memory for advanced processing.)

With those models, start at a lower number of CPU and GPU cores, and only pay for more cores if you can reasonably afford them. If you add more CPU cores, in Lightroom Classic you speed up previews and general use. If you add more GPU cores, you speed up Develop, bulk export, generative AI, and AI denoise.

 

[Thom Hogan]

For a photographer doing image processing: memory before cores.

I disagree Thom, because Apple has been charging exorbitant amounts for storage.

How much Apple charges isn't relevant to performance.

For example, you can buy a MacBook Pro with the M3 Pro, 18 GB RAM and 512 GB of storage at Costco for $1,500. Your swap file will definitely not fill with 512 GB of storage.

As an SSD fills and as things like temporary cache files get stacked up, performance can and will drop faster than you think. The macOS itself has a threshold at about 90% of storage used where you absolutely will start seeing performance hits.

Apple Thunderbolt is extremely fast — 40 gbps — and they give you three independent 40 gbps connections.

Still not as fast as the internal SSDs, and now you have another set of things in the pipeline and can reduce performance. Hook up an external monitor, for instance, and you're taking lanes out of play.

You can easily purchase 8 TB of Thunderbolt SSD storage for less than the price of upgrading a MacBook Pro to 1 TB, and that will give you enough storage for a very large photo library.

I long ago learned that "cheaping" out in the short term tends to come back and bite you in the long term. Moreover, the OP has posited that they want longevity for this system.

The real question that I keep having to answer is whether most users can see the performance decline by under buying. To that I'd say no, they can't. That's partly because many are moving from Intel to Apple Silicon and there's some clear performance benefits just in that. That doesn't mean they're not there and they won't eventually see them, though.

From time to time I'll actually run real tests between my base MacBook Air and my juiced up MacBook Pro and see what I can measure. I've noticed that with a couple of programs recently that their updates made the Pro run faster than the Air than it used to. I looked more closely at one of those and discovered it was all about temporary caches. There's a huge advantage to those being generated on chip as opposed to off chip.

 

[MarshallG]

For a photographer doing image processing: memory before cores.

I disagree Thom, because Apple has been charging exorbitant amounts for storage.

How much Apple charges isn't relevant to performance.

For example, you can buy a MacBook Pro with the M3 Pro, 18 GB RAM and 512 GB of storage at Costco for $1,500. Your swap file will definitely not fill with 512 GB of storage.

As an SSD fills and as things like temporary cache files get stacked up, performance can and will drop faster than you think. The macOS itself has a threshold at about 90% of storage used where you absolutely will start seeing performance hits.

Apple Thunderbolt is extremely fast — 40 gbps — and they give you three independent 40 gbps connections.

Still not as fast as the internal SSDs, and now you have another set of things in the pipeline and can reduce performance. Hook up an external monitor, for instance, and you're taking lanes out of play.

You can easily purchase 8 TB of Thunderbolt SSD storage for less than the price of upgrading a MacBook Pro to 1 TB, and that will give you enough storage for a very large photo library.

I long ago learned that "cheaping" out in the short term tends to come back and bite you in the long term. Moreover, the OP has posited that they want longevity for this system.

The real question that I keep having to answer is whether most users can see the performance decline by under buying. To that I'd say no, they can't. That's partly because many are moving from Intel to Apple Silicon and there's some clear performance benefits just in that. That doesn't mean they're not there and they won't eventually see them, though.

From time to time I'll actually run real tests between my base MacBook Air and my juiced up MacBook Pro and see what I can measure. I've noticed that with a couple of programs recently that their updates made the Pro run faster than the Air than it used to. I looked more closely at one of those and discovered it was all about temporary caches. There's a huge advantage to those being generated on chip as opposed to off chip.

 

[Greg Edwards]

But I'm so confused.

Generations: We currently have three generations of Apple Silicon available on Macs: M1 (the original), M2, and M3. Late this fall we'll get M4. Apple has been using the reduction of semiconductor process size to drive each new generation. In general, they've been balancing performance versus power usage.

Apologies to the OP for massively derailing this thread, but this comment make me wonder how long until the fundamental limits of physics are met.

M1 and M2 are 5nm, whereas the M3 and M4 are 3nm (IIRC), with 2nm processor technology expected next year (that I'm guessing is good for the M5 and M6). Then maybe 1.5nm or the theoretical limit of 1nm .

With Moore's law currently running on a 3-year cycle, it could be dead in the next 10 years.

What comes next? Could an M7 or M8 be the last computer I'll ever buy?

 

[dv9973]

But I'm so confused.

Generations: We currently have three generations of Apple Silicon available on Macs: M1 (the original), M2, and M3. Late this fall we'll get M4. Apple has been using the reduction of semiconductor process size to drive each new generation. In general, they've been balancing performance versus power usage.

Apologies to the OP for massively derailing this thread, but this comment make me wonder how long until the fundamental limits of physics are met.

M1 and M2 are 5nm, whereas the M3 and M4 are 3nm (IIRC), with 2nm processor technology expected next year (that I'm guessing is good for the M5 and M6). Then maybe 1.5nm or the theoretical limit of 1nm .

With Moore's law currently running on a 3-year cycle, it could be dead in the next 10 years.

What comes next? Could an M7 or M8 be the last computer I'll ever buy?

Quantum computing

 

[MarshallG]

But I'm so confused.

Generations: We currently have three generations of Apple Silicon available on Macs: M1 (the original), M2, and M3. Late this fall we'll get M4. Apple has been using the reduction of semiconductor process size to drive each new generation. In general, they've been balancing performance versus power usage.

Apologies to the OP for massively derailing this thread, but this comment make me wonder how long until the fundamental limits of physics are met.

M1 and M2 are 5nm, whereas the M3 and M4 are 3nm (IIRC), with 2nm processor technology expected next year (that I'm guessing is good for the M5 and M6). Then maybe 1.5nm or the theoretical limit of 1nm .

With Moore's law currently running on a 3-year cycle, it could be dead in the next 10 years.

What comes next? Could an M7 or M8 be the last computer I'll ever buy?

Quantum computing

No, quantum computers cannot replace conventional CMOS CPUs, because quantum computers can operate at temperatures close to absolute zero. You will not ever have a quantum computer laptop or cell phone.

 

[graybalanced]

Greg Edwards wrote:

Apologies to the OP for massively derailing this thread, but this comment make me wonder how long until the fundamental limits of physics are met.

M1 and M2 are 5nm, whereas the M3 and M4 are 3nm (IIRC), with 2nm processor technology expected next year (that I'm guessing is good for the M5 and M6). Then maybe 1.5nm or the theoretical limit of 1nm .

With Moore's law currently running on a 3-year cycle, it could be dead in the next 10 years.

What comes next? Could an M7 or M8 be the last computer I'll ever buy?

Apologies for continuing the derailment, but it could be that future generations are less about the CPU. Already we see that as apps (especially for photography) use the GPU more, maybe we don’t need more/faster CPU cores, maybe a supercharged GPU is needed. Also with the rise of AI, the NPU becomes more important. The Mac’s NPU is the Apple Neural Engine and is already being used by the AI photo features of some apps. For video editors, the advances needed are not just CPU but GPU and video coprocessors for codecs.

Maybe the M8 CPU is not as advanced as much because it’s already all the CPU we need, but the GPU, NPU, and other coprocessors will be massively advanced.

 

[Thom Hogan]

As photographers, even 2 TB of internal storage isn’t enough to store our photos. So most photographers will need external storage anyway, wouldn’t you agree?

Absolutely. However you're mixing storage needs versus processing needs. Photoshop is not running on that external drive, it's running on Apple Silicon.

In my workflow, I use my Thunderbolt card reader to dump the card to my internal, do all the edits there, and then MOVE that folder to the external SSD, which gets backed up (mirrored) to the Synology. The two bottlenecks I encounter are copying GB’s of files, and culling through hundreds of thumbnails. To copy faster, for some odd reason, always copy a folder and not the files inside the folder. If you open a folder, select 500 photos and copy them to another drive, it takes several times longer than just copying the folder. To browse thumbnails faster (I use DxO), I limit my folders to about 500 photos, give or take. With this, I can cull my images without any lag.

Again, here you're talking about storage related things, not processing related. Specifically, when you ask Photoshop to open an image that's stored on your external SSD, then the following is true: Photoshop, Photoshop's cache, the image data, and any temporary working space needed to make a change to the image are all in the Apple Silicon chip. (Yes, you can point Photoshop to use an external cache, but that will create a clear performance hit on large files.) As you work on that image, as long as everything stays in that chip, you get maximal, very high performance response. The minute you start to spill out of that chip, if the spill goes to the internal SSD (and there's plenty of room on it) it will still be fast, but if it goes to an external SSD it will be slower.

You can minimize spills by working on one image a time and by having only Photoshop open (no other apps, even in background).

The alternative scenarios (multiple apps running, spilling to external SSD, etc.) may be "good enough" for you, but that's one of those things I'd want to verify. Moreover, as I pointed out before, the OP wants 10 years out of his new machine. Software bloats over time, the image files created by our cameras are getting bigger, and much more. What applies as "good enough" today may not apply in the future when you start stressing the memory chain in the future.