Use Cases for 128GB RAM

[Beau10]

The thing is though, Macs last really long. Case in point, I am playing with a 2003 PowerBook G3 with Panther installed and it still works great. There are still Intel Machines out there doing great, hence why Apple is encouraging them to upgrade to M3. With machines these powerful, Apple gonna have a super hard time getting many users off them when they introduce planned obsolescence. I'm keeping my M1 until 2027 because of what I do on it.

Sure, but back to my statement: "Conversely, if you're coming from a 6+ year old machine to get a machine that is 3x the performance at twice the cost of one that is 2.5x the performance makes little sense."

If you plan to keep a machine a long time, your needs are moderate and a lower spec machine (within reason) should be absolutely fine.

 

[ondioline]

I think one of them was MaxTech's Lightroom test video. Lightroom Classic is a notorious RAM hog, and I've seen it bog down due to RAM (doing a huge catalog operation) on a 128 GB Mac Studio a friend has... That experience is one reason I'm leaning towards 128 GB.

Yes pretty much any raster art program is going to suck up memory like no tomorrow at high resolutions. Other than like 70B parameter LLMs this is the main thing I can think of.

 

[raythompsontn]

Lightroom Classic is a notorious RAM hog, and I've seen it bog down due to RAM (doing a huge catalog operation) on a 128 GB Mac Studio a friend has

I use Lightroom on a 13" M2 Air, 16 Gig, 1 TB without issue. I can import 3K photos in under 3 minutes. Out of that I can run 2 export operations of about 150 photos with multiple adjustments applied, and a web export to a private website without issues. The exports take less than a minute to accomplish, the website a little more as I use FTP in Lightroom to do the upload.

Lightroom, and Photoshop, will use as much memory as they can grab, even though the apps don't need that much memory. There are settings in Photoshop to limit the amount of memory grabbed. If the system needs memory, Photoshop will release some of the memory.

Any decent OS will load as much into memory as possible, maximum the use of the memory. Then when other apps need the memory the OS will discard unused portions of memory. So how much memory is used when applications are open is not really a useful metric. The real metric is swap file usage (not size) as that indicates memory is being swapped due to a lack of memory space for the apps.

 

[danwells]

I use Lightroom on a 13" M2 Air, 16 Gig, 1 TB without issue. I can import 3K photos in under 3 minutes. Out of that I can run 2 export operations of about 150 photos with multiple adjustments applied, and a web export to a private website without issues. The exports take less than a minute to accomplish, the website a little more as I use FTP in Lightroom to do the upload.

Lightroom, and Photoshop, will use as much memory as they can grab, even though the apps don't need that much memory. There are settings in Photoshop to limit the amount of memory grabbed. If the system needs memory, Photoshop will release some of the memory.

Any decent OS will load as much into memory as possible, maximum the use of the memory. Then when other apps need the memory the OS will discard unused portions of memory. So how much memory is used when applications are open is not really a useful metric. The real metric is swap file usage (not size) as that indicates memory is being swapped due to a lack of memory space for the apps.

You don't mention the resolution of your camera...

 

[raythompsontn]

You don't mention the resolution of your camera...

45 Megapixels. Using RAW or JPG makes little difference. I have a high speed card and a high speed reader. The reader is the real bottleneck, not the Mac.

 

[danwells]

At this point, I think it's pretty darned possible to project five year needs for still photography (I'd hesitate in some newer categories like VR). I've gotten four years out of my last-generation Intel MBP (which is getting really slow with modern noise reduction algorithms, for example). I think a relatively loaded M3 Max will give me at least five years...

I would be shocked if I'm using a camera with more than twice the resolution of what I have now in five years. I use a 102 MP camera today, am pretty sure the next generation will be ~160 MP (a 40 MP APS-C sensor exists and the medium format version of that will be ~160 MP), and that we won't see a second additional generation within five years.

Imaging software needs more and more power and RAM (due to things like AI-powered noise reduction), but linearly, not exponentially.

My printer's not getting any bigger - a 44" printer is already a very large object.

We're not going to see 16K displays become common (a few semi-experimental ones may appear, but they won't be something I'm trying to fill on a daily basis), and I'm already comfortably out-resolving 8K.

I wouldn't be comfortable projecting 10 years out, because of the possibility of something like a fundamentally redesigned sensor or a compete architecture change on the computer side (Intel to Apple Silicon is recent enough that I'm confident about Apple Silicon in the five year term.

 

[jennyp]

I use Lightroom on a 13" M2 Air, 16 Gig, 1 TB without issue. I can import 3K photos in under 3 minutes. Out of that I can run 2 export operations of about 150 photos with multiple adjustments applied, and a web export to a private website without issues. The exports take less than a minute to accomplish, the website a little more as I use FTP in Lightroom to do the upload.

Lightroom, and Photoshop, will use as much memory as they can grab, even though the apps don't need that much memory. There are settings in Photoshop to limit the amount of memory grabbed. If the system needs memory, Photoshop will release some of the memory.

Any decent OS will load as much into memory as possible, maximum the use of the memory. Then when other apps need the memory the OS will discard unused portions of memory. So how much memory is used when applications are open is not really a useful metric. The real metric is swap file usage (not size) as that indicates memory is being swapped due to a lack of memory space for the apps.

ArtIsRight gives a demo of why you should use "at least 32 GB of RAM" better than I can. It depends what you're doing, but I find even more is better.

 

[Goodrich]

The thing is that with the m3 max you dont need to generate previews. I loved the size / weight of my M2 Air, but it took just a bit more than I was comfortable with to generate high quality previews with DXO Photolabs. Lightroom seems to be a memory hog.

 

[raythompsontn]

ArtIsRight gives a demo of why you should use "at least 32 GB of RAM" better than I can.

I have run Photoshop and Lightroom for years on a PC. I was working with 16 Gig of RAM on the PC with dozens of images open in Photoshop with 20-30 layers. I never had a performance issue. Now I have 64 Gig on the PC. I notice no difference in the way Photoshop operates.

I don't need, nor do most people, "at least 32 Gb of ram". What is more important is fast external storage. Photoshop does a good job of using memory and disc storage. A SSD improved the speed more than the memory.

Unless a person has actual experience within the environment, using the applications, then all that the person states is unfounded rhetoric copied from another source with the same rhetoric.

Large catalog operations in Lightroom are disc intensive, not memory. Lightroom imports are disc intensive. Lightroom exports are disc intensive, CPU intensive if the format is being changed and multiple modifications are being applied. I can export a couple hundred images on my M2 Air with 16 Gig, with multiple modifications, copyright overlay, and changing format in a couple of minutes. I doubt that increasing the memory would improve that process.

 

[ksj1]

I'd argue it's overkill for that. As powerful as the M3 Max GPU is, it's underpowered for situations where you'd need that much RAM, whether training or inference. These are pretty good though if you want to play with 7B/13B models.

Hmm, I think it is a bit more capable than that. I run inference on code-llama 34B with 5 bit quantization with pretty good performance.

 

[Beau10]

Hmm, I think it is a bit more capable than that. I run inference on code-llama 34B with 5 bit quantization with pretty good performance.

I've played with 34B models and 10-12 tok/sec is too slow for me after working w/ commercial models, but I can understand how that might be acceptable for others. Running even larger models would only make sense for eval purposes.

Also with a 5 bit quant 34B model you'll still have 24 gigs free on the 16/40 base model. Bumping to 64 makes sense if one feels a bit better for only $200, but spending another whopping $800 on top to get 128 is silly if the main purpose is LLM work. Save that $$ to purchase something more capable for this kind of stuff in another 2-3 generations.

 

[danwells]

I have run Photoshop and Lightroom for years on a PC. I was working with 16 Gig of RAM on the PC with dozens of images open in Photoshop with 20-30 layers. I never had a performance issue. Now I have 64 Gig on the PC. I notice no difference in the way Photoshop operates.

I don't need, nor do most people, "at least 32 Gb of ram". What is more important is fast external storage. Photoshop does a good job of using memory and disc storage. A SSD improved the speed more than the memory.

Unless a person has actual experience within the environment, using the applications, then all that the person states is unfounded rhetoric copied from another source with the same rhetoric.

Large catalog operations in Lightroom are disc intensive, not memory. Lightroom imports are disc intensive. Lightroom exports are disc intensive, CPU intensive if the format is being changed and multiple modifications are being applied. I can export a couple hundred images on my M2 Air with 16 Gig, with multiple modifications, copyright overlay, and changing format in a couple of minutes. I doubt that increasing the memory would improve that process.

How big are the images you're using? Since it's working on one image at a time, PS is only memory intensive for certain operations (or for very large pano stitches and the like). Lightroom(Classic - CC is much more memory efficient) is memory intensive for catalogs with a ton of images, or where the individual images are large.

 

[jennyp]

I have run Photoshop and Lightroom for years on a PC. I was working with 16 Gig of RAM on the PC with dozens of images open in Photoshop with 20-30 layers. I never had a performance issue. Now I have 64 Gig on the PC. I notice no difference in the way Photoshop operates.

I don't need, nor do most people, "at least 32 Gb of ram". What is more important is fast external storage. Photoshop does a good job of using memory and disc storage. A SSD improved the speed more than the memory.

Unless a person has actual experience within the environment, using the applications, then all that the person states is unfounded rhetoric copied from another source with the same rhetoric.

Large catalog operations in Lightroom are disc intensive, not memory. Lightroom imports are disc intensive. Lightroom exports are disc intensive, CPU intensive if the format is being changed and multiple modifications are being applied. I can export a couple hundred images on my M2 Air with 16 Gig, with multiple modifications, copyright overlay, and changing format in a couple of minutes. I doubt that increasing the memory would improve that process.

Ah sorry I thought I could actually see the tool stutter while he was working. Didn't realise it was rhetoric.

 

[Harry Haller]

Content creation, VMs. and numerous browser tabs.
Large coding and Unreal Engine projects.
200+ audio tracks in music DAWs.
Unlimited multitasking.
The fact that memory is shared with the cpu and gpu.
Apps needing more and more memory because of feature bloat.
Non-upgradeable memory.
Future proofing for 5 year ownership.

 

[raythompsontn]

Ah sorry I thought I could actually see the tool stutter while he was working. Didn't realise it was rhetoric.

Stutter from what resource? Memory or Disc? File operations, especially conversions consume much CPU and disc activity. Even does so on my I9 with 64 Gig. I can white balance, crop, and exposure correct 2,000 images in less than 10 seconds on my M2 Air with 16 Gig. Memory pressure never leaves the green. I don’t think doubling the memory would change that scenario.

 

[Macshroomer]

It's not unusual for me on an out of town / onsite job to have DXO open cleaning up Mavic 3 Pro files while I stitch panos from a Z9 in Lightroom and process other types of raw files in Photoshop, all at the same time, bouncing from one open app to the next and then uploading those images as organized by PhotoMechanic for immediate upload.

In similar fashion of doing 2-3 things at once, these could be with 102MP Hasselblad X2D files. Some of these stitched medium format files can get enormous. After 35+ years in this biz, I have no issues putting up a few more bucks for hardware with *lots* of power to more than cover my bases. A good example of this would be back in July when the hired motion guy on my job needed my M2max for a bit to do video work because his laptop was glitching.

I always get hardware with lots of headroom because even at the price they are charging for these now, it is more than money well spent, it's one less thing to worry about.

 

[danwells]

Agreed about the 102MP files (GFX100S shooter here) - they are gorgeous, but they do step on your computer. All the "still photography doesn't use a ton of RAM" posts aren't using really high resolution files. If you're just running Photos (or even Lightroom CC - the newer cloud-based Lightroom is much less RAM-intensive than Lightroom Classic, BUT it doesn't do everything - notably, it still lacked a "print" command when I last looked at it a year or so ago) on iPhone files, or even 24-30 MP camera files, it fits in 16 GB.

Get into 40 MP+ images (especially 102 MP medium format), Lightroom Classic, Capture One and DxO, especially if you're running some big plug-ins, stitching or have a RIP going in the background, and 16 GB looks awfully small...

 

[ksj1]

This is interesting...

MLX framework released for AS

 

[iHorseHead]

Having just decided on the configuration for an M3 MBP, I am curious to know what use cases have caused people to specify 128GB RAM.

I could perhaps understand running (large) LLMs, heavy duty video / special effects and high-end music stuff (with many complex tracks)? Although for those sorts of workloads, I wonder if at least a Studio (for better ventilation) or an Ultra CPU might also be needed.

So, if you have ordered a MacBook Pro with 128GB RAM, it would be interesting to know why.

(I'm asking purely out of curiosity, by the way, I have no need of that much memory myself. It also strikes me that an extra £800 (GBP) here in the UK to go from 64 to 128GB is not particularly good value for money, unless it's an absolute necessity.)

Thanks .... Andrew

My school had PCs with 256GB of RAM and it was mainly for game development.