Apple Hardware Questions for Davinci Resolve Studio

Hello hello!

I'm currently working on a 2019 MacBook Pro with a 2.6 GHz 6 Core Intel i7 processor, Radeon Pro 555X 4GB card, 256GB of Internal SSD, and 32GB DDR4 RAM. While I feel like these specs should be sufficient for video editing, I definitely feel like my workflow is quite slow at the moment. I'm a concert/landscape/wedding videographer and photographer, and I want to start improving the quality of my videos to the point where I'm using slow motion, VFX, and other more intense areas of Davinci Resolve studio (especially for the concert work). Thus, I'm considering an upgrade to an M2 MacBook or Studio. Regardless of the form factor of laptop vs desktop, I'm curious about what recommendations others might have about spec upgrades. Specifically, I'm curious about:

- CPU and GPU Core Upgrades: How many cores does Davinci Resolve utilize? Will I reach a point where performance will plateau?

- RAM: I'm currently thinking 64GB of RAM. My idea is to build a Mac Studio that I can have for years to come. Will 64GB of RAM be enough to make the computer I get feasible for long term use (ie: 5-10 years)?

- Internal Storage: I currently have some SanDisk Pro SSDs, so I definitely have proper external storage. How useful would upgrading the internal SSD on an Apple computer be? Currently looking to get 1TB of internal storage, and I could always load the media I need onto the desktop while I work on the project, and then offload it once done. Open to suggestions

Happy to provide more details if needed, looking forward to hearing what you all say!

Thank you!

1. DR is using all the GPU cores you can throw at it, 76 in my current testing.
2. With that many GPUs, 8K sources and heavy effects in an UHD timeline, it can use close to 100 GB in DR.
3. Internal storage is not much of an issue on Mac Studio, one TB would suffice. It has enough ports to use far cheaper external storage.

If you don't desperately need a laptop, I'd recommend the Mac Studio. If it needs to be a laptop, have a look at the new MB Air with the 15" screen, but get 16 GB RAM and not the smallest SSD (which is slow).

P.S. Over the next few days I'll be posting more results. Questions welcome, but the machine goes back soon.

...Questions welcome, but the machine goes back soon.

I have a top-spec M1 Ultra Mac Studio. If you want to contact me directly about doing some comparative tests, let me know. I am most interested in certain Resolve GPU and encode/decode areas that may have benefited from M2 Ultra improvements (vs M1 Ultra).

In particular, there were allegedly some hardware GPU scalability improvements and video engine parallelism improvements, but this has not been well tested by 3rd parties. I'd suggest these be done on the release version of 18.5 if possible.

Sure, I'm using the release version.
I can send you a .drp of my test. Most of the sources are demo footage from camera manufacturers, so I don't need to keep all those gigabytes in the cloud.

I have fiber internet, so if you have media test files on the cloud, it's no problem to download those. If they are on DropBox, I could do a cloud-to-cloud transfer.

We should use the same test procedure and cache config to avoid variation.

I'm interested in several areas:

- M1 Ultra vs M2 Ultra performance increase for Resolve compute-intensive operations such as TSNR, Magic Mask, Face Refinement, Depth Mask, Depth Map, etc. Reason: evaluate alleged GPU scalability improvements in M2 Ultra.

- Single-stream encode performance for common codecs. Reason: evaluate any hardware acceleration improvement, esp. if there is any sign of using multiple accelerators in parallel on a single stream. I don't really expect much, but at least on ProRes, segmented encoding of a single output file is theoretically possible.

- Frame corruption/wrong order on H265/HEVC export. Unknown how widespread that is, whether it's hardware/OS or software, but given these machines, it's a good idea to test it. See this thread for details: viewtopic.php?f=21&t=184873&p=965248#p965248

- I also have an M1 Max MacBook Pro 16, so I can test that as well.

I'm using heavy effects that are active at render time to be comparable. NR with variations in settings, Speed Warp, SuperScale enhanced, Relight, Depth Map, plus some lighter ones like FilmGrain, Halation, Aperture.
I'm not going to change that any more, since the big guy is going back very soon.

If you want a specific test of Magic Mask, Stabiliser or Face Refinement, send me the footage and your settings ASAP, and I'll try.

Most of my sources are RAW and massive, I'd rather like that all who want to try download them straight from the source. Publishing them to the forum might pose a copyright issue too, since the manufacturers want you to register.

1. DR is using all the GPU cores you can throw at it, 76 in my current testing.
2. With that many GPUs, 8K sources and heavy effects in an UHD timeline, it can use close to 100 GB in DR.
3. Internal storage is not much of an issue on Mac Studio, one TB would suffice. It has enough ports to use far cheaper external storage.

If you don't desperately need a laptop, I'd recommend the Mac Studio. If it needs to be a laptop, have a look at the new MB Air with the 15" screen, but get 16 GB RAM and not the smallest SSD (which is slow).

P.S. Over the next few days I'll be posting more results. Questions welcome, but the machine goes back soon.

Appreciate all of the advice! The current MacBook I use works fine for photo editing and light video editing (the battery isn't great but so be it), so definitely not in desperation to get a new laptop. My main goal with the purchase is to get the most longevity I can for my investment, which is why I'm leaning towards the studio.

You mentioned that there isn't really a GPU plateau. Are you aware of any CPU plateau? Just curious

For the Studio, I'm looking at a few configurations:
1) M2 Max, 30 Core GPU, 64GB RAM, 1 TB SSD
2) M2 Max, 38 Core GPU, 64GB RAM, 1 TB SSD
3) M2 Max, 38 Core GPU, 96GB RAM, 1 TB SSD
4) M2 Ultra, 60 Core GPU, 64GB RAM, 1 TB SSD

Based on your experience, what would you recommend? I mainly shoot in 4k using Canon's IPB or IPB Light compression modes from some of their newer mirrorless cameras, if that provides any helpful insight.

Thanks again!

Number one would be a very good basic configuration.

Number four would need 128 GB RAM and be an optimum. It'll simply be faster with demanding DR tasks, but number one is as good if you are not under tight deadlines.

For DR, which is hammering all GPU cores, you'll need enough RAM in relation to the number of GPU cores. IMHO, 64 GB will be fine for 30 cores, while 60 will need 128. Less would not only slow down the machine, but ruin the internal SSD in the long run. Fusion, OTOH, will use all RAM you can afford.

What do you mean by plateau? DR is using all GPU cores to the limit, and Blender (for example) all the CPUs at 100%.

...If you want a specific test of Magic Mask, Stabiliser or Face Refinement, send me the footage and your settings ASAP, and I'll try...

Thanks very much for the offer. In hindsight, I'm too busy on a project right now to follow up. I really wish I could, because this is an interesting area. Maybe there will be a later opportunity.

Number one would be a very good basic configuration.

Number four would need 128 GB RAM and be an optimum. It'll simply be faster with demanding DR tasks, but number one is as good if you are not under tight deadlines.

For DR, which is hammering all GPU cores, you'll need enough RAM in relation to the number of GPU cores. IMHO, 64 GB will be fine for 30 cores, while 60 will need 128. Less would not only slow down the machine, but ruin the internal SSD in the long run. Fusion, OTOH, will use all RAM you can afford.

What do you mean by plateau? DR is using all GPU cores to the limit, and Blender (for example) all the CPUs at 100%.

With regards to the plateau perspective, I just want to ensure that I'm not buying too much power to the point where it's sitting idle, but it seems based on what you're saying that there really is no such thing as too much power when it comes to using the advanced capabilities of Resolve. Am I correct in my interpretation of what you're saying?

When you say number one would be as good as number four if you aren't under tight deadlines, do you mean to say that both will get the job done but number 4 will do it significantly faster? When I shoot concerts, I get video clips out within 24 hours (just color grading, stabilizing, etc; nothing crazy). I'd like to start doing more complicated edits (ie: for reels) within that time period, though. Which machine, in your opinion, fits that situation better?

If you have the time/willingness, I have more of a learning question for you, as this is not my area of expertise. When you say that option #4 would "need" more RAM, what exactly do you mean? It sounds like you're saying that your RAM should be around 2x the amount of GPU cores, I'm interested to learn why that is. Additionally, can you explain how having insufficient RAM would slow down the machine and ruin the SSD?

Thank you again for all of your thoughts on the matter!

Regarding RAM, it's relatively simple to explain: those new Apple chips have no VRAM in the usual sense any more, but 'unified RAM'. So, each GPU core is grabbing same RAM for itself when active, and twice the number of GPU cores will grab about twice as much RAM. I have not yet tested a Mac with 30 or 38 GPU cores, but my laptop works with the identical project with 16 cores within the limits of its 32 GB, while the 76 core machine needs up to 100 GB.
In the next two days I'll have a laptop with an M2 coming, which has 38 GPU cores, but only 32 GB. That might be getting too tight. I'll report.
What happens when the RAM is insufficient? Those Macs will still be very stable, without giving out-of-memory warnings. The OS starts to swap memory to the internal SSD, which is fast, but much slower than RAM. So, insufficient RAM will slow the machine down. What's even worse, an SSD will not live forever , but has a limited number of read/write cycles. Consequently, such swapping will shorten its lifetime if happening regularly over extended periods.

Regarding the 30 vs. 60 cores machine: even if they don't scale linearly, the latter will give you close to double performance of the first for DR. IMHO, the 76 core machine, just like the 38 one, is overpriced if you calculate the performance added, even under optimal conditions.

Regarding RAM, it's relatively simple to explain: those new Apple chips have no VRAM in the usual sense any more, but 'unified RAM'. So, each GPU core is grabbing same RAM for itself when active, and twice the number of GPU cores will grab about twice as much RAM. I have not yet tested a Mac with 30 or 38 GPU cores, but my laptop works with the identical project with 16 cores within the limits of its 32 GB, while the 76 core machine needs up to 100 GB.
In the next two days I'll have a laptop with an M2 coming, which has 38 GPU cores, but only 32 GB. That might be getting too tight. I'll report.

I'd love to hear how that M2 laptop treats you when it comes in.

What happens when the RAM is insufficient? Those Macs will still be very stable, without giving out-of-memory warnings. The OS starts to swap memory to the internal SSD, which is fast, but much slower than RAM. So, insufficient RAM will slow the machine down. What's even worse, an SSD will not live forever , but has a limited number of read/write cycles. Consequently, such swapping will shorten its lifetime if happening regularly over extended periods.

That makes sense, thank you for the explanation!

Regarding the 30 vs. 60 cores machine: even if they don't scale linearly, the latter will give you close to double performance of the first for DR. IMHO, the 76 core machine, just like the 38 one, is overpriced if you calculate the performance added, even under optimal conditions.

The only reason I'd consider the 38 core would be to get the RAM up to 96GB, but sounds like you're saying that amount of RAM would be more useful when used with the 60 core machine. Am I inferring correctly?

Within your experience, if I were to narrow down my selection choices to either Option 1 or Option 4, what do you think would be the maximum use case for each one? If I were to jump from Option 1 to Option 4, what would be the additional capability that I would gain (ie: ability to smoothly edit 8k, 2x render speeds, etc)? I feel like I'm at the point where I'm considering two very different machines and I'd like to understand what I'll really be able to do with each.

RAM is always useful, but in this case you need to weight carefully cost vs effect. If Fusion is (or will be) important to you, I'd always get the maximum you can afford.
But if it's DR only, you may see diminishing returns. I'll get that machine tonight, so I'll report. From my results until now, I'd expect 32 GB to be insufficient for 38 cores, but 64 should be plenty.

So test is running and 32 GB is definitely not enough if you have 38 GPU cores. Whenever DR encounters a demanding effect the machine is swapping and everything is getting very slow. This happens with a timeline in 8K.
If the timeline is set to 4K/UHD, it's getting better, but you'd still be living on the edge. Memory pressure is high and it's getting very close to using all the 32 GB, occasionally swapping again.

IMHO, one should not buy a Mac with 38 GPU cores with less than 64 GB, just like you should have 128 GB for 76. With 30 cores and 32 GB or 60 cores and 64 GB, you'll still encounter swapping in rare cases.

Nevertheless, still not a single crash.

So test is running and 32 GB is definitely not enough if you have 38 GPU cores. Whenever DR encounters a demanding effect the machine is swapping and everything is getting very slow. This happens with a timeline in 8K.
If the timeline is set to 4K/UHD, it's getting better, but you'd still be living on the edge. Memory pressure is high and it's getting very close to using all the 32 GB, occasionally swapping again.

IMHO, one should not buy a Mac with 38 GPU cores with less than 64 GB, just like you should have 128 GB for 76. With 30 cores and 32 GB or 60 cores and 64 GB, you'll still encounter occasional swapping.

Nevertheless, still not a single crash.

Great to hear so far! Where I’m currently leaning is towards 38 cores and 96gb or 30 cores and 128gb. The swapping matter is definitely something I want to avoid, thus the desire for the 96gb with the 38 cores. Let me know your thoughts, and thanks for running some tests!

38 cores should be fine with 64 GB, 96 is overkill. The laptop here has only 32 and is running on the edge.
But for 60 or more cores, I'd get 128 GB.

38 cores should be fine with 64 GB, 96 is overkill. The laptop here has only 32 and is running on the edge.
But for 60 or more cores, I'd get 128 GB.

The reason I included the 96gb option in my previous message is I want to avoid the issue of memory swapping/degrading of the SSD due to lack of RAM so I can preserve the longevity of the machine. As I start to use more intense Resolve effects or as I use larger video files, I’m curious if that extra bit of RAM would help to future proof the machine a little bit. Is the studio with the 38 cores even capable of utilizing that much RAM? If it is, maybe that’s a safer long term play.

Also, with regards to the 60 core, I’m interested in the base model of the Ultra, but I’m not sure I’m open to spending USD $4800 on the machine with the upgraded at the current moment. That extra performance from the boost in cores, though, is definitely of interest to me. If I were to get the Ultra with 64GB of RAM, what real world effect would that have? How quickly could that potentially degrade an SSD? Would we be looking at a reduction of the computers long term usability?

Hi,

I am left with a wonder after the M4 announcement. It seems weird that you can only get the 64gb and 128gb unified ram option with the maxed out M4 Max... You cannot get it with M4 Pro.

It left me with wondering if the ram is that much faster.

After a bit of digging I found that the M4 has LPDDR5X-7500 compared to the M3's LPDDR5-6400. What I couldn't find is what the actual real life performance and comparison is.

So for example, the M4's 48gb is equivalent to the M3's 64gb or 48gb, etc?

I'd love to get your takes.

BTW I am a videographer and use Davinci and having a hard time deciding what is the best configuration.

Speed of the RAM has nothing to do with the size needed in relation to GPU cores. I'm pretty sure the same rule discussed above still applies. Of course, I have not tested that yet, with those machines even on pre-order.
But IMHO the maxed out one with 40 GPUs is overpriced and the RAM does not fit well. The second highest version would be the sweet spot, with 32 GPU cores and 64 GB. But even the Pro would be a nice machine. The biggest advantage of the Max are dual encoder/decoder units.

Hi,

I am left with a wonder after the M4 announcement. It seems weird that you can only get the 64gb and 128gb unified ram option with the maxed out M4 Max... You cannot get it with M4 Pro.

It's because the M series are single chip designs. You don't get any choice after manufacture; whatever comes off of the production line is what you get.

It left me with wondering if the ram is that much faster.

After a bit of digging I found that the M4 has LPDDR5X-7500 compared to the M3's LPDDR5-6400. What I couldn't find is what the actual real life performance and comparison is.

Due to the all in one chip factor, it's irrelevant to the end user. What you see is what you get.

That said, the main improvement in the M4 compared to the M3 is the GPU... which is part of the reason that all of the comparison are to the M1, and to the M3.

So for example, the M4's 48gb is equivalent to the M3's 64gb or 48gb, etc?

I'd love to get your takes.

BTW I am a videographer and use Davinci and having a hard time deciding what is the best configuration.

More is better. Period. There's no other variable to care about with Apple Silicon.

As Uli pointed out since the GPU and CPU share memory, using the GPU heavily can eat up your system memory and crush your performance due to swapping.

How much that will affect the software you're using depends on the software. Some will handle it by becoming less responsive as it has to wait longer to get data, some will crash. The OS can't control that.

Resolve usually just becomes less responsive, but some more latency sensitive software like RedShift (at least in Houdini) tends to self-destruct That may have been cured by now, but I can't test it yet because my license isn't currently active, but my machine also only has 32GB... which for Houini/Redshift is way too small. It's plenty for Resolve, but I have a discrete GPU with 16GB, which is equivalent to 48GB of unified memory, in case that helps.

When we tested my benchmark on a machine with too low RAM in relation to the GPU cores, it just became slow under Resolve by swapping, but it was still running stable.

The 128GB model of the M4 Max is too much of a price jump IMO, however, the 40-core GPU with 64GB RAM option falls a bit below your 2GB / GPU core - hopefully it will be good enough - you mentioned once that about 1.5GB/core might be enough - what do you think, Uli?

I would think so, we found very few situations driving an Mx Mac over that ratio.
What are the source formats you are usually using and which are the most demanding functions in DR?
And then, your SSD won't die right away from occasional swapping.

Good considerations. HEVC and ProRes editing/delivery and I do make use of the Fusion page but not a heavy Fusion user at the moment. I'm happy with my M1 Max (see signature) but I am also excited to see where the latest M4 Max MacBooks are going. Too bad my pocket book doesn't like them, lol. I took a look at maxing-out a 16" M4 Max with 128GB RAM and 8TB SSD and it came to over $9200 (CAN). Dialling it back to 64GB RAM and a 2TB SSD was just over $6000 (CAN). In both cases I'm talking about the 16 CPU Core with 40 core GPU version.

Maxed-out models were always crazy expensive by Apple. I think it's their policy to make money from those who want to show off. I normally found the middle-ground much more balanced.

Even the new Mac mini will handle DR pretty well, if you follow the 'rule' I gave above. Heck, my wife has been doing quite a bit of AR and video work with her basic M1 MBP, and the cheapest Mac mini M4 is stronger than that.

Maxed-out models were always crazy expensive by Apple. I think it's their policy to make money from those who want to show off. I normally found the middle-ground much more balanced.

It's also related to the single-chip design. The big chips aren't going to get great yields even at TSMC, and especially on TSMC's bleeding edge fabs. Truthfully, if these chips weren't so heavily redundant they'd be sheer idiocy. Their design however allows for disabling bad cores, bad memory, bad CPU cores, etc in order to get a working processor to compensate for yields.

This is the reason that AMD's Ryzen AI monstrosities aren't single chip designs. The packaging is a lot more complicated, but the yields on a single 8-core compute chiplet will ALWAYS be higher than on a single chip with 14-16 cores (even if a bunch of them are compact), 128 GB of embedded memory, a GPU, I/O complex, AND an AI chip.

Also, AMD has to allow for customizable memory configurations because it sells processors to integrators, unlike Apple.

Even the new Mac mini will handle DR pretty well, if you follow the 'rule' I gave above. Heck, my wife has been doing quite a bit of AR and video work with her basic M1 MBP, and the cheapest Mac mini M4 is stronger than that.

If you're running Houdini and rendering with Redshift (which is the only GPU rendering engine that I know of that supports nVidia, AMD AND Apple GPUs), the 2x1 system to GPU memory is going to be a much bigger issue, but for video, it's not nearly as critical.

I've been doing some reasonably big composites in Nuke on my aging Flow X13; that only has 32GB of system memory + a 16GB GPU, and that's been doing well. Houdini however runs the system out of memory very easily.

Fusion will probably be fine on a 64GB M4. Houdini, not so much. I definitely need more memory.

It depends. For complex compositing tasks, a Mx Mac is not the best solution. In most cases you'll need heaps of RAM and that's too expensive for that SoC design. A strong desktop PC will be the better solution.
But in our experience Mx Macs are really nice machines for DR.

True. And that's why there are no macs in any major VFX studios, even though some professional compositors use macs to remote into their virtual workstations, which is becoming more common according to one of our local VFX producers.

The tradeoff for Apple Silicon is that it's designed primarily around mobile, so while its performance is largely identical when on battery vs on mains power, it can't scale up like some of the higher spec gaming systems. Some of the ones I'm eyeing constrain the CPU to 20 watts when on battery and can't even power the dGPU without mains power.

Or maybe you can convince those to power the dGPU on battery so that you can watch the batter meter go from full to flat in six minutes

Exactly. Apple's silicon shines in mobile devices, but buying a Mac Pro with it is simply a waste of money (and space, with a nearly empty box). I even doubt that a Mac Studio makes a lot of sense, now that the tiny Mac mini is out with enough fast connectors.
If you need to be very mobile, but not off mains power, it's attractive. For DR. Not for heavy compositing.

That sounds about right.

I don't really believe that anyone else will overtake the Apple battery life leadership, but as far as performance goes, the Windows/Linux side will almost certainly run away from Apple again. Especially since if Qualcomm keeps its ARM license we'll have four competitors in that space instead of two for once.

Competition is good. And long overdue.

I bought my Studio Max at a time when I could not get a GPU upgrade for my PC and if available was roughly the same price as a 3090. I have been happy with it as it is silent and with the hardware decoder/encoder makes editing with my GH6 files easy that the PC found difficult. However when it comes to rendering with noise reduction and fusion scrolling credits the PC wins by a lot. An example of short edit with NR then fusion scrolling credit . iPad less the 1 fps, Studio Max about 3 fps, PC with 4070Ti 13fps. Of course a simple encode of 3840x2160 GH6 files with simple colour then both Studio Max and PC do almost 200 fps. But I think it shows the difference between 10GPU cores on the iPad, 24 GPU cores on the Studio Max and the 4070Ti when the hardware encoders are not enough and the real GPU needs to be used. Actually if it is simple stuff the iPad is fast enough !! Almost wish one could get an iPad with 32G of memory.

Almost wish one could get an iPad with 32G of memory.

That would kill off MacBook sales!

Maybe. Personally, I still prefer a large screen, standing upright, that doesn't get touched by me all the time.
And then BM still doesn’t officially support all pages on the iPad. Who else would need that much RAM if not DR users?

But IMHO the maxed out one with 40 GPUs is overpriced and the RAM does not fit well. The second highest version would be the sweet spot, with 32 GPU cores and 64 GB.

I tried to configure 16" Max with 32 GPU cores and option for 64GB isn't available, site says "48GB, 64GB, or 128GB available with M4 Max with 40‑core GPU only".

Let me suggest 20 cores and 48 GB then. For me, stability matters more than speed, since in the end you may even save time. And then, more cores will not scale linearly for everyday tasks in DR. There are just a few where you’ll feel those 50 something percent. If I would need a new machine right now, I’d go for that.
Oh, I just noticed, are we talking MBP or Mac mini here?

Let me suggest 20 cores and 48 GB then. For me, stability matters more than speed, since in the end you may even save time. And then, more cores will not scale linearly for everyday tasks in DR. There are just a few where you’ll feel those 50 something percent. If I would need a new machine right now, I’d go for that.
Oh, I just noticed, are we talking MBP or Mac mini here?

How about the new Mac Mini M4 Pro chip with 14‑core CPU, 20‑core GPU, 64GB & 1TB SSD storage? $2400USD for this configuration.

If you don’t need off-mains operation, I’d consider that seriously. Even 48 GB would be plenty for DR alone.

If you don’t need off-mains operation, I’d consider that seriously.

What do you mean by “off-main operation”? Like portability?

Off the power grid.

Off the power grid.

Got it. No, I’m always working in studio. I go off the grid if I want to get away from all things technology and just zen.

Let me suggest 20 cores and 48 GB then. ... Oh, I just noticed, are we talking MBP or Mac mini here?

I was talking MBP. To get 48GB on 20 cores GPU one would have to scale CPU down to M4 Pro starting at $2899. M4 Max with 40 core GPU and 64GB of memory starts at $4199, over 40% more. So question is how much faster later one will be, especially once heat (and in turn thermal throttling) of Max chip kicks in, will it justify that cost premium.

What did you use until now? My experience always was that the best value for money from Apple is the middle ground, not the maxed-out ones. And then, speed in everyday work depends so much on your source media, the features you need regularly, and your deadlines, it's hard to judge.
For example, if your sources are in HEVC 10 bit UHD, even the base model will be about as smooth as ProRes is for others, since it has the necessary hardware. OTOH, if you regularly need to decode and encode GOP codecs at the same time under a tight schedule, the Max has twice the engines for this.
But then, you may be right about some thermal throttling of the Max.

I guess we'll have to wait and see. Difficult to predict at this point.

I think the value is in the hardware decode/ encoders for smooth editing. Once GPU effects or Fusion is used then memory and GPU cores are important and then I think a PC is a better deal for costs and performance. I think the Mini also has competition from the many new Windows NUC's that are about the same size, are cheaper, upgradable memory and some even have the ability to use a dock for full size GPU's. The Intel NUC's also have the hardware decode/ encode for 10bit 4:2:2 H265 etc. OF course they do not have ProRes encode if that is what you want.

I still believe the Studio is a better deal. In Canada comparing the kitted out Mini costs with the Studio Max it is just $200 cheaper but for 10 cores less on the GPU every thing else the same. Here that amounts to $3500CAN. For that one could get a full size PC and an Intel NUC !! For example a M4 Mini with 10 core CPU 10 core GPU 24G RAM and 512SSD is $1300 CAN. A Beelink SEi14 Intel Mini PC is $925 with 14core Intel CPU, 32G RAM ( can be user upgraded to 96G ), 1T SSD other interfaces much the same as Mini.

What did you use until now? ... OTOH, if you regularly need to decode and encode GOP codecs at the same time under a tight schedule, the Max has twice the engines for this.

I am still on Windows but keep myself informed about MBP for the day I will need it and be able to afford it ...

I'm looking to move from a MBP M1 to the new Mini because I was a larger screen and to update in general.

I'm a light DR user and I'm thinking about the 10 CPU 10 GPU with 24G memory. Is the 24 enough or should I step up to 32?

thanks,
chris

Should be perfectly fine for DR.

Come on, don't get all obsessed with that 1.5/2x RAM to GPU count idea.
Sure, if you're going to torture your system with extremely heavy workloads, you'll get 128GB RAM anyway and be happy, but then you'll not be asking these kind of questions.
If you do middle-of-the road DR work with some NR, some magic masks, color grading and stabilization, sure enough, 64GB will easily cut it, even 48GB will in most cases.
Skimping on GPU cores simply for not meeting that "1.5-2x rule" will slow down your overall timeline performance. It can make the decisive difference between "realtime playback stutters" and "realtime playback is smooth".
Render-caching also uses all available GPU cores to the fullest, so waiting for those red lines to turn blue also is faster with more GPU cores.

That's correct, in particular if there seems to be an option to replace the SSD instead of the MoBo now.

Come on, don't get all obsessed with that 1.5/2x RAM to GPU count idea...

Good point. Where does this 1.5/2x number come from?

A key performance constraint in many GPU-related tasks using VRAM is bandwidth between the GPU and main memory. IOW the CPU can't just do a one-way, one-time data transfer to the GPU VRAM. Buffers are constantly being copied back and forth. On an RTX 4090 in a motherboard supporting PCIe 4.0 x16, the maximum transfer rate is 64 gigabytes per second per direction, and almost all apps use unidirectional transfers.

By contrast, in a unified memory architecture, the maximum transfer rate on an M4 Max is 546 gigabytes per second, 8.5x as fast.

Also, GPU memory management was fundamentally changed starting with Apple Silicon M3. On M3 and newer, it uses Dynamic Caching that allocates the use of local on-GPU memory buffers (likely tile cache) in hardware in real time. Any prior GPU memory performance test on earlier Apple Silicon machines would have to be re-run on M3 or newer to be valid. This improvement may also involve GPU-assisted address translation to reduce GPU memory bottlenecks in a unified system, as documented in Apple's patent US20210271606A1: https://patents.google.com/patent/US20210271606A1/en

It's likely possible to determine GPU core vs main memory constraints by actual testing, but it would (1) Be for a specific workload and (2) Require taking an XCode Instruments trace of the workload suspected of GPUs starving main memory and (3) It would only make sense to do that on M3 or newer systems.