PIOK wrote:
So are you saying I would be ok with base Mac Studio 36 GB Memory... I think I read your comment before that 64 GB Memory is what people should get ?
I've advocated for some time that folks should be on the Max in order to have the memory bandwidth double the Pro, etc. ... and a minimum of 32GB. The 32GB represents the size of the storage (think size of a truck), and the bandwidth for memory operations to shuttle info back / forth to / from memory GPU / CPU is kinda like the speed limit on the highway. Bigger trucks can store more, higher speed limit can transfer at a faster rate. The matrix combination of storage / bandwidth renders how quickly operations can be performed ... while not the most technically correct presentation (conceptual) ... it's how I approach the amount of memory operations I need (maximum payload delivery, of sorts), when making comparisons / evaluations.
The first row would be the minimum that I've advocated since the M2. The third row is my personal configuration. The second row sits between my minimum recommendation and personal setup. I'm pretty sure that if you don't know a specific reason that you need MORE than 36GB, coupled with an M4 processor (i.e. heavy video editing, etc.), your work for stills should be fine with 36GB and 546GB/s bandwidth.
Just for reference, a Pro configuration (2X Modules) would have 1/2 the bandwidth of the Max configuration (4X Modules), and when combined with the lesser limits of memory storage, combined transfer operations are significantly less than even the "entry" 36GB of the Max. This "double reduction" is where I felt the "pain point" when I dropped back from the 32GB Max to the 24 GB Pro, when I was demo'ing different configurations. I didn't "feel" any difference between the 32GB Max vs. 64GB Max, but I got the 64GB for "future considerations".
On a similar note. When I bumped my ThinkPad Extreme from 32GB to 64GB ... I didn't "feel" a performance bump there either. Just another data point why I advocate for 32GB (or more if so inclined) if folks don't have a specific (known) need for 64GB. Dropping back to 24GB puts you into the Pro model, and the hit for 1/2 the bandwidth can be felt in different "smoothness" of operations ... particularly in context of shared memory between CPU Ops / GPU Ops / OS overhead.
24GB just isn't enough for serious (and future looking) ops that it won't be "felt" at some point ... only to leave you in the mode of needing / wanting to upgrade. And, given the fact that with Mac you can't just upgrade components ... you have to "repurchase" the case, power supply, CPU, etc. just to get your "upgrade" which effectively doubles your upgrade cycle costs because you upgrade sooner.
Yeah, folks gripe about the Mac's pricing / component upgrade approach ... but, better to "overpay" once for part of it, than "overpay" twice for all of it.
RustyBug wrote:
I've advocated for some time that folks should be on the Max in order to have the memory bandwidth double the Pro, etc. ... and a minimum of 32GB. The 32GB represents the size of the storage (think size of a truck), and the bandwidth for memory operations to shuttle info back / forth to / from memory GPU / CPU is kinda like the speed limit on the highway. Bigger trucks can store more, higher speed limit can transfer at a faster rate. The matrix combination of storage / bandwidth renders how quickly operations can be performed ... while not the most technically correct presentation (conceptual) ... it's how I approach the amount of memory operations I need (maximum payload delivery, of sorts), when making comparisons / evaluations.
The first row would be the minimum that I've advocated since the M2. The third row is my personal configuration. The second row sits between my minimum recommendation and personal setup. I'm pretty sure that if you don't know a specific reason that you need MORE than 36GB, coupled with an M4 processor (i.e. heavy video editing, etc.), your work for stills should be fine with 36GB and 546GB/s bandwidth.
Just for reference, a Pro configuration (2X Modules) would have 1/2 the bandwidth of the Max configuration (4X Modules), and when combined with the lesser limits of memory storage, combined transfer operations are significantly less than even the "entry" 36GB of the Max. This "double reduction" is where I felt the "pain point" when I dropped back from the 32GB Max to the 24 GB Pro, when I was demo'ing different configurations. I didn't "feel" any difference between the 32GB Max vs. 64GB Max, but I got the 64GB for "future considerations".
On a similar note. When I bumped my ThinkPad Extreme from 32GB to 64GB ... I didn't "feel" a performance bump there either. Just another data point why I advocate for 32GB (or more if so inclined) if folks don't have a specific (known) need for 64GB. Dropping back to 24GB puts you into the Pro model, and the hit for 1/2 the bandwidth can be felt in different "smoothness" of operations ... particularly in context of shared memory between CPU Ops / GPU Ops / OS overhead.
24GB just isn't enough for serious (and future looking) ops that it won't be "felt" at some point ... only to leave you in the mode of needing / wanting to upgrade. And, given the fact that with Mac you can't just upgrade components ... you have to "repurchase" the case, power supply, CPU, etc. just to get your "upgrade" which effectively doubles your upgrade cycle costs because you upgrade sooner.
Yeah, folks gripe about the Mac's pricing / component upgrade approach ... but, better to "overpay" once for part of it, than "overpay" twice for all of it.
Excellent synopsis and rationale, though the only additional consideration I would add with respect to the amount of memory is that additional memory also facilitates running multiple programs simultaneously.
10:20 ... comment about Max is TWO generations behind the PRO, yet it still has performance that exceeds the newer M4 (on exports). Part of my point about extending your upgrade cycle by going with Max vs. Pro levels (for the added expense of a few $$)
11:16 panomerge tells a good story. Notice that the M2M is 1/2 the time of the M2P and the M4P. The memory is the same in the older gen vs. gen, and the newer gen's M4 processor doesn't do much to offset / improve with 24GB. Goes back to my recommendation that 32GB is the breakpoint that I recommend to not go below.
11:40 and 12:16 story repeats itself with AI denoise, that the older Max still bests the newer Pro.
15:35 C1 export, the older Max thumps the newer Pro
18:29 story repeats in video, older max thumps newer Pro
A few different places where they are closer, but the salient point is (imo), the 4X (vs. 2X) of a Max vs. Pro configuration (Ultra if $$$$$ = 8X), provides a 200% benefit for certain tasks, while a faster / newer processor in Pro (half as many) can't keep up. Think of it like having 4 1 ton trucks that can go 70 mph vs. only 2 trucks, but they can go 80 mph. For hauling things that don't need more two tons, you can move two tons faster. But, for things that need more than two tons, you're gonna have to make extra trips, and that'll take longer than putting 3 or 4 truck on the road at the same time.
While the video is targeting the Ultra for max performance, I think it brings a nice presentation of the diff's between the Max vs. Pro, also.
Nothing here to shift my perspective ... for many folks, the break point of bang / buck (longer term) is the Max with at least 32GB. Pro with 24GB, just say "no" today, and you'll enjoy the Max 32GB (or more) longer, tomorrow. If the Pro / 24GB is truly your "tapped out" budget, its a fine enough machine. But, for a few more $$$ (certainly not 2X $$), you can get 2X performance (above the Pro 24GB) for a variety of tasks that can benefit from it.
armd wrote:
Excellent synopsis and rationale, though the only additional consideration I would add with respect to the amount of memory is that additional memory also facilitates running multiple programs simultaneously.
Thanks.
+1 for more applications / programs simultaneously.
Although, many will argue that if you want / need better photo performance (RAM related), just close those other programs. True enough, but it's also nice to have more RAM when you do run certain tasks that can harness it well. My main perspective is that 24GB isn't enough to do much with, so it takes a performance hit ... and will be even less so, as OS / GPU / CPU sharing demands will increase in future software development.
Apple computers are okay, but they only perform in the mid-range. Considering that you can get two faster PCs for the same price, the price-performance ratio is rather poor. To illustrate this, here are several benchmarks for comparison: Four predefined RAW files were denoised, and the time was measured. https://www.dslr-forum.de/threads/lrc-benchmark-ki-entrauschen.2148453/page-19#post-16772054
Stefan Official wrote:
Apple computers are okay, but they only perform in the mid-range. Considering that you can get two faster PCs for the same price, the price-performance ratio is rather poor. To illustrate this, here are several benchmarks for comparison: Four predefined RAW files were denoised, and the time was measured. https://www.dslr-forum.de/threads/lrc-benchmark-ki-entrauschen.2148453/page-19#post-16772054
Don’t be silly.
Apple computers _perform_ at a very high level, particularly with typical photography apps.
If you prefer some other system and OS, that’s absolutely fine. But please don’t spread misinformation.
Check out the real-world benchmarks. I’ve already linked them. They show everything without the need for discussion.
Instead of talking theoretically, I prefer real results.
They are okay, but not workstations when it really comes to speed and performance.
There is nothing better for video denoising than Neat Video – it is the market leader. Check out the results! You can also download it and test it on your own machine. https://www.neatvideo.com/download/neatbench
With Apple, it's not possible to denoise a 4K video track in DaVinci Resolve with Neat Video in real-time. If I feel like it, I'll throw in another 5090 and double the performance – and it still won't cost more than a studio Mac. The operating system doesn't matter anyway since I spend 98% of my time working in a program.
I just upgraded from the M1 Studio Max to the M4 Studio Max (base specs) and I can attest that the new model is far more powerful and efficient. Very happy with the upgrade and TBH it's the first time I've been "holy crap" at the difference when upgrading hardware.....and I've had a lot of upgrades since 2001 😎
James Burden wrote:
I just upgraded from the M1 Studio Max to the M4 Studio Max (base specs) and I can attest that the new model is far more powerful and efficient. Very happy with the upgrade and TBH it's the first time I've been "holy crap" at the difference when upgrading hardware.....and I've had a lot of upgrades since 2001 😎
You did not add extra Memory ?
Are you doing some Final Cut videos as well or this is mostly for LR and PS ?
I gonna get Studio M4 Max but I will think about Memory 48 GB vs 64 GB RAM > and of course 1TB SSD is what I want and need
James Burden wrote:
I just upgraded from the M1 Studio Max to the M4 Studio Max (base specs) and I can attest that the new model is far more powerful and efficient. Very happy with the upgrade and TBH it's the first time I've been "holy crap" at the difference when upgrading hardware.....and I've had a lot of upgrades since 2001 😎
Ballpark ... 40% bump in memory bandwidth, 2X bump in GPU performance, 60% bump in CPU performance ... sounds like the pieces added together, make for a nice jump from M1 > M4.
I'm sittin' pat on my 16" MBP M2 Max (64GB / 2TB)... at least till M6. On paper, I'd like to hold out till about the M8 (i.e. 6 years) or more, but we'll see how it goes over the next few years. GAS and all that stuff, ya know.
Stefan Official wrote:
Check out the real-world benchmarks. I’ve already linked them. They show everything without the need for discussion.
Instead of talking theoretically, I prefer real results.
They are okay, but not workstations when it really comes to speed and performance.
There is nothing better for video denoising than Neat Video – it is the market leader. Check out the results! You can also download it and test it on your own machine. https://www.neatvideo.com/download/neatbench
With Apple, it's not possible to denoise a 4K video track in DaVinci Resolve with Neat Video in real-time. If I feel like it, I'll throw in another 5090 and double the performance – and it still won't cost more than a studio Mac. The operating system doesn't matter anyway since I spend 98% of my time working in a program....Show more →
Check out real world performance, not "benchmarks" designed to show that OS A (or B) is miles ahead of the other for fanboys.
Your PC machines are fine, excellent even. Enjoy using them. But your OS Wars fetishizing dates you and makes you irrelevant.
PIOK wrote:
You did not add extra Memory ?
Are you doing some Final Cut videos as well or this is mostly for LR and PS ?
I gonna get Studio M4 Max but I will think about Memory 48 GB vs 64 GB RAM > and of course 1TB SSD is what I want and need
I did not since I never hit the wall with the 32gb on the M1, I did get the 1tb drive to have a blistering partition for raw files before moving them to storage on a 40gbps m.2 external. I do edit an occasional drone video although it's not enough to tax the system. I always do the exports while "getting coffee" I am wondering if proxies are even necessary now. DxO Pure Raw is exponentially faster with my A7R5 files though!
Another note, saving multi layer PS files was also exponentially faster to my m.2 drives. I don't know how to equate the speed increase since the m.2 dives aren't transferring faster but it's night and day faster. Things like a finder search for PSD files used to be 20 thumbnails with the rest as PS file icons, now it rips through then and renders the thumbnails like they are jpegs.
The final bonus is I am running a Dell 40" curved 5K screen for editing and under the M1 it was locked to 60 mhz. I had to use Better Display to edit the eeid file to get 100 mhz. The M4 delivers 120mhz natively....
Stefan Official wrote:
Check out the real-world benchmarks. I’ve already linked them. They show everything without the need for discussion.
Instead of talking theoretically, I prefer real results.
They are okay, but not workstations when it really comes to speed and performance.
There is nothing better for video denoising than Neat Video – it is the market leader. Check out the results! You can also download it and test it on your own machine.
The results that these "benchmarks" produce show performance a specific piece of software running on a specific operating system and specific piece of hardware. The results of these benchmarks are about the combination of these pieces. With this in mind, one might think, that running the same software and OS on different pieces of hardware would allow to compare the performance between different pieces of hardware. Unfortunately that's not the case. All these pieces, but especially the software part, involves a stack of many concepts and technologies. As Video denoising was mentioned, which is mostly ran on GPU's these days, let's focus on the abstraction layers/interfaces used by software developers to run their code on those GPUs:
OpenCL: available for many different types of GPU's and OS'es, with varying performance, especially on halfway modern hardware
CUDA: Developed by NVIDIA in order to better use their own GPU's performance. Only available on Windows and Linux
Metal: Developed by Apple for iOS and macOS, thus only available on their own systems
Software aiming at good performance targeting both Windows and macOS would use CUDA and Metal. The concepts behind these two are quite different. Software developers not only have to implement CUDA and Metal and optimize their code, they might have to optimize their software architecture (i.e. shared memory between CPU and GPU on modern Macs) or even adjust their algorithms in order to get optimized performance.
In practice, the amount of effort required to do this equally as well for Windows+NVIDIA GPU and Mac is something not many companies are willing to invest in or, in case of smaller companies/independent developers, even able to.
This is a simplification of some aspects that made the type of performance comparison the way it seemed to be intended to essentially useless.
There's a german saying: "Wer misst, misst Mist". It's a pun, playing with similar sounding words and their meaning in german. Roughly translated meaning "the one who measures measures crap".
Stefan Official wrote:
With Apple, it's not possible to denoise a 4K video track in DaVinci Resolve with Neat Video in real-time. If I feel like it, I'll throw in another 5090 and double the performance – and it still won't cost more than a studio Mac. The operating system doesn't matter anyway since I spend 98% of my time working in a program.
As outlined above, amongst other things, the OS does matter.
It seems obvious that you're primarily using Windows, at least for these things. What I don't get is why you seem to feel the need to comment in a thread that's clearly Mac specific?
gdanmitchell wrote:
If you are at the point where deleting less than 50GB from your SSD is necessary…
… you really need a larger SSD, not deleted files.
Quiet true. There's additional reasons to get a larger SSD.
There's an impact on the SSD lifetime if it's filled close to nominal capacity. Due to the way an SSD controller performs wear leveling in order to increase the limited lifetime of its storage cells.
SSD's internally have more storage capacity than available to the operating system, but it's limited. However, with many or most SSDs, it can use storage cells no by the User/Operating System for wear leveling as well. If the SSD is almost full, it has less capacity available to wear leveling, thus potentially wearing the SSD out sooner.
Some SSDs have more spare internal capacity available then others and Apple's internal ones seem to be doing better on that front than the average consumer SSDs. Still, I would suggest to avoid filling SSDs to capacity.
This is also one reason why I would recommend against using external USB SSDs: Even if they support what's called TRIM, macOS can't use it. With actual thunderbolt SSDs macOS can use that.
Then there's filesystems file systems, like HFS or APFS used by macOS. They slow down before they're full. (Overly) simplified: it's due to the way data and metadata is structured in order to handle read/write operations well.
I'm not aware of any official guideline by Apple, but this effect is not limited to Apple or the Mac and exists for essentially all systems today. The general rule of thumb: Filled beyond 80% performance is impacted, beyond 90% it's often getting nasty.
And, even more so ... to a degree of very specific machine / configurations considerations.
Kinda like saying I'm trying to decide between baking an apple or a cherry pie, to serve after a brisket dinner ... and someone comes along trying to tell you, that you should be eating tofu.
torstenb wrote:
Quiet true. There's additional reasons to get a larger SSD.
There's an impact on the SSD lifetime if it's filled close to nominal capacity. Due to the way an SSD controller performs wear leveling in order to increase the limited lifetime of its storage cells.
SSD's internally have more storage capacity than available to the operating system, but it's limited. However, with many or most SSDs, it can use storage cells no by the User/Operating System for wear leveling as well. If the SSD is almost full, it has less capacity available to wear leveling, thus potentially wearing the SSD out sooner.
Some SSDs have more spare internal capacity available then others and Apple's internal ones seem to be doing better on that front than the average consumer SSDs. Still, I would suggest to avoid filling SSDs to capacity.
This is also one reason why I would recommend against using external USB SSDs: Even if they support what's called TRIM, macOS can't use it. With actual thunderbolt SSDs macOS can use that.
Then there's filesystems file systems, like HFS or APFS used by macOS. They slow down before they're full. (Overly) simplified: it's due to the way data and metadata is structured in order to handle read/write operations well.
I'm not aware of any official guideline by Apple, but this effect is not limited to Apple or the Mac and exists for essentially all systems today. The general rule of thumb: Filled beyond 80% performance is impacted, beyond 90% it's often getting nasty.
Kinda reminds me of trying to defrag an old spinner drive. If you tried to do it when the drive was nearly full ... it was a royal pain compared to keeping it clean before it was full up. Yeah, I know, not the same ... but, the point being that when things get full, there's not as much "wiggle room" for things to move around, etc. and it takes more "micro-steps" of sorts > more reads / writes ops, etc.
BTW ... I just had an SSD crash on me at work last week. Fortunately, I keep it on the network and didn't lose any work.
It was a PC with a Samsung drive, that was almost 4 years of service.
torstenb wrote:
Quiet true. There's additional reasons to get a larger SSD.
There's an impact on the SSD lifetime if it's filled close to nominal capacity. Due to the way an SSD controller performs wear leveling in order to increase the limited lifetime of its storage cells.
SSD's internally have more storage capacity than available to the operating system, but it's limited. However, with many or most SSDs, it can use storage cells no by the User/Operating System for wear leveling as well. If the SSD is almost full, it has less capacity available to wear leveling, thus potentially wearing the SSD out sooner.
Some SSDs have more spare internal capacity available then others and Apple's internal ones seem to be doing better on that front than the average consumer SSDs. Still, I would suggest to avoid filling SSDs to capacity.
This is also one reason why I would recommend against using external USB SSDs: Even if they support what's called TRIM, macOS can't use it. With actual thunderbolt SSDs macOS can use that.
Then there's filesystems file systems, like HFS or APFS used by macOS. They slow down before they're full. (Overly) simplified: it's due to the way data and metadata is structured in order to handle read/write operations well.
I'm not aware of any official guideline by Apple, but this effect is not limited to Apple or the Mac and exists for essentially all systems today. The general rule of thumb: Filled beyond 80% performance is impacted, beyond 90% it's often getting nasty.
Even with spinning disks, I've always gotten a bit uneasy about going past 80%, and I think that's probably a pretty good boundary for SSDs, especially those that get a lot of read/write cycles.
I'll confess that I've pushed things a few times in the past, perhaps with a backup drive or similar. And it didn't take long for the slowdowns to rear their heads.
It is kind of like EV range specifications. (For the record, I like EVs.) You are never going to drive 350 miles on a charge in your EV that claims a 350 mile range. And if you try, the last few miles are going to be very, very unpleasant. Roughly the same holds true for hard drives — you aren't going to actually store 8TB on a a 8TB drive, and if you try you will not be very happy.
I am wondering if proxies are even necessary now. DxO Pure Raw is exponentially faster with my A7R5 files though!