Skip to main content
Participant
January 27, 2024
Question

Computer advice. Sorry if this has been done 1000's of times already

  • January 27, 2024
  • 1 reply
  • 2302 views

I currently have a 2019 i7 16gb MacBook pro. 

I need to upgrade as this is getting on a bit now and running denoise AI is taking roughly 7 mins on a 85mb raw image. ( 42mp Sony a7R iii ) 

 

I'm torn between getting another MacBook pro, Air  or Mac mini with a M1 / M2 chip & 16gb ram. Would this be a big time improvement to run denoise AI ?

 

I would potentially be open to windows options too if the money is right. 

 

Any information would be greatfully received. 

 

Many thanks 

 

1 reply

Conrad_C
Community Expert
Community Expert
January 27, 2024

AI Denoise is almost exclusively GPU-based, so performance goes up a lot on more recent GPUs and more powerful GPUs. All of the options below will do AI Denoise faster than your current Mac. The conclusions below are from watching a number of stress tests by the ArtIsRight and MaxTech channels on YouTube, so you can check those out if you want.

 

The way it works out for your choices are:

 

MacBook Air: Probably avoid. Although it will process AI Denoise much faster than your current Mac, it will be slower than any of the Macs below.

 

Mac mini: This can work. Upgrading the processor to M2 Pro may cut AI Denoise time a little more because it has a third more GPU cores than the M2. Important: If the cost of the Mac mini approaches $1999, consider getting the base Mac Studio instead, because at that price the Studio has an M3 Max with almost double the GPU cores of the Mac mini M2 Pro, and 32GB Unified Memory.

 

MacBook Pro: Consider the 14"/16" MacBook Pro, because they have a good balance of GPU power and cooling. But for AI Denoise, consider avoiding the base M3 MacBook Pro, and going up to an M3 Pro or M3 Max instead.

 

Current M2/M3 models may bring your AI Denoise time to around one minute or below per image. For even faster AI Denoise processing, upgrade the above Macs to an M2/M3 Max or Ultra, which have many more GPU cores and can cut that time down to around 30 seconds per image or less, depending on the number of megapixels.

 

If you are going to do batch processing (run AI Denoise on a large number of images at once), you have to think about cooling. Batch AI Denoise will max out the GPU, and if you do that for at least a few minutes, the Mac will reach its maximum temperature. If the Mac has no fan (MacBook Air), it must slow down to stay within temperature limits. The Macs with fans can continue at full speed. Note that the base 14" M3 MacBook Pro has only one fan, so if you are serious about batching AI Denoise, upgrade to the M3 Pro or M3 Max model which have two fans and more GPU cores. If you want the fastest possible bulk AI Denoise in a Mac laptop, upgrade to the 16" because the larger size gives it a higher capacity two-fan cooling system. For the fastest possible bulk AI denoise in a Mac desktop, you probably want the Mac Studio with M2 Ultra (60 GPU cores, or 6 times more than a MacBook Air) and its beefier cooling system.

 

If you go Windows, same thing: Discrete GPU is strongly preferred over Intel integrated graphics, and the more powerful the better. Relatively few Windows laptops have a discrete GPU. Also be aware that many Windows laptops throttle down both the CPU and GPU on battery because of both wattage and heat concerns. That is much less of a problem with Apple Silicon Mac laptops, which can run CPU and GPU at full power on battery largely because they can get the same work done with fewer watts, which also generates much less waste heat, for less fan noise. The fastest AI Denoise on any platform is, I think, a PC with effective cooling and a monster new graphics card (e.g. RTX 40xx) where the GPU alone can cost as much as a Mac, and the whole thing consumes hundreds of watts. But that can get AI Denoise down to 10-15 seconds per image.

 

This linked chart summarizes a lot of this, using an A7r IV raw file:

https://www.fredmiranda.com/forum/topic/1804640/17#16330477

 

In all cases above, it is strongly recommended that it have 32GB Unified Memory or more. The MacBook Air and Mac mini can only take a maximum of 24GB, but that can be OK if you’re on a budget.

D Fosse
Community Expert
Community Expert
January 28, 2024
quote

a PC with effective cooling and a monster new graphics card (e.g. RTX 40xx) where the GPU alone can cost as much as a Mac, and the whole thing consumes hundreds of watts. But that can get AI Denoise down to 10-15 seconds per image.


By @Conrad_C

 

It's not nearly as bad as that.

 

A Nvidia RTX 4060 will do 42 MP in less than 20 seconds and sells at B+H for a little over $300.

 

That is a price anyone can afford, if they can afford a computer at all, and the most cost-effective solution all things considered. It works off a standard power supply.

 

You're thinking of the top end RTX 4090, which is massive overkill for this (but will do Denoise in 5-10 seconds).

 

 

Participant
January 7, 2025
quote

So I've had a list out together to build a computer myself.


By @scott88m14

 

I'm a huge fan of building yourself. Not to save money (although you do) - but mainly because you get a clean Windows installation without any "helpful" vendor modifications and customizations. And no additional software that wants to organize your life. What you get is a reliable workhorse that never fails.

 

As for the list - note the warning about BIOS compatibility! You don't want that. Go back one generation on the CPU. The motherboard/CPU/RAM combo is where you need to spend some time to ensure they will work together.

 

The golden rule is to pick components that have been on the market for 6 months to a year. Then the bugs have been ironed out, BIOSes have been updated etc. Avoid the just-released bleeding edge.

 

Make sure the RTX 4070 has physical room in this small case. I prefer full ATX motherboards to have room for any eventuality and good cooling (although I see the temptation of a smaller box). Also make sure the power supply has the necessary PCI outlets and connectors for it. 650 watts is a bit on the low side, but it should work.

 

I'd up the NVMe drive to 2 TB, and generation 4 (PCIe 4.0). The speed increase is significant, and you'll need the space. Also consider your storage needs in the future - do you need to add more disks? Relying on external drives cluttering your desk is both impractical and unsafe, considering cables and connectors are exposed. Putting your drives inside the box is much safer.

 

 


i just did a 'stopwatch' test on my M1 mac mini vs a new prebuilt gaming pc with the 4060 that my son bought (right at 1000$) because building it would have been about 1200$. 

 

Test results are :

pc - 15 seconds (ish)

mac mini - 1 minute ish.

 

Mac mini does great for general processing/editing of photos in lightroom. But clearly struggles with Denoise.
The mac mini M1 has 16 GB ram (not sure if getting more makes that big of a difference).

 

the pc specs are :

  • AMD Ryzen 5 7600X3D (4.1GHz) Processor
  • Gigabyte B650M C V2 Motherboard
  • 32GB DDR5-6000 RAM
  • NVIDIA GeForce RTX 4060 Graphics Card