PL5/deep prime and mac studio ultra

lcubed11

Leading Member
Messages
982
Solutions
1
Reaction score
197
Location
US
has anyone had the chance to try this combination yet?

it will be a while before my ordered mac studio arrives and i'm looking

forward to what software to install for handling my photo needs.

does PL5 and deep prime take advantage of all the extra cores and gpu?
 
has anyone had the chance to try this combination yet?

it will be a while before my ordered mac studio arrives and i'm looking

forward to what software to install for handling my photo needs.

does PL5 and deep prime take advantage of all the extra cores and gpu?
What I can tell you is that PL5 does DeepPRIME processing on one of my ISO 25,600 a7RIII 42MP RAW files in 10 seconds on my $1099 16/512 Mac mini with the base M1 chip.

On the DxO forums, there's a spreadsheet where users report benchmark test results. The M1, M1 Pro and M1 Max times for batch export of D850 RAWs are 35, 31 and 28 respectively. The M1 mini is a killer bargain for DxO work.
 
can you give a pointer to that location in the DXO forums?

i'm probably searching for the wrong keywords.
 
can you give a pointer to that location in the DXO forums?

i'm probably searching for the wrong keywords.
DxO DeepPRIME Processing Times - Google Sheets

Just tested my two systems on a D850 Raw Image and I get:-

For my Apple 14" Laptop:-

Apple M1 Pro 16C / 32GB - 70 Seconds

And for my PC Editing Rig:-

RTX 3080 OC / 5950X / 64GB - 28 Seconds

I would think the Studio will be close to my PC rig.
 
Last edited:
can you give a pointer to that location in the DXO forums?

i'm probably searching for the wrong keywords.
DxO DeepPRIME Processing Times - Google Sheets

Just tested my two systems on a D850 Raw Image and I get:-

For my Apple 14" Laptop:-

Apple M1 Pro 16C / 32GB - 70 Seconds
Is that for ONE image??? If so, check preferences and make sure DeepPRIME is set to auto select Neural Engine rather than CPU or GPU. My M1 mini takes only 10 seconds for a 42MP a7RIII RAW.
And for my PC Editing Rig:-

RTX 3080 OC / 5950X / 64GB - 28 Seconds

I would think the Studio will be close to my PC rig.
--
"I learned long ago, never to wrestle with a pig. You get dirty, and besides, the pig likes it." - George Bernard Shaw
http://jacquescornell.photography
http://happening.photos
 
Last edited:
can you give a pointer to that location in the DXO forums?

i'm probably searching for the wrong keywords.
DxO DeepPRIME Processing Times - Google Sheets

Just tested my two systems on a D850 Raw Image and I get:-

For my Apple 14" Laptop:-

Apple M1 Pro 16C / 32GB - 70 Seconds
Is that for ONE image??? If so, check preferences and make sure DeepPRIME is set to auto select Neural Engine rather than CPU or GPU. My M1 mini takes only 10 seconds for a 42MP a7RIII RAW.
And for my PC Editing Rig:-

RTX 3080 OC / 5950X / 64GB - 28 Seconds

I would think the Studio will be close to my PC rig.
5 X D850 RAW Images.

Thanks for the tip, I now get 11 seconds for 5 X D850 RAW images on my RTX3080 PC with GPU selected, I will test my MacBook Pro when I get home with Neural selected:-

Screenshot-2022-03-28-143014.jpg
 
Last edited:
can you give a pointer to that location in the DXO forums?

i'm probably searching for the wrong keywords.
DxO DeepPRIME Processing Times - Google Sheets

Just tested my two systems on a D850 Raw Image and I get:-

For my Apple 14" Laptop:-

Apple M1 Pro 16C / 32GB - 70 Seconds
Is that for ONE image??? If so, check preferences and make sure DeepPRIME is set to auto select Neural Engine rather than CPU or GPU. My M1 mini takes only 10 seconds for a 42MP a7RIII RAW.
And for my PC Editing Rig:-

RTX 3080 OC / 5950X / 64GB - 28 Seconds

I would think the Studio will be close to my PC rig.
5 X D850 RAW Images.

Thanks for the tip, I now get 11 seconds for 5 X D850 RAW images on my RTX3080 PC with GPU selected, I will test my MacBook Pro when I get home with Neural selected:-

Screenshot-2022-03-28-143014.jpg
FWIW, that RTX3080 graphics card alone costs more than my entire M1 Mac mini with 16GB RAM and 512GB SSD.

--
"I learned long ago, never to wrestle with a pig. You get dirty, and besides, the pig likes it." - George Bernard Shaw
 
interesting. i just downloaded the PL5 trial and i just noticed that "Deep Prime" now allows the option of M1 Ultra GPU.

i haven't given that a try yet since i'm trying to figure out how to use ViewPoint with PL5 and discovered the GPU option under PL5 preferences.
 
interesting. i just downloaded the PL5 trial and i just noticed that "Deep Prime" now allows the option of M1 Ultra GPU.

i haven't given that a try yet since i'm trying to figure out how to use ViewPoint with PL5 and discovered the GPU option under PL5 preferences.
You can choose to designate the GPU for DeepPRIME processing with any Mac. However, with M1 variants, you'll get better performance from the Neural Engine on a base M1, roughly similar performance with NE or GPU on an M1 Pro, and possibly better performance with GPU on an M1 Max or Ultra.
 
i was surprised that the M1 Ultra was specifically called out vs a generic M1 option.
 
i was surprised that the M1 Ultra was specifically called out vs a generic M1 option.
I would have thought that they would just call out GPU acceleration versus Neural Engine acceleration. (Assuming that the software is not designed so as to be able to send some photos through one path, and some through another, at the same time.)

For the M1, people report that the Neural Engine path delivers better performance – but that might change by the time you move up to a M1 Max (with all of its GPU cores).
 
To follow up – if
  • You had a number of photos to process, and
  • You didn''t mind the possibility of small differences between GPU-accelerated and ANE-accelerated processing
an AI photo processing application could probably deliver the fastest speedup by sending some photos through the GPU while simultaneously sending others through the ANE.

Based on Anandtech's article about the M1 Pro and M1 Max, it sounds as though the M1 Max and M1 Ultra have the sort of memory bandwidth to pull this off. The M1 Pro might, too. I'm not as sure about the M1, but I can't see harm in an application vendor doing an experiment. Worst case, you just use one path on the M1, and both paths on the others.
 
To follow up – if
  • You had a number of photos to process, and
  • You didn''t mind the possibility of small differences between GPU-accelerated and ANE-accelerated processing
an AI photo processing application could probably deliver the fastest speedup by sending some photos through the GPU while simultaneously sending others through the ANE.

Based on Anandtech's article about the M1 Pro and M1 Max, it sounds as though the M1 Max and M1 Ultra have the sort of memory bandwidth to pull this off. The M1 Pro might, too. I'm not as sure about the M1, but I can't see harm in an application vendor doing an experiment. Worst case, you just use one path on the M1, and both paths on the others.
DeepPRIME batch times reported by users on DxO Forums are:

M1 - 35

M1 Pro - 31

M1 Max - 28

If I split a large batch between my M1 mini and M1 MBA, I get down to 35/2 = 17.5, for less total cost than a base-model 14" MBP.
 

Keyboard shortcuts

Back
Top