Speed Up Processing Using Prime Noise Reduction

Prime NR is very effective but when it’s added to processing RAW files it takes far too long to render. I’m using a machine with plenty enough chest hair and custom built to run graphics. It’s dreadful how long a batch of RAW renders will take using Photo Lab Prime NR.

Have you tried other noise reduction products that produce comparable results? I sometimes use Neat Image, which works great - but makes PL Prime noise reduction look like a speed demon :slight_smile:

Also, if you’re trying to speed things up with your GPU, PL relies on OpenCL, which, in turn relies on the driver supplied by the GPU manufacturer. It might be as simple as updating that driver.
I’m still trying to sort out the hardware metrics that matter to PL performance, but it doesn’t seem to be super multi-processed or multi-threaded, so throwing CPU cores at the problem won’t make things faster, but single-core speed might. Also, the speed of the storage device the cache is on seems to affect performance.So RAM drive or PCie NVME SSD. I’d pop for that latter just to test but my current chipset doesn’t support NVME. Sigh…

Good assessment of the situation, Bob, and thanks for the info.

I’ve used Nik Dfine for years and use NR on nearly all images. Fine works very well and quickly. However, I’d rather run the RAW files into PhotoLab and get most of the initial work done there, including NR. Prime NR seems to be a bit better than Dfine is another reason.

Admittedly Prime can seem pretty slow, but of course the results are worth it. I understand that Prime in PhotoLab is faster then it was in Optics Pro 11, Hopefully they will continue to fine tune it.

Mark

On my ThinkPad, processing a ISO200 6000x4000 RAW image with Photolab PRIME usually takes more than 2 minutes, my Neat Image 6.1 (rather old ) is much faster (but doesn’t process RAW).

In my opinion, the results from NeatImage are even better than Photolab PRIME in many cases.

Oliver

i am used to use dfine in case of high iso jpeg’s after processing of a raw including that default NR, (somehow the manual fine tuning of all NR sliders costed me more time then i wanted in my former rawprocessor as learning and working out what’s best setting speaking, and got just lazy by run them through PSE/Dfine, (no batch processing so time consuming clicking) Not on all but only the one’s i liked be cleaner.

So now i use Prime Noise on default and that’s it.
(i tested PN against my best effort manual on the other application and Dfine (export in tiff 16bit open in pse and apply dfine. and export in jpeg and dfine.)
I could equal PN but not really do better. only ofcoarse when i got jpeg’s out the camera (panorama shot’s and such)
As far as time /speed of process?
This is just export and walk away or do something else on the pc. The other was a repeatable action which was far more time consuming.

I rather want best possible outcome (IQ wise) then giving that up for more speed.
if i want more speed the other nr setting will do.
And i am not sure but the processing time is depending on the work it needs to clean up so if default Prime noise hits a clean base iso it is far quicker then processing a 3200 iso noisy one.
So if it can quicker great! as long as it not bring a lesser outcome. (1 minut waiting extra for a better image you keep for ever? any time! :smile: )

Prime is a speed demon compared to how it used to run in Optics Pro. And, of course, it all depends upon the speed of your processor, the memory available, the characteristics of your graphics card and the like.

Prime can take as long as 20+ seconds on my Mac Mini, and that is a long time, but I put that down to how slow the Mini is in general. All of this used to be much faster when I ran it on a MacBook Pro with a faster chip and a real graphics card with real memory.

In my opinion, the results from NeatImage are even better than Photolab PRIME in many cases.

Makes sense. For anyone not familiar with it, they have (or you provide) noise characteristics for each ISO of your camera, they then process based on those characteristics. My experience is that it works better when you build your own noise profiles with the camera you’re going to use, and better yet if you build two sets, one with a normal temp camera, the other with a hot sensor. Major PITA, but great results.
But I’ve since moved to a camera with a lot less noise (D750), shoot raw-only and ETTR.

BTW, if you have 6.1, you should be able to upgrade to 6.[whatever the letest minor version is] for free - so long as they haven’t changed their policy.

@MikeFromMesa I wonder how it can be possible that your Mac Mini is six times faster than my 2.3GHz Core i5-6200U CPU. Are you really using PRIME? How large are your RAW images?

@bobkoure the NeatImage profiles are obviously not applicable after processing a RAW image with DxO because there is no information about the tone curve applied. Therefore NeatImage is no alternative to DxO noise filtering but only an additional tool if DxO’s results are not good enough.

In my previous posting, I wanted to express that good noise filtering is possible without PRIME’s “magic”.

@obetz
I suppose it is possible that I remembered the High processing time instead of the Prime processing time. I am traveling now but will check when I get back home.

@obetz
I have returned from my trip and ran some tests using Prime on my raw images.

I shoot with an Olympus E-M1 Mk II, which is an m4/3 20MP camera. The Micro FourThirds sensors are small and sensitive to noise, so I generally do not shoot at high ISOs unless I am birding at some of the campgrounds we stop at. Then the birds are only around in the late afternoon and the trees make for dim light and I end up shooting at the high end of the available ISO values for the Olympus, which is 25,600. At that value the images are mostly unusable so the max I generally use is 6400 or, upon occasion, 12,800.

Since those are birding photos they generally are heavily cropped and those are the images I process, so I repeated the processing using both the full images (which I almost never use) and the cropped one. The results are that the cropped images that I process take about 20-25 seconds to process (which is the value I remembered and posted) using Prime while the full images take about 80 seconds.

My Mini is a 3 GHz i7 chip with 16GB of RAM. The chips in the Late 2013 Mini are only dual core, not quad core, and I have PhotoLab set up to process 2 images simultaneously.

Hi Mike - - This is completely off-topic (for which I apologise in advance), but … you might be interested in this; Taking advantage of ISO-invariance characteristics

Regards, John M

@John-M

That is interesting. As the author says, the common practice lately has been to shoot to the right while he is advocating shoot to the left which, I think, is more practical with camera systems like the M43 with its small and noise sensitive sensor, but also requires a very good raw engine.

What makes this so interesting for me is that I just returned from a camping trip in California and found myself taking a lot of low light birding photos in the late afternoon. The sun was out but the campground was full of trees, so little of it made it to the ground where the birds were. To make a long story short, I ended up under-exposing some of the most interesting photos that I took and they came out pretty well with PhotoLab’s raw engine.

https://www.dropbox.com/s/vc2vahwetvhvri3/OLY01386.jpg?dl=0

Thanks for the link.

Thanks for the image, Mike - - - That’s a great result from your under-exposed capture.

John