jhapeman wrote:
So for those who care...the additional neural engines in the Studio Ultra models really do make a difference in overall speed. My 16" M2 Max has 16 neural engines vs. the 32 in the Studio Ultra, but they also run at a faster clock speed. On the Studio Ultra the total process from hitting the Denoise button until the DNG is saved on average takes 21 seconds; on the M2 Max it takes on average 26 seconds.
As a reminder my Nvidia RTX 4080 takes the same 21s to process the same files. The only consumer level card that would probably perform better would be an Nvidia RTX 4090, but those are $2000 video cards.
BTW, it's hard to tell when Adobe is doing the actual denoise calculations, but I think it's part of the whole DNG creation process. It's definitely different on my PC and Studio Ultra--the Studio will load an image preview of the Denise in less than 2 seconds; it takes the PC with the Nvidia card about 6 seconds to do that, but ultimately the total time to finish is very similar for both. ...Show more →
You said you're testing on A1 raw files?
Been interesting testing the M1 Pro against various files.
A7R4 Compressed RAW - 61mp 61mb - 51 seconds
Nikon Z9 HE* RAW - 45mp 31mb - 40 seconds
Fuji X100v RAW - 26mp 56mb - 22 seconds
I expected to see a little more correlation to file size, but that doesn't seem to be the case, it correlates much closer to pixel count, which sort of makes sense I guess. I don't have any old uncompressed Sony files but would be interesting to see if there is a compressed vs uncompressed difference.
RoamingScott wrote:
You said you're testing on A1 raw files?
Been interesting testing the M1 Pro against various files.
A7R4 Compressed RAW - 61mp 61mb - 51 seconds
Nikon Z9 HE* RAW - 45mp 31mb - 40 seconds
Fuji X100v RAW - 26mp 56mb - 22 seconds
I expected to see a little more correlation to file size, but that doesn't seem to be the case, it correlates much closer to pixel count, which sort of makes sense I guess. I don't have any old uncompressed Sony files but would be interesting to see if there is a compressed vs uncompressed difference.
Yes, A1 lossless compressed raw files. Given that's doing a Bayer demosaicing process as well, the correlation to pixel count is logical. I tested it out with some older files I have from a variety of cameras and it mostly holds true, although it's also dependent on the total noise in the image as well.
RoamingScott wrote:
I honestly don't know much about Ryzen, although I'm noticing the people with the longest times (anecdotally on FM) seem to have that in common. I have a markedly worse GPU than you, and an ancient i7 and my times are around 60 seconds on my old desktop and half that on my newer M1 mac.
Update:
Finally got back to my system and verified I had "Use GPU" selected. I found and downloaded a new driver for my GeForce RTX 2060 6 GB and ran several more test. I also used the GeForce Experience program to optimize LR.
It's now ripping through Canon R5 45mp images in about 25 seconds now.
By Ian Lyons "Currently, the Apple Neural Engine found in M1/M2 silicon Macs is not used by Denoise. It's expected that performance will be even better when Denoise does use the Apple Neural Engine."
"Apparently, an issue on Apple side means that the Neural Engine is not used by Adobe Denoise. When the issue is addressed, then Adobe Denoise is ready to take advantage of the Neural Engine."
Then confirmed by Rikk Flohr "information is correct regarding the Neural Engine on Apple devices"
lighthound wrote:
Update:
Finally got back to my system and verified I had "Use GPU" selected. I found and downloaded a new driver for my GeForce RTX 2060 6 GB and ran several more test. I also used the GeForce Experience program to optimize LR.
It's now ripping through Canon R5 45mp images in about 25 seconds now.
lighthound wrote:
Update:
Finally got back to my system and verified I had "Use GPU" selected. I found and downloaded a new driver for my GeForce RTX 2060 6 GB and ran several more test. I also used the GeForce Experience program to optimize LR.
It's now ripping through Canon R5 45mp images in about 25 seconds now.
I feel a lot better now.
How did you use the GeForce Experience to optimize LR? I have my latest drivers for my GTX, but I don't have Experience installed. Easy to do? What does it do?
I ran this photo below in both Topaz and LR new denoise.... LR did a better job (at 50) in removing moire than Topaz. But here is the kicker...
Time to run Topaz was maybe 10 seconds.
Time to run LR was 25 minutes (and it makes my PC choppy, and flashes)
Yep... I have some experiences where LR flies fast... but most of the time... I dunno why it varies so darn much.
Anyone with GTX cards? What are you experiencing?
Oh, my GPU is slammed at 100% and my CPU is dead asleep. At this point, I can't use LR.. it appears a tad better but not by much. I see that Adobe is saying to use an RTX card... well... given the bit coin issue that drove up prices and super scarcity, I can't believe there are that many RTX cards out there... this has got to be a real issue for Adobe... Topaz can do pretty much the same thing and it screams fast compared to LR... Hmmm.......
Bruce n Philly wrote:
How did you use the GeForce Experience to optimize LR? I have my latest drivers for my GTX, but I don't have Experience installed. Easy to do? What does it do?
Peace
Bruce in Philly
When you go into GeForce Experience you should see two tabs. (at least on my older version) One "HOME" and the other "DRIVERS". When you are in the HOME tab you should see all available apps that are on your system that can be optimized for your GPU. Mine showed LR plus all the DXO apps. You can click on each one then select optimize or (as I later discovered) there is a drop down in the top right corner where you can select optimize all.
I don't know if the fix for me was simply getting the latest drive or a combination of the latest driver plus this optimization using GeForce Experience. I checked the preferences setting in LR afterwards and it was still set the same as it was.
Ondrej wrote:
From other forum - 24Mpx ISO 12800 RAW, default Denoise with GTX 1050 Ti - 4 min / RTX 3060 - 10 sec
Very interesting... thanx for posting this. I have a GTX 980 Ti SC card... 4 min compared with 25 mnts and longer. Something is just wrong... I don't know what it is. The RTX 3060 card is approx $375 today.
Most of my timings are stupid long and unusable, but sometimes I get a much faster time. Just odd. As I posted prior, my PC as a whole is in Passmark's 99th percentile of performance and my graphics card by itself in my machine still rates highly. I guess per Adobe, they wrote their code to use features that first appeared in the RTX GPUs.... boy that cuts out a ton of Adobe customers with lower end cards.
In short, even with say 4 mnts... LR Denoise is unusable for me, especially when Topaz is so darn fast and works virtually the same for me.
Being a geek, I am compelled to run out and get a new graphics card... but I still don't see the benefits of this over Topaz. I don't want to use another product and continue to pay for Topaz support. But buying a new card and still not getting the speed of Topaz...... about $375... not really worth it to me.
Just odd that Adobe would release a product feature that requires the latest cards... I wonder what percent of their user base has these cards... it can't be more that 50% and I suspect it is way lower than that. Topaz makes it happen with lower-end cards, why did Adobe so poorly implement their feature? Oh well...
I will continue to work on this because I can't help myself.
It's about Machine Learning, Ray Tracing and Tensor Cores, so it's more about the age of the card.
Even a weaker modern one will definitely do a better service with AI acceleration.
Bruce n Philly wrote:
Very interesting... thanx for posting this. I have a GTX 980 Ti SC card... 4 min compared with 25 mnts and longer. Something is just wrong... I don't know what it is. The RTX 3060 card is approx $375 today.
Most of my timings are stupid long and unusable, but sometimes I get a much faster time. Just odd. As I posted prior, my PC as a whole is in Passmark's 99th percentile of performance and my graphics card by itself in my machine still rates highly. I guess per Adobe, they wrote their code to use features that first appeared in the RTX GPUs.... boy that cuts out a ton of Adobe customers with lower end cards.
In short, even with say 4 mnts... LR Denoise is unusable for me, especially when Topaz is so darn fast and works virtually the same for me.
Being a geek, I am compelled to run out and get a new graphics card... but I still don't see the benefits of this over Topaz. I don't want to use another product and continue to pay for Topaz support. But buying a new card and still not getting the speed of Topaz...... about $375... not really worth it to me.
Just odd that Adobe would release a product feature that requires the latest cards... I wonder what percent of their user base has these cards... it can't be more that 50% and I suspect it is way lower than that. Topaz makes it happen with lower-end cards, why did Adobe so poorly implement their feature? Oh well...
I will continue to work on this because I can't help myself.
Being a "geek", you'd think the basic differences between a modern card and 8 year old card would be obvious when it comes to how Adobe utilizes the GPU for heavy processing. Benchmarks mean literally nothing compared to real world usage, especially with bleeding edge apps and features.
How long do you plan on blaming Adobe for your 8 year old GPU? Either way, you are right about something...you're between a rock and a hard place, and spending money is your only option if you want to use LR for this particular task.
I read through your post and the comments - sounds more like getting an understanding of what and how Denoise works than an actual bug.
My understanding is:
1. Denoise works on the raw data (which has no adjustments made to it, just the raw data from the camera-processed raw sensor data). So the results of Denoise (including the time it takes) will be no different if all the adjustments are zeroed out or if any or all of them are set to whatever.
Of course the denoised image that you see will be different depending on the settings/adjustments selected, but this is because of those settings / adjustments, not because Denoise is doing anything different.
2. It's best to use Denoise first, not because it does anything different then but because then you will be working on the denoised image and will be better able to make subsequent image enhancements. Also - some settings may look different on the denoised DNG file than how the look on the original raw file - so again best to use Denoise first.
I read through your post and the comments - sounds more like getting an understanding of what and how Denoise works than an actual bug.
My understanding is:
1. Denoise works on the raw data (which has no adjustments made to it, just the raw data from the camera-processed raw sensor data). So the results of Denoise (including the time it takes) will be no different if all the adjustments are zeroed out or if any or all of them are set to whatever.
Of course the denoised image that you see will be different depending on the settings/adjustments selected, but this is because of those settings / adjustments, not because Denoise is doing anything different.
2. It's best to use Denoise first, not because it does anything different then but because then you will be working on the denoised image and will be better able to make subsequent image enhancements.
Yes that appears to be accurate. They recommend to Denoise first as editing later edits can affect results. So some sharpening won't hurt anything as Denoise uses RAW data, ignores it but applies it to the new DNG.
I tried a file with sharpening at 150. It came back at 150 and it was over-sharpened. NR looked OK. I reduced it to normal levels and it looked fine. From now on I'll just use Reset before sending the file.
Like I said in that thread seeing a difference in Enhance with two different levels of sharpening threw me off.
So.... if this is how LR denoise works, can you just denoise as you import? If you are out shooting high ISOs, for example, and you know you are going to have noise, well then it may make sense to do just that. Experienced shooters kinda know what they will need given experience no?
Bruce n Philly wrote:
So.... if this is how LR denoise works, can you just denoise as you import? If you are out shooting high ISOs, for example, and you know you are going to have noise, well then it may make sense to do just that. Experienced shooters kinda know what they will need given experience no?
Possible?
Peace
Bruce in Philly
I'm pretty sure you can't because it isn't a preset bit you can run the files through as a batch after import.
