@mcbroomf Thanks for the info, that was very helpful!
Edit:
It's going to be interesting to see what happens with the M5 GPU. Right now, it's looking like the A19 GPU has about a 30% improvement over the A18 GPU. If that carries over to the M5, Apple could be very close to catching up to Nvidia for mobile GPUs.
@EB-1 I got to test LR vs DXO vs Topax Photo AI on my 2019 Mac this weekend, and the results were more surprising than expected.
With a 60mp a7rIV file I got:
LR AI Denoise = Approximately 90 seconds
DXO DeepPrime = Approximately 30 seconds
DXO DeepPrime XD = Approximately 60 seconds
Topaz Photo AI Raw Denoise = Approximately 180 seconds
It would be interesting if we could have a few people who have more than 1 Raw AI noise reduction program on their system, benchmark their speed. Interestingly, Topaz used a higher percentage of my GPU than LR or DXO, yet was by far the slowest. Also, I was most disappointed in the Topaz results. This was the first time I used Topaz Photo AI on a raw file instead of a tif file, which is a different AI model. DXO DeepPrime was also a nice surprise for speed, coming in at 30 seconds. Deep Prime XD, though, was a closer match to LR results (I actually liked it the best of all 4 models).
We would need to run comparisons on at least a few machines to know if these speeds translate across hardware. For example, it could be possible that Topaz would have run much faster if my GPU had more VRAM.
Just got an email fromTopaz. From what I can tell since my subscription for Photo AI was recently renewed I've been labeled a Founder (ie grandfathered) and been given Topaz Photo. In my account on line, Topaz Photo shows active with next billing date of Aug 2026 and the download button is also active. The Founder price will remain in effect as long as the subscription does not lapse otherwise the price will return to current level ie a lot more. So this looks like from a $ perspective it's all about the same as the Photo AI I'm currently using. Granted it's only the Photo version of the Studio 3 App combo but that might be an ok thing, at least for me.
DWOfPaul wrote:
@EB-1@ I got to test LR vs DXO vs Topax Photo AI on my 2019 Mac this weekend, and the results were more surprising than expected.
With a 60mp a7rIV file I got:
LR AI Denoise = Approximately 90 seconds
DXO DeepPrime = Approximately 30 seconds
DXO DeepPrime XD = Approximately 60 seconds
Topaz Photo AI Raw Denoise = Approximately 180 seconds
It would be interesting if we could have a few people who have more than 1 Raw AI noise reduction program on their system, benchmark their speed. Interestingly, Topaz used a higher percentage of my GPU than LR or DXO, yet was by far the slowest. Also, I was most disappointed in the Topaz results. This was the first time I used Topaz Photo AI on a raw file instead of a tif file, which is a different AI model. DXO DeepPrime was also a nice surprise for speed, coming in at 30 seconds. Deep Prime XD, though, was a closer match to LR results (I actually liked it the best of all 4 models).
We would need to run comparisons on at least a few machines to know if these speeds translate across hardware. For example, it could be possible that Topaz would have run much faster if my GPU had more VRAM....Show more →
I had to dig deep into the anals of my files, but was able to find some ARW a7rIV files from 2019.
I'm getting 2.77 sec. (n=50, max=2 simultaneous) with DXO XD/XD2 ARW to 16-bit TIFF. Global Sharpness, Chromatic Aberration, Vignetting were on.
Regardless of the fact your computer is older, the relative performance is not making me want to upgrade my Topaz to the latest pay plan. What version were you testing?
EB-1 wrote:
Regardless of the fact your computer is older, the relative performance is not making me want to upgrade my Topaz to the latest pay plan. What version were you testing?
Photo AI 3. I haven't given Photo AI 4 a try yet since there was no mention of Denoise improvements, and the main feature, Dust and Scratch removal, is not supported on Intel Macs. Just typing out that sentence is making me wish DXO would make a model that works on TIF files more and more. Denoise AI practically started Topaz Labs transformation and now image denoising seems like an afterthought as they try to push us towards subscriptions and cloud rendering
Topaz Photo AI 3.2.2 is the last version I'm licensed to use so I just updated and rebooted.
With only the AI denoise module selected, using the same 50 ARW files, the average time per ARW to 16-bit TIFF is 10.4 seconds.
I don't know about newer versions of Topaz but in this build DXO PL 8.8 is nearly 4x as fast with their slowest, most AI intensive model. I did not even apply any of the other basic corrections used in DXO.
EB-1 wrote:
Topaz Photo AI 3.2.2 is the last version I'm licensed to use so I just updated and rebooted.
With only the AI denoise module selected, using the same 50 ARW files, the average time per ARW to 16-bit TIFF is 10.4 seconds.
I don't know about newer versions of Topaz but in this build DXO PL 8.8 is nearly 4x as fast with their slowest, most AI intensive model. I did not even apply any of the other basic corrections used in DXO.
EBH
What is Topaz doing? If the results were better than DXO, maybe it would be worth the time, but I definitely preferred the DXO results. I need to dig out Topaz Denoise AI and see if it has Raw file support. On tif files, I found it faster, but did not time it, and on most photos, I preferred the results over Photo AI.
I may also have to retest Luminar Neo Noiseless AI. I got it as part of a bundle deal about two years ago. When I tried it at the time, the results didn't seem better than Denoise AI, and I haven't messed with it since. It sounds like it may have improved since that time. I believe it should work on tif files too.
They offered me a discounted rate to their studio package and some freebies, but while I can justify a subscription to a core program like Lightroom, I can’t to a package of add-ons, so I did not accept their subscription offer. I think they’re going in the wrong direction.
DWOfPaul wrote:
What is Topaz doing? If the results were better than DXO, maybe it would be worth the time, but I definitely preferred the DXO results. I need to dig out Topaz Denoise AI and see if it has Raw file support. On tif files, I found it faster, but did not time it, and on most photos, I preferred the results over Photo AI.
I may also have to retest Luminar Neo Noiseless AI. I got it as part of a bundle deal about two years ago. When I tried it at the time, the results didn't seem better than Denoise AI, and I haven't messed with it since. It sounds like it may have improved since that time. I believe it should work on tif files too.
The last version of DeNoise AI is 3.7.2 and supports older cameras in RAW, but newer ones were not added. They wanted you to use Photo AI. That ON1 NoNoise would be cheap and probably not need updating for TIFFs, but the examples show artifacts.
DWOfPaul wrote:
@EB-1@ I got to test LR vs DXO vs Topax Photo AI on my 2019 Mac this weekend, and the results were more surprising than expected.
With a 60mp a7rIV file I got:
LR AI Denoise = Approximately 90 seconds
DXO DeepPrime = Approximately 30 seconds
DXO DeepPrime XD = Approximately 60 seconds
Topaz Photo AI Raw Denoise = Approximately 180 seconds
It would be interesting if we could have a few people who have more than 1 Raw AI noise reduction program on their system, benchmark their speed. Interestingly, Topaz used a higher percentage of my GPU than LR or DXO, yet was by far the slowest. Also, I was most disappointed in the Topaz results. This was the first time I used Topaz Photo AI on a raw file instead of a tif file, which is a different AI model. DXO DeepPrime was also a nice surprise for speed, coming in at 30 seconds. Deep Prime XD, though, was a closer match to LR results (I actually liked it the best of all 4 models).
We would need to run comparisons on at least a few machines to know if these speeds translate across hardware. For example, it could be possible that Topaz would have run much faster if my GPU had more VRAM....Show more →
I have the latest DXO Pureraw (5.2). I use all the default settings and I am processing a file once every 6 seconds from the A1. This is with a AMD 6950 XT GPU and i7 12700K CPU
Fred Miranda wrote:
Just my luck...I bought Topaz Photo AI a couple of months ago and haven’t received a single update since. I do like how the app works right now, but I just hope they continue supporting it so it stays compatible with future Adobe updates.
From what I read, the existing perpetual licenses such as PhotoAI won't be supported and don't get updates anymore, unfortunately. You can use them at the current status as long as they work with Adobe updates and/or your camera (given you want to use it for your Raw files)
WJaekel wrote:
You can use them at the current status as long as they work with Adobe updates and/or your camera (given you want to use it for your Raw files)
and/or the servers that support your license continue to function.
I just bought PhotoLab9 as a replacement for both Topaz and CaptureOne. It has all the functions that I need, I think their DeNoise is better than Topaz.
After a little cursory playing I purchased Topaz Photo AI on Jan 1, 2025, which I believe was the last day of their annual "sale". It may have been December 31, 2024. Because an extra year of updates was also available at a significant discount, I purchased that at the same time.
I later discovered that Photo AI was too heavy handed for my taste. Through a careful comparison that I should have done before I purchased, I found that I could do a better job that's less destructive at the pixel level with ACR and Photoshop, which I already subscribe to. Live and learn. At least when my Topaz updates were done I could continue to use it, perhaps on occasions when an image was marginal.
Then I discovered through an email from Topaz that my account was converted to a subscription and the extra year of updates I bought was a year's subscription. That's the last thing I wanted. Fortunately they cancelled that extra year for me. I was told that I can continue using what I currently have (V 4.04) for as long as it'll work. It's not the first photo software I've tried and abandoned. It may not be the last, but it probably should be.
Although I haven't always liked what Adobe has done as a business, I've got to say that since I began using Photoshop v 4.0 in 1996 it (later along with ACR) has been a fantastic product that does everything I've needed it to do. YMMV.
I know I'm a grumpy old man but I just don't the whole subscription thing for software. Same as I don't get renting a place to live, leasing a vehicle, etc.
I've decided recently to just keep my old unsupported windows desktop going offline as my main photo editing machine. Still running CS6 on it and it's more than enough for me. I tend to stick with things a long time and just cannot wrap my mind about the whole subscription scheme.
Apple announced the M5 today. We will need to see how it performs in the real world, but it appears the GPU has gotten about 30-50% more powerful vs the M4. If this holds with the eventual M5 Pro and M5 Max, we could be talking about serious AI Denoise speed improvements.
The M4 Max was already ahead of basically all mobile CPU's and a lot of Desktop CPU's. The M5 Max will probably, at a minimum, help Apple keep its lead in CPU performance. On the GPU side, it appears like the M5 Max could very well end up taking on the best mobile GPU's from Nvidia and AMD for everything outside of gaming, which is more of a software support problem than a hardware problem.
Looks like 2026 is going to be an interesting year for computer hardware for power users.
In the announcement they talk about Neural Accelerators built into each GPU core. This is different to the Neural Engine (which is still there but with no additional cores). It will be interesting to see how this affects AI processing (eg Adobe Denoise). Let's hope it doesn't break it ...
https://www.apple.com/newsroom/2025/10/apple-unleashes-m5-the-next-big-leap-in-ai-performance-for-apple-silicon/ Apple today announced M5, delivering the next big leap in AI performance and advances to nearly every aspect of the chip. Built using third-generation 3-nanometer technology, M5 introduces a next-generation 10-core GPU architecture with a Neural Accelerator in each core, enabling GPU-based AI workloads to run dramatically faster, with over 4x the peak GPU compute performance compared to M4.1 The GPU also offers enhanced graphics capabilities and third-generation ray tracing that combined deliver a graphics performance that is up to 45 percent higher than M4.1 M5 features the world’s fastest performance core, with up to a 10-core CPU made up of six efficiency cores and up to four performance cores.2 Together, they deliver up to 15 percent faster multithreaded performance over M4.1 M5 also features an improved 16-core Neural Engine, a powerful media engine, and a nearly 30 percent increase in unified memory bandwidth to 153GB/s
mcbroomf wrote:
In the announcement they talk about Neural Accelerators built into each GPU core. This is different to the Neural Engine (which is still there but with no additional cores). It will be interesting to see how this affects AI processing (eg Adobe Denoise). Let's hope it doesn't break it ...
https://www.apple.com/newsroom/2025/10/apple-unleashes-m5-the-next-big-leap-in-ai-performance-for-apple-silicon/ Apple today announced M5, delivering the next big leap in AI performance and advances to nearly every aspect of the chip. Built using third-generation 3-nanometer technology, M5 introduces a next-generation 10-core GPU architecture with a Neural Accelerator in each core, enabling GPU-based AI workloads to run dramatically faster, with over 4x the peak GPU compute performance compared to M4.1 The GPU also offers enhanced graphics capabilities and third-generation ray tracing that combined deliver a graphics performance that is up to 45 percent higher than M4.1 M5 features the world’s fastest performance core, with up to a 10-core CPU made up of six efficiency cores and up to four performance cores.2 Together, they deliver up to 15 percent faster multithreaded performance over M4.1 M5 also features an improved 16-core Neural Engine, a powerful media engine, and a nearly 30 percent increase in unified memory bandwidth to 153GB/s
Didn't I read somewhere that Adobe LR has again turned off using the Neural Engine capability? Their Denoise AI on my Sony A1 files usually takes 30 sec, on my M1 Max MBP, argh!!! Whatever the reason, one would think that taking full advantage of the latest M5 capabilities ought to significantly improve Denoise processing time. At this point I'm waiting for the M5 Max in spring and hopefully Adobe getting their Denoise aligned to fully work with it. 🤞
