New pc lightroom cc denoise performance

BigDaveE

Active member
Messages
71
Reaction score
26
My current 3 year old ryzen based laptop takes about 9mins to denoise a 60mpix sony raw file in lightroom classic using AI denoise.

I am going to build a new desktop pc with the following spec

Intel core i7 14700k

64gb ddr5 ram

2tb internal ssd @ 6000mhz

Rtx 4060ti graphics card

My question is, how long would a similar denoise operation take on this new pc?

I'm asking if there is anyone that has a similar speced machine and what their denoise performance is like in lightroom classic.

Thanks in advance.
 
Last edited:
I'm not an expert, but if I compare your possible results relative to my computer, I'd say denoising should be pretty fast. I have an HP TE01-1020 (i7 core) and have just replaced its integrated card with a RTX-4060. Running Photolab7's DeepPrime before the new card, would take around 45-60 sec on a 22MB image. It now takes about 10-15 sec. As the RTX-4060Ti is faster than the RTX-4060, I think you will be happy.
 
The denoise performance depends on the GPU and not the CPU. Therefore, you should consider upgrading the graphic card of your system. A thee years old Ryzen CPU is usually more than enough for a smooth Photoshop/Lightroom performance.

Cheers,

Ivan

----------------
 
I am reading this using an ancient Skylake i7 with an AMD 580 GPU.

It took barely 30 seconds to denoise a 24mp 16 bit raw image. Actually 28.

That is about the same time it takes on a Ryzen 5800/nvidia 3060 desktop, the other one available at this moment so I ran the same file. My experience with other desktops with a bit higher tiered GPUs is in the same ballpark timewise. Not exactly the same, not hugely different. I have been surprised that times in Topaz Sharpen AI, while immensely faster with any discrete GPU, seem about the same regardless of GPU. Something to consider when reaching for your credit card.

Integrated graphics in a laptop like the OP presumably has are indeed a limiting factor in GPU offloaded processing. Less processing units and slower external system RAM can only do what they do.

If the OP wants to stay with a laptop then getting one with a discrete GPU should cut processing times significantly but a desktop will be somewhat faster because, you know, gobs more electricity juicing the silicon. Usually.

One other piece of unsolicited advice: if you really are going with 64gb of DDR5 RAM I hope it was on sale, has reasonable latencies and occupies only two sticks.
 
Yup, the 64gb ram is fast and 32gb per stick.

I have seen lots of people saying that they can get 20secs for lightroom denoise, But not what the system configuration is or even the file type and size.
 
I could definitely see something close to that being possible with a 4060 Ti, the issue is that there's no set tests out there from a big tech outlet or Puget Systems for GPU performance in Lightroom AI with a standard workload across multiple cards.

The closest I can do is an 80MP image in Lightroom Classic from an Olympus camera with a 3060 Ti.

There's a bit of a start up/wind down process it seems and a single image was 41 seconds and doing 5 averaged out to 36.2 each.

The confusing part is how much faster are NVIDIA 40 series tensor cores in this specific task. And then how much if that offset by NVIDIA reducing the 3060 Ti's 152 cores to the 4060Ti's 136.

So there's a bunch of variables at play but hopefully this can give you a rough ballpark guestimate.
 
Lightroom Classic denoise of 60 MP Sony file with 4060Ti is aboout 25 sec.

See table in this link: https://www.fredmiranda.com/forum/topic/1804640/17
Wow, what a great link. Thanks so much!

Well i built my PC yesterday...Core i7-14700k, 64gb 6000mhz RAM, 2tb fast SSD, RTX 4060ti

My denoise set to 50%. Using a A7R5 raw file taken at 5000 ISO it took 27 seconds...exactly as the table in Fred Miranda predicted!

Running several more tests and its very predictable 26-28 seconds. I'm very happy with that!

Thanks
 
Lightroom Classic denoise of 60 MP Sony file with 4060Ti is aboout 25 sec.

See table in this link: https://www.fredmiranda.com/forum/topic/1804640/17
Thanks for the link. I'm also looking for a gpu for my new PC for Photoshop AI tasks like denoise. Does the amount of VRAM matter?

Is there any difference in speed between a 3060 12 GB version vs a 4060 Ti 8 gb? There's no mention on the impact of the amount VRAM.

Thanks.
 
Last edited:
Sorry I don't know if the VRAM matters it is apparently all about the number of tensor cores...what ever they are :-)

My 4060ti Card has 8gb VRAM. I bought it for the price/performance tradeoff. If you don't have a budget then best go with the 4090 !
 
Sorry I don't know if the VRAM matters it is apparently all about the number of tensor cores...what ever they are :-)

My 4060ti Card has 8gb VRAM. I bought it for the price/performance tradeoff. If you don't have a budget then best go with the 4090 !
For anyone who doesn't spot the joke:

A 4090 is 3+ slots wide (for the air cooled versions), eats 450-600W, and is so heavy that a support for the card is pretty much mandatory unless it's mounted vertically.

nVidia recommends at least an 850W PSU. That would be pretty marginal for one of the 600W cards.

I don't see any at NewEgg for what was the $1.6k MSRP for some. Still cheap compared to (say) a Canon big prime.
 
Sorry I don't know if the VRAM matters it is apparently all about the number of tensor cores...what ever they are :-)
Lightroom Denoise uses the Nvidia Tensor cores aka AI processor aka NPU (neural processing unit) aka Apple Neural Engine aka Ryzen AI processor, etc.

Eric Chan of Adobe wrote this Denoise AI paper that was published on 2023/4/18:

https://blog.adobe.com/en/publish/2023/04/18/denoise-demystified

He wrote this:

Finally, we built our machine learning models to take full advantage of the latest platform technologies, including NVIDIA’s TensorCores and the Apple Neural Engine. Using these technologies enables our models to run faster on modern hardware.

For those who do not know about the Apple Silicon GPU which is similar to Nvidia Cuda cores and the Apple Neural Engine which is similar to Nvidia Tensor cores then see this:

What are Apple's GPU cores?

https://www.dpreview.com/forums/post/67117906

Apple Neural Engine - Nvidia CUDA Cores vs. Tensor Cores: What's the Difference?

https://www.dpreview.com/forums/post/67118578

Note that for the last 1.5 years an Apple Neural Engine bug introduced into MacOS has not been fixed by Apple so still Adobe cannot use the Neural Engine. Instead on Apple Silicon Macs it is forced to use the much slower GPU for Denoise.
 
Last edited:
Sorry I don't know if the VRAM matters it is apparently all about the number of tensor cores...what ever they are :-)

My 4060ti Card has 8gb VRAM. I bought it for the price/performance tradeoff. If you don't have a budget then best go with the 4090 !
Thanks for the reply. I do have a budget, I'm also looking at second hand cards ;).

I've also read about the tensor cores (but there are also different generations of these cores). I haven't read any information about the differences between 8 gb, 10 gb, 12 gb or 16 gb versions. Personally I'm looking at the 3060 12 GB, 3070 8 GB, 4060 Ti 8 GB versions. Not sure what the best option is for Photoshop and maybe (in the future) some occasional 4k editing.
 
We have on route some laptops with the 4070 mobile chipset, 64Gb of 5400 RAM, and i9-13980HX13th Gen HX55 processor.

I will update once I've done some tests with LR, PS and Topaz but whilst searching for machines Asus do have a machine with a 4090 mobile chip. In the two Asus laptops with that graphics configuration they have a limit of 175W (65 for the processor).

I wonder if other variations may be available for desktop machines that are less power hungry / take up less room.
 
We have on route some laptops with the 4070 mobile chipset, 64Gb of 5400 RAM, and i9-13980HX13th Gen HX55 processor.

I will update once I've done some tests with LR, PS and Topaz but whilst searching for machines Asus do have a machine with a 4090 mobile chip. In the two Asus laptops with that graphics configuration they have a limit of 175W (65 for the processor).

I wonder if other variations may be available for desktop machines that are less power hungry / take up less room.
The differences in the standard RTX 4090 cards are things like fans, how much overclocking, how the manufacturer supports the card, number of fans/cooling and modified connections for some of the newest ones. They are all very big/heavy, need support and officially draw 450 watts, though it can be more in spikes. A RTX 4090D was recently rumored with reduced capacity for the Chinese market to avoid the restrictions on the regular 4090 for export to China.
 
We have on route some laptops with the 4070 mobile chipset, 64Gb of 5400 RAM, and i9-13980HX13th Gen HX55 processor.

I will update once I've done some tests with LR, PS and Topaz but whilst searching for machines Asus do have a machine with a 4090 mobile chip. In the two Asus laptops with that graphics configuration they have a limit of 175W (65 for the processor).

I wonder if other variations may be available for desktop machines that are less power hungry / take up less room.
The differences in the standard RTX 4090 cards are things like fans, how much overclocking, how the manufacturer supports the card, number of fans/cooling and modified connections for some of the newest ones. They are all very big/heavy, need support and officially draw 450 watts, though it can be more in spikes. A RTX 4090D was recently rumored with reduced capacity for the Chinese market to avoid the restrictions on the regular 4090 for export to China.
 
We have on route some laptops with the 4070 mobile chipset, 64Gb of 5400 RAM, and i9-13980HX13th Gen HX55 processor.

I will update once I've done some tests with LR, PS and Topaz but whilst searching for machines Asus do have a machine with a 4090 mobile chip. In the two Asus laptops with that graphics configuration they have a limit of 175W (65 for the processor).

I wonder if other variations may be available for desktop machines that are less power hungry / take up less room.
The differences in the standard RTX 4090 cards are things like fans, how much overclocking, how the manufacturer supports the card, number of fans/cooling and modified connections for some of the newest ones. They are all very big/heavy, need support and officially draw 450 watts, though it can be more in spikes. A RTX 4090D was recently rumored with reduced capacity for the Chinese market to avoid the restrictions on the regular 4090 for export to China.
Nvidia do have a mildly confusing naming system when it comes to comparing cards and mobile chips. I don't know what performance to expect from the 4090M for LR Denoise but perhaps more energy efficient than it's big desktop cousin.
The 4090M is more like the desktop RTX 4080 with a few things more limited with power draws from 120 to 150 watts. Cooling would be a significant problem if it could draw the 450+ watts of the desktop RTX 4090. The desktop RTX 4080 can draw up to 320watts.
 
We have on route some laptops with the 4070 mobile chipset, 64Gb of 5400 RAM, and i9-13980HX13th Gen HX55 processor.
The Lightroom denoise function seems to depend only on tensor cores on Nvidia chips. The RTX 4070 mobile has 144 tensor cores so I would guess the denoising of a 60 MP files takes about 25 sec.
I will update once I've done some tests with LR, PS and Topaz but whilst searching for machines Asus do have a machine with a 4090 mobile chip. In the two Asus laptops with that graphics configuration they have a limit of 175W (65 for the processor).
The RTX 4090 mobile with its 302 tensor cores should take about 14 sec for a 60MP file.
I wonder if other variations may be available for desktop machines that are less power hungry / take up less room.
 

Keyboard shortcuts

Back
Top