Question GPU upgrade for Topaz Photo AI ?

[jeremy.p.dykes]

Hi,
Looking for advice for an upgrade to reduce image processing time in Topaz Photo AI, particularly DeNoise and Sharpen applications. The build only runs Photoshop and Topaz, no gaming.

I'm struggling to find any decent benchmark data on improvements. There's a good article from Puget Systems:

Topaz AI: CPU & GPU Performance Analysis

Topaz AI is known for how exceptionally well they can iomprove images and videos, but they can be very hardware intensive. Exactly what CPU and GPU will give you the best performance?

www.pugetsystems.com

And I have found some Lightroom Denoise AI:

My Recent GPU Experience (Gigabyte Nvidia RTX 4070 Ti 12GB)

+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ "This is a relatively long discussion. It is worth reading the discussion in sequence, as posted." To assist people returning to this discussion, the following are direct links to frequently used posts...

www.lightroomqueen.com

Neither are extensive enough to show the RTX 4060 vs 3060 TI.

Current specs:
CPU - AMD 5600X
RAM - 16GB
PSU - 650W
HDD - Many!
SSD - Samsung PCIE x4
GPU - GTX 1650
Screen - Dell 4k

Question:
Budget would cover an RTX 4060 or RTX 3060 Ti.

Is there any data that suggests which one is preferable for photo editing?

Is there a significant step up to a RTX 4070 or is it just incremental?

Thanks!

 

[Ralston18]

The starting point is to understand the hardware requirements of the software being run.

Most software companies provide some listing of hardware requirements in the form of "minimal", "recommended", and "best".

You do not want minimal and you do want as much "best" as you can afford.

If the objective is to minimize processing time then you need to identify ways to measure and compare processing times. As objectively and quantifiably as possible. Keeping in mind accuracy and margins of error.

There are a number of tools available to help understand system performance: Task Manager, Resource Monitor, and Process Explorer. All via Windows/Microsoft.

Do some tests on your system using the tools to understand how well (or not) the system is currently performing photo editing tasks.

Then determine how much more of an improvement you require. Quantify: Edit requires X seconds and you wish to reduce that time to a smaller number of Y seconds. May or may not be viable.

Knowing what your system is doing know along with delving into what the system may be capable of doing is important. Could be just a few tweaks required. For example, remove unecessary apps running in the background. Perhaps temporarily while photo editing. Disconnect some of those "many" HDD's might help.

Or you may discover that any given editing tool will only use some certain amount of RAM and, no matter how much more RAM you add - performance will not improve.

Look at all the hardware specs for your system and the GPU's being considered. Read the User Guides/Manuals and pay attention to the fine print and other caveats.

When reading peformance evaluations compare the test environment to your environment. Many tests are conducted under ideal circumstances sometimes favoring the product being touted. You results may be different.

Bottom line: take a closer look at what you have and what might be done with what you have.

Another GPU or other hardware changes may make little or no performance difference. Even if detectable and measurable.

 

[Nighthawk117]

Hi,
Looking for advice for an upgrade to reduce image processing time in Topaz Photo AI, particularly DeNoise and Sharpen applications. The build only runs Photoshop and Topaz, no gaming.

I'm struggling to find any decent benchmark data on improvements. There's a good article from Puget Systems:

Topaz AI: CPU & GPU Performance Analysis

Topaz AI is known for how exceptionally well they can iomprove images and videos, but they can be very hardware intensive. Exactly what CPU and GPU will give you the best performance?

www.pugetsystems.com

And I have found some Lightroom Denoise AI:

My Recent GPU Experience (Gigabyte Nvidia RTX 4070 Ti 12GB)

+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++ "This is a relatively long discussion. It is worth reading the discussion in sequence, as posted." To assist people returning to this discussion, the following are direct links to frequently used posts...

www.lightroomqueen.com

Neither are extensive enough to show the RTX 4060 vs 3060 TI.

Current specs:
CPU - AMD 5600X
RAM - 16GB
PSU - 650W
HDD - Many!
SSD - Samsung PCIE x4
GPU - GTX 1650
Screen - Dell 4k

Question:
Budget would cover an RTX 4060 or RTX 3060 Ti.

Is there any data that suggests which one is preferable for photo editing?

Is there a significant step up to a RTX 4070 or is it just incremental?

Thanks!

I use a number of Topaz products, Gigapixel AI, Sharpen AI, Denoise AI and Video AI but not Photo AI. I can't really answer your question as to what will perform the best between the 3060 Ti and 4060. The former has more shading units and significantly higher memory bandwidth. The latter has greater floating point performance which I believe is important for machine learning workloads.

While the Puget benchmarks don't truly tell you exactly what you need to know, they do demonstrate diminishing returns the higher you go. In Video AI for example, the 4090 shows a 45% increase in performance over the 3060 for 6x the price. Clearly there are limits to how these apps scale.

Based on my experience with these apps, my two cents would be it's unlikely to matter too much, they will both perform similarly. Unless every last drop of performance counts, there's a good chance you will be happy with just an ordinary 12GB 3060.

I think the Sharpen and Denoise apps are more influenced by the CPU than the image upscaling stuff. I find those quite a bit slower than Gigapixel AI for example which is generally quite fast. Which is probably why the 4090 is only 12% faster than the 3060.

I would say if you had both the 4060 and 4070 in front of you, your unlikely to be impressed by the performance uplift of the latter for the additional money your spending. That money would probably be better spent on a faster CPU.

For reference I have a i9 10850K, 32GB RAM and Nvidia 3080.

 

[jeremy.p.dykes]

Thanks Nighthawk - that's a very helpful perspective.

I've gone for a 4060 on the basis it will have better longevity of driver support and hopefully some performance improvements through life.

Hoping it makes a significant difference!

 

[jeremy.p.dykes]

As a quick follow up having installed the new RTX 4060. Processing time has improved c20-fold.

For example Sharpen AI has gone from 20 seconds to less than 1 second on a crop of a 40mp image.

Well worth the upgrade!

 

[Nighthawk117]

As a quick follow up having installed the new RTX 4060. Processing time has improved c20-fold.

For example Sharpen AI has gone from 20 seconds to less than 1 second on a crop of a 40mp image.

Well worth the upgrade!

I can understand why you went with the 4060 as it's just come out. That's one hell of an upgrade, well worth it then. I wasn't really sure to be honest what the improvement would be, I mostly use Gigapixel AI and Video AI so I know those will be much faster. I'm a bit taken back how much of an improvement you've got tbh. There could be multiple factors at play. I used to use the 1060 6GB before I upgraded to the 3080, I compared it to the 1060 3GB at the time. The 6GB version was 3x faster despite only being 10% faster on paper, my conclusion was it must have been the VRAM.

They have been doing a lot of updates to these apps lately so I'm not sure if they are using the Tensor Cores now which your previous card wouldn't have had.

Thanks for sharing on the performance improvement.