About the Hardware Acceleration for windows version

[Aftemplate]

How is it achieved? CUDA or OpenCL and OpenCV?

Thanks.

[Mark Ingram]

8 hours ago, Ωmega said:

How is it achieved? CUDA or OpenCL and OpenCV?

Thanks.

The Windows versions don't currently support hardware acceleration.

[Aftemplate]

1 hour ago, Mark Ingram said:

Windows版本目前不支持硬件加速。

Yes, I understand. I heard that you are already doing this work. How do you plan to support it in the future? By cuda or something else? Thanks.

[Mark Ingram]

34 minutes ago, Ωmega said:

Yes, I understand. I heard that you are already doing this work. How do you plan to support it in the future? By cuda or something else? Thanks.

It won't be CUDA, as that is NVIDIA only.

[R34V3r]

On 12/12/2018 at 1:02 PM, Mark Ingram said:

It won't be CUDA, as that is NVIDIA only.

Is hardware acceleration planned for 1.7 like on the Mac version?

[ashf]

This article mentions that hardware acceleration is coming to Windows soon.
Is this already available on 1.7.1 or coming to 1.7.2 or later?
https://www.dpreview.com/news/2037962810/affinity-photo-1-7-brings-hardware-acceleration-hdr-support-and-more

[fde101]

14 hours ago, ashf said:

Is this already available on 1.7.1

no

[RyanJEC]

Hardware/GPU Acceleration on Windows should be top priority for 1.9 update in all the apps.

[Aftemplate]

As of the latest 1.8.4beta, it is still not supported.

[Aftemplate]

MacOS Affinity GPU accelerated development took less than 1 year. And it took you 2 years to transplant it, and it hasn't been done yet.
This feature has a significant impact on performance up to 10 times. You should do your best to achieve it, not to make it go away.
Thank efforts for Mark Ingram.

Windows has the following 4 GPU acceleration Api: 
OpenCL  OpenCV  Vulcan  CUDA.
MacOS: only has Metal. (The rest was declared deprecated by Apple)
Windows has more sufficient prerequisites for GPU acceleration.

@Alex_M  @MancDan  @Frozen Death Knight  @Zekez  @sbp  @R34V3r  @ashf  @fde101  @RyanJEC    

The current GPU acceleration appears to have been lost...

[fde101]

OpenCL driver support on Windows is inconsistent and probably shouldn't be relied upon.

OpenCV is a computer vision library, not a compute acceleration API, so that isn't really in the same category as the others.

CUDA is proprietary to Nvidia and best avoided as it would not work on GPUs from other companies.

So no, you only really have one GPU compute API on Windows too, or at least only one that can be reasonably expected to be available and usable on the majority of systems.

There are wrapper implementations which implement Vulkan on the Mac over top of Metal - meaning that it doesn't really offer anything that you can't do with Metal, so Vulkan on Windows doesn't really have any technical benefit here compared to Metal on the Mac.  It is at best a draw.

[Aftemplate]

9 hours ago, fde101 said:

OpenCL driver support on Windows is inconsistent and probably shouldn't be relied upon.

OpenCV is a computer vision library, not a compute acceleration API, so that isn't really in the same category as the others.

CUDA is proprietary to Nvidia and best avoided as it would not work on GPUs from other companies.

So no, you only really have one GPU compute API on Windows too, or at least only one that can be reasonably expected to be available and usable on the majority of systems.

There are wrapper implementations which implement Vulkan on the Mac over top of Metal - meaning that it doesn't really offer anything that you can't do with Metal, so Vulkan on Windows doesn't really have any technical benefit here compared to Metal on the Mac.  It is at best a draw.

https://github.com/YukihoAA/waifu2x_snowshell

This program uses OpenCV or Vulkan or CUDA for GPU computing acceleration (up to 10 times)
Blender and many programs are compatible with Nvidia and AMD using OpenCL and work well.
MacOS Vulcan is realized above metal. I know this.
I know a lot about these aspects. You don't need to prompt me. (I also know that OpenCV is a visual library, But acceleration)
As you said, a "draw". But the current lack of GPU acceleration is clearly unfair.
I agree with you: CUDA should be avoided. only.
Vulcan is the best choice (compatible with possible future Linux versions)
And the calculation acceleration performance is more powerful. (But I think Mark Ingram may lack the ability to control this complex low-level API)

@fde101  ☺️

Edited July 9, 2020 by Aftemplate
Add strikethrough.

[Mark Ingram]

36 minutes ago, Aftemplate said:

But I think Mark Ingram may lack the ability to control this complex low-level API

You are very much wrong. 

I understand you would like to see GPU support in the Windows products, so would I, but spamming, tagging, and calling me out on the forum are not the best methods for motivating people.

[Aftemplate]

9 minutes ago, Mark Ingram said:

You are very much wrong. 

I understand you would like to see GPU support in the Windows products, so would I, but spamming, tagging, and calling me out on the forum are not the best methods for motivating people.

I honestly admit my mistake, please allow me to apologize to you. very sorry.
Please listen to me. . .
I do this only because. You seem to be evading the issue of GPU acceleration. Maybe my guess is incorrect.
I want to know the current progress, it's that simple. You only need to disclose a number, if convenient. (Optional)
What you may not know is that I use the MacOS version instead of the Windows version.
My friends are using the Windows version. I hope the affinity is better, this is my wish.
I don't want windows to miss features that are so important to performance.
Finally, I sincerely apologize to you again.

  @Mark Ingram   

[Mark Ingram]

Don't worry, it's high on my priority list, there's no way I could forget about it!

[Aftemplate]

9 hours ago, Mark Ingram said:

You are very much wrong. 

I understand you would like to see GPU support in the Windows products, so would I, but spamming, tagging, and calling me out on the forum are not the best methods for motivating people.

I'm not motivating and urging you (Not to be faster)
You can complete this function at any future time.
I just want to know the latest progress about it. That's it.

[Aftemplate]

9 minutes ago, Mark Ingram said:

Don't worry, it's high on my priority list, there's no way I could forget about it!

You're the best! Thank you for not letting it absent.

[Aftemplate]

I'll stop focusing on this GPU issue.

(it's definitely going to be supported sometime in the future! )

(Efforts by Mark Ingram, Chief Developer of Windows affinity)

[grunnvms]

I've been reading this thread with interest, and my first question is what kind of Affinity work would benefit from hardware acceleration. I only do simple work with Affinity products, and I've never come across a moment that It thought, well this is going too slow. Hence my question.

And then my second question, isn't hardware acceleration provided by DirectX on Windows, in other words you don't need to know too much about the underlying hardware?

 

[Aftemplate]

14 hours ago, grunnvms said:

I've been reading this thread with interest, and my first question is what kind of Affinity work would benefit from hardware acceleration. I only do simple work with Affinity products, and I've never come across a moment that It thought, well this is going too slow. Hence my question.

And then my second question, isn't hardware acceleration provided by DirectX on Windows, in other words you don't need to know too much about the underlying hardware?

 

1. All
2. No

[Novak]

Well we know now it's OpenCL. Does anyone know how well it performs vs metal and vs non accelerated version? 

[grunnvms]

On 7/30/2020 at 1:45 AM, Aftemplate said:

1. All
2. No

Well, directx does hardware acceleration, bit only in decoding, and not in encoding so it seems.

[Aftemplate]

On 9/18/2020 at 12:09 AM, Novak said:

Well we know now it's OpenCL. Does anyone know how well it performs vs metal and vs non accelerated version? 

I've already replied to you, In another topic.
Windows Acceleration is garbage and is currently 2 times slower.

[Aftemplate]

Use the same hardware and affinity version.

[CoEl]

Re: Glitches in Photo exports with Open CL

If it is not working *yet*, why is the default to have Open CL hardware acceleration ticked *on*? It would make more sense for it to be off by default, and for a dialogue box to inform the user that it's off and why when they first open in package after installation.

This is absolutely baffling to me. Not only does it cause me problems (I'm an infrequent user, and by the look of things updates must be re-ticking the box because I have to Google and remind myself what the problem is when I know I've fixed it before), but I also have friends to whom I have recommended Affinity products who complain to me that Photo leaves glitches in their files.

Does the company not understand the terrible impression this gives to customers, and then potential customers by word-of-mouth? I understand the problem may lie with Windows rather than Serif, but few users will find that out, or accept it. All they will see is that even with the latest and greatest Nvidia cards, they have to turn off hardware acceleration, and that's if they find their way to the solution. If not, they will be left with the impression that Photo is unusable.