Jump to content
You must now use your email address to sign in [click for more info] ×

AMD Radeon RX Hardware Acceleration


Recommended Posts

Just my last 2 cents on this topic.

Please read this my intervention. I have not any clue about the influences of these parameters into this context, but unexpectedly i solved the freezing long lasting moments.

I hope it can help you as well.

Link to comment
Share on other sites

I hope nvidia get the memo and drop the 1630 to something like 80 euros so I can get it for affinity or intel release the cards and be good or something idk.

Current Workstation:
CPU: AMD Ryzen 5 5500 - MOBO: Asus B450 - RAM: 16GB DDR4 2667Mhz - GPU: AMD Radeon 7850 1GB
NVMe SSD: Crusial P3 1TB M.2 -  SSD: Samsung Evo 850 256GB  - PSU: XFX TS450 - OS: Win10

Link to comment
Share on other sites

6 minutes ago, GenewalDesign said:

AMD is providing the driver for the cards we use. The driver is bugged, so Affinity had to disable HA. If AMD doesn't fix the driver, it's solely AMD's problem.

I have a whole host of other GPU accelerated applications that run just fine on AMDs cards. Serif chose to use that API, they were not forced. After more than a year they still have not chosen an API that works like everyone else has.

The problem is firmly in Serif's court far more than AMDs.

 

Link to comment
Share on other sites

2 minutes ago, ziplock9000 said:

I have a whole host of other GPU accelerated applications that run just fine on AMDs cards. Serif chose to use that API, they were not forced. After more than a year they still have not chosen an API that works like everyone else has.

The problem is firmly in Serif's court far more than AMDs.

 

I do understand your point but I must imagine switching from OpenCL to any other API must be very long to code, am I wrong?

Link to comment
Share on other sites

4 minutes ago, ziplock9000 said:

I have a whole host of other GPU accelerated applications that run just fine on AMDs cards. Serif chose to use that API, they were not forced. After more than a year they still have not chosen an API that works like everyone else has.

The problem is firmly in Serif's court far more than AMDs.

 

Care to explain a bit in detail? I though the only way to have a GPU ventor agnostic GPGPU is to use openCL?

What other options are out there? Can Vulkan and DirectX do the same?

We may not get the acceleration but at least we can learn something new.

 

 

  

Just now, GenewalDesign said:

I do understand your point but I must imagine switching from OpenCL to any other API must be very long to code, am I wrong?

 

it is possible that it is true, but since they already work with metal for apple maybe they have not hardcoded openCL that much in affinities? I just hope so at least.

Current Workstation:
CPU: AMD Ryzen 5 5500 - MOBO: Asus B450 - RAM: 16GB DDR4 2667Mhz - GPU: AMD Radeon 7850 1GB
NVMe SSD: Crusial P3 1TB M.2 -  SSD: Samsung Evo 850 256GB  - PSU: XFX TS450 - OS: Win10

Link to comment
Share on other sites

59 minutes ago, GenewalDesign said:

I do understand your point but I must imagine switching from OpenCL to any other API must be very long to code, am I wrong?

Adobe products use OpenCL on AMD GPUs, they have zero issues running on there. DaVinci Resolve uses OpenCL, but surprisingly it doesn't have any issues on AMD either.

Blender switched from OpenCL to HIP on AMD GPUs, because this API is more reliable, versatile, faster, future proof and very similar to CUDA. Also OpenCL as a whole has really big problems in general, it's hard and annoying to implement for hardware manufacturers, that's why everyone sticks to 2.0. OpenCL 3.0 is a big reset which makes most features optional because Khronos (standardisation organisation) has realized 2.0+ was a big mistake.

Serif has 3 options to handle this:

- fix the OpenCL bug like all other creative apps, make use of OpenCL 3.0's feature query function in the future.

- think straigh forward, future proof and switch to a better API (HIP and CUDA)

- or just do nothing and blame AMD for not supporting a broken API

Obviously the latter option does not fix anything.

Link to comment
Share on other sites

25 minutes ago, Nox said:

Adobe products use OpenCL on AMD GPUs, they have zero issues running on there. DaVinci Resolve uses OpenCL, but surprisingly it doesn't have any issues on AMD either.

Blender switched from OpenCL to HIP on AMD GPUs, because this API is more reliable, versatile, faster, future proof and very similar to CUDA. Also OpenCL as a whole has really big problems in general, it's hard and annoying to implement for hardware manufacturers, that's why everyone sticks to 2.0. OpenCL 3.0 is a big reset which makes most features optional because Khronos (standardisation organisation) has realized 2.0+ was a big mistake.

Serif has 3 options to handle this:

- fix the OpenCL bug like all other creative apps, make use of OpenCL 3.0's feature query function in the future.

- think straigh forward, future proof and switch to a better API (HIP and CUDA)

- or just do nothing and blame AMD for not supporting a broken API

Obviously the latter option does not fix anything.

Thank you!

Link to comment
Share on other sites

15 hours ago, Nox said:

Adobe products use OpenCL on AMD GPUs, they have zero issues running on there. DaVinci Resolve uses OpenCL, but surprisingly it doesn't have any issues on AMD either.

Blender switched from OpenCL to HIP on AMD GPUs, because this API is more reliable, versatile, faster, future proof and very similar to CUDA. Also OpenCL as a whole has really big problems in general, it's hard and annoying to implement for hardware manufacturers, that's why everyone sticks to 2.0. OpenCL 3.0 is a big reset which makes most features optional because Khronos (standardisation organisation) has realized 2.0+ was a big mistake.

Serif has 3 options to handle this:

- fix the OpenCL bug like all other creative apps, make use of OpenCL 3.0's feature query function in the future.

- think straigh forward, future proof and switch to a better API (HIP and CUDA)

- or just do nothing and blame AMD for not supporting a broken API

Obviously the latter option does not fix anything.

Well, it's pretty summed up. I guess this problem will be resolved with 2.0. We're not out of the woods yet ¯\_(ツ)_/¯ 

Link to comment
Share on other sites

  • Staff
On 8/29/2022 at 3:21 PM, Nox said:

Adobe products use OpenCL on AMD GPUs, they have zero issues running on there. DaVinci Resolve uses OpenCL, but surprisingly it doesn't have any issues on AMD either.

Just to clarify this slightly, OpenCL is not utilised widespread across an entire app and is mainly used for specific functionality—e.g. Photoshop uses hardware acceleration for a small subset of its functionality like the blur gallery, camera raw development, neural filters and some selection tools (source: https://helpx.adobe.com/uk/photoshop/kb/photoshop-cc-gpu-card-faq.html). In fact, on that page OpenCL appears to be used very sparsely:

Quote

Use OpenCL: Enable to accelerate the new Blur Gallery filters, Smart Sharpen, Select Focus Area, or Image Size with Preserve Details selected (Note: OpenCL is available only on newer graphics cards that support OpenCL v1.1 or later.)

Affinity Photo leverages hardware acceleration for practically all raster operations in the software—from basic resampling to expensive live filter compositing—requiring many different permutations of kernels to be compiled on the fly. Every time you add a new live filter or apply a destructive filter, paint with a brush tool, add an adjustment layer or perform many other operations, these all need to load kernels for hardware acceleration. With the majority of GPUs and driver combinations, this kernel compilation time is more or less negligible, but as Mark has posted previously with benchmarks, the Navi architecture with its current drivers appears to be problematic here. Any kind of comparison to Photoshop's OpenCL implementation is not appropriate, as the two apps use it very differently.

I previously had a 5700XT, and loading a single RAW image was painfully slow because a number of kernels needed to be compiled simultaneously (for all the standard adjustments you can apply in the Develop Persona). We're talking almost a minute from loading in a RAW file to being able to move sliders around. The previous Polaris generation of cards are, to my understanding, absolutely fine with OpenCL kernel compilation.

Product Expert (Affinity Photo) & Product Expert Team Leader

@JamesR_Affinity for tutorial sneak peeks and more
Official Affinity Photo tutorials

Link to comment
Share on other sites

I purposely nixed AMD as possibility of GPU upgrade because of the driver issues. Not acceptable. Unfortunately, even on gaming side they are known for issues.

Microsoft Windows 10 Home (Build 19045)
AMD Ryzen 7 5800X @ 3.8Ghz (-30 all core +200mhz PBO); Mobo: Asus X470 Prime Pro
32GB DDR4 (3600Mhz); EVGA NVIDIA GeForce GTX 3080 X3C Ultra 12GB
Monitor 1 4K @ 125% due to a bug
Monitor 2 4K @ 150%
Monitor 3 (as needed) 1080p @ 100%

WACOM Intuos4 Large; X-rite i1Display Pro; NIKON D5600 DSLR

Link to comment
Share on other sites

20 minutes ago, debraspicher said:

I purposely nixed AMD as possibility of GPU upgrade because of the driver issues. Not acceptable. Unfortunately, even on gaming side they are known for issues.

on gaming side they are fine, this is what hurts actually. I enjoy my games with the 5500XT for the last 2 years with zero issues, but this thing with the openCL is biting bad right now.

Current Workstation:
CPU: AMD Ryzen 5 5500 - MOBO: Asus B450 - RAM: 16GB DDR4 2667Mhz - GPU: AMD Radeon 7850 1GB
NVMe SSD: Crusial P3 1TB M.2 -  SSD: Samsung Evo 850 256GB  - PSU: XFX TS450 - OS: Win10

Link to comment
Share on other sites

28 minutes ago, nitro912gr said:

on gaming side they are fine, this is what hurts actually. I enjoy my games with the 5500XT for the last 2 years with zero issues, but this thing with the openCL is biting bad right now.

That's what I have, 5700XT Red Devil and I've had no issue with the many games I've played over the last ~2 year or with any GPU accelerated applications (of which I use a lot as a game developer). Affinity Designer is the ONLY software that has issues for me unfortunately.

Link to comment
Share on other sites

  • Staff
17 minutes ago, ziplock9000 said:

That's what I have, 5700XT Red Devil and I've had no issue with the many games I've played over the last ~2 year or with any GPU accelerated applications (of which I use a lot as a game developer). Affinity Designer is the ONLY software that has issues for me unfortunately.

What issues are you having with Affinity Designer? OpenCL is only used to accelerate raster paint brushing in Designer, all the vector operations are performed on CPU anyway. As far as I'm aware there are no issues with DirectX view presentation on Navi cards.

Product Expert (Affinity Photo) & Product Expert Team Leader

@JamesR_Affinity for tutorial sneak peeks and more
Official Affinity Photo tutorials

Link to comment
Share on other sites

19 hours ago, James Ritson said:

What issues are you having with Affinity Designer? OpenCL is only used to accelerate raster paint brushing in Designer, all the vector operations are performed on CPU anyway. As far as I'm aware there are no issues with DirectX view presentation on Navi cards.

this can't be truth, I have noticed considerable lag in vector and fonts affinity designer files I work in similar systems in office and home, the only difference is the home one have the 5500XT and as of that work without GPU acceleration.

Aside that the one in office a bit tad slower in CPU and the SSD, while a bit faster in RAM speed (2400 in home, 2666 in office).

Current Workstation:
CPU: AMD Ryzen 5 5500 - MOBO: Asus B450 - RAM: 16GB DDR4 2667Mhz - GPU: AMD Radeon 7850 1GB
NVMe SSD: Crusial P3 1TB M.2 -  SSD: Samsung Evo 850 256GB  - PSU: XFX TS450 - OS: Win10

Link to comment
Share on other sites

  • Staff
3 hours ago, nitro912gr said:

this can't be truth, I have noticed considerable lag in vector and fonts affinity designer files I work in similar systems in office and home, the only difference is the home one have the 5500XT and as of that work without GPU acceleration.

Aside that the one in office a bit tad slower in CPU and the SSD, while a bit faster in RAM speed (2400 in home, 2666 in office).

Vector operations are not processed on the GPU. Perhaps display resolution could be a factor as it will influence the document view rasterisation resolution, e.g. a 1920x1080 display versus 2560x1440 (or even 3840x2160, which requires rasterisation at 4x the spatial resolution of 1080p).

It may be worth benchmarking in Affinity Photo if you have it and comparing the CPU Vector Single and Multi scores (the top two entries) between your home and work machines to see which one is faster in practice?

Product Expert (Affinity Photo) & Product Expert Team Leader

@JamesR_Affinity for tutorial sneak peeks and more
Official Affinity Photo tutorials

Link to comment
Share on other sites

10 minutes ago, James Ritson said:

Vector operations are not processed on the GPU. Perhaps display resolution could be a factor as it will influence the document view rasterisation resolution, e.g. a 1920x1080 display versus 2560x1440 (or even 3840x2160, which requires rasterisation at 4x the spatial resolution of 1080p).

It may be worth benchmarking in Affinity Photo if you have it and comparing the CPU Vector Single and Multi scores (the top two entries) between your home and work machines to see which one is faster in practice?

I dont know if benchmarking photo will help because I don't use it that match to notice any lag in my workflow. But since I literally used file A at office and then open it again in home to continue and it felt unbearable.

But that was back in March and if I remember correctly I mostly had trouble in documents with lot of text, it was some restaurant menu I think. Does the font rendering go through GPU? Because I just opened a complex vector file with the bare minimum of text and it was fine now.

Current Workstation:
CPU: AMD Ryzen 5 5500 - MOBO: Asus B450 - RAM: 16GB DDR4 2667Mhz - GPU: AMD Radeon 7850 1GB
NVMe SSD: Crusial P3 1TB M.2 -  SSD: Samsung Evo 850 256GB  - PSU: XFX TS450 - OS: Win10

Link to comment
Share on other sites

  • Staff
1 minute ago, nitro912gr said:

I dont know if benchmarking photo will help because I don't use it that match to notice any lag in my workflow. But since I literally used file A at office and then open it again in home to continue and it felt unbearable.

But that was back in March and if I remember correctly I mostly had trouble in documents with lot of text, it was some restaurant menu I think. Does the font rendering go through GPU? Because I just opened a complex vector file with the bare minimum of text and it was fine now.

The benchmark in Photo will be representative for all the Affinity apps—the CPU single and multi vector scores illustrate the kind of performance to expect in Designer as it primarily deals with vector operations (which are all CPU-based).

Text will be rendered on CPU as it's vector. I believe layer effects (e.g. drop shadow, gaussian blur) are also rendered on CPU in Designer but I'll double check. There could be other factors at play—it's worth doing the benchmark first on both machines to compare the scores, as it may highlight an area where one is weaker...

Product Expert (Affinity Photo) & Product Expert Team Leader

@JamesR_Affinity for tutorial sneak peeks and more
Official Affinity Photo tutorials

Link to comment
Share on other sites

ok tell me now that I was based on a single event that happened with a single file and all this time I wasn't using affinity at home at all, and I'm going to do some seppuku or something :P because with that in mind I left the comfort of my home many times to go to the office "waiting for a fix in openCL" :(

Current Workstation:
CPU: AMD Ryzen 5 5500 - MOBO: Asus B450 - RAM: 16GB DDR4 2667Mhz - GPU: AMD Radeon 7850 1GB
NVMe SSD: Crusial P3 1TB M.2 -  SSD: Samsung Evo 850 256GB  - PSU: XFX TS450 - OS: Win10

Link to comment
Share on other sites

On 8/30/2022 at 11:50 AM, nitro912gr said:

on gaming side they are fine, this is what hurts actually. I enjoy my games with the 5500XT for the last 2 years with zero issues, but this thing with the openCL is biting bad right now.

To be clear, "on gaming side", I mean that even that crowd is aware the driver issues that can exist or just bad updates. Vast majority of people play games fine on the hardware. The hardware itself is excellent. Just drivers in specific situations/workstation setups and maybe specific graphics settings. My husband was using AMD in his rig until very recently, but moved to Nvidia when I updated mainly because there was a hack around it for pool-mining. The GPU will probably end up in a Linux rig eventually.

Microsoft Windows 10 Home (Build 19045)
AMD Ryzen 7 5800X @ 3.8Ghz (-30 all core +200mhz PBO); Mobo: Asus X470 Prime Pro
32GB DDR4 (3600Mhz); EVGA NVIDIA GeForce GTX 3080 X3C Ultra 12GB
Monitor 1 4K @ 125% due to a bug
Monitor 2 4K @ 150%
Monitor 3 (as needed) 1080p @ 100%

WACOM Intuos4 Large; X-rite i1Display Pro; NIKON D5600 DSLR

Link to comment
Share on other sites

Not sure how much I can contribute to this topic that hasn't yet been—but as of right now OpenCL hardware acceleration is the only source of crashes while using affinity products. Turning it off has taken my complex design sessions from 2-5 minutes max before a crash, to indefinite. I am using an Nvdia RTX3070 as render engine and an AMD Ryzen 5950X. Very much hoping for some news here as hardware accel does significantly help performance and quality processing, yet is un-useable after too long even on a 3070.

Edited by Justin Hawkes
Link to comment
Share on other sites

Hi, I'm using RX6800 and RX570.

I see this issue and I think the AMD driver will have bug or limitation due to Navi.
But Navi2 is too or not?
Benchmark results of RX5000 are not improved enough, but RX6000's one seems to be enough.
Is RX6000 enough or not enough?
If RX6000 is enough (as the default as enabled) or enough for user environment (as the default as disabled, and user selectable enabled), does Serif change this restriction for only RX5000?

Link to comment
Share on other sites

Join the conversation

You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.

Guest
Reply to this topic...

×   Pasted as rich text.   Paste as plain text instead

  Only 75 emoji are allowed.

×   Your link has been automatically embedded.   Display as a link instead

×   Your previous content has been restored.   Clear editor

×   You cannot paste images directly. Upload or insert images from URL.

Loading...
×
×
  • Create New...

Important Information

Terms of Use | Privacy Policy | Guidelines | We have placed cookies on your device to help make this website better. You can adjust your cookie settings, otherwise we'll assume you're okay to continue.