1.9 slow

[Mark Ingram]

Just now, Petryk said:

I understand, but with OpenCL AF enabled, it should use the GPU core and the CPU, and it only uses the CPU and 50% of it, so with openCL enabled performance drops,it should be the other way around. 

Not really, you would notice sections of your document rendering much slower in comparison to others if you split the work between CPU and GPU. You would end up waiting for the slowest device each time. e.g. Render 90% of the document on GPU quickly, the remaining 10% on CPU takes 20X longer (depending on GPU/CPU combo). So now most of your document is rendered, but you still have to wait for the remaining 10% on the CPU to finish rendering.

[Petryk]

3 minutes ago, Mark Ingram said:

Not really, you would notice sections of your document rendering much slower in comparison to others if you split the work between CPU and GPU. You would end up waiting for the slowest device each time. e.g. Render 90% of the document on GPU quickly, the remaining 10% on CPU takes 20X longer (depending on GPU/CPU combo). So now most of your document is rendered, but you still have to wait for the remaining 10% on the CPU to finish rendering.

Ok I understand how this work, but still in My opinion, this function not work properly.  For now i disable OpenCL. Maybe in new update this function will work fine.

[Mark Ingram]

57 minutes ago, Petryk said:

Ok I understand how this work, but still in My opinion, this function not work properly.  For now i disable OpenCL. Maybe in new update this function will work fine.

It looks like there is definitely a problem according to your original report, as your GPU should be ~40X faster than your CPU. We will investigate. What I was suggesting is that looking at CPU / GPU usage can't always be relied on as a way of identifying problems.

[Petryk]

22 minutes ago, Mark Ingram said:

It looks like there is definitely a problem according to your original report, as your GPU should be ~40X faster than your CPU. We will investigate. What I was suggesting is that looking at CPU / GPU usage can't always be relied on as a way of identifying problems.

If you need some more information from Me: logs, system configuration, maybe some tests, write to me. I try to help to resolve this problem.

[phil_k]

I had an issue on a previous system that showed low gpu usage with hardware acceleration set on (not affinity photo). The issue was my gpu AMD 5500XT was significantly reduced due to its 8 bit pcie 4.0 bus and my motherboards pcie 3.0 bus. My previous generation gpu card used a 16 bit bus but some of the newer ones are going with 8 bit bus at the higher pcie clock rate.  Does the affinity benchmark take into consideration the transfer of large amounts of data that is typical in a photo that may be slowing down some systems? Although I have issues with affinity photo rendering square artifacts I followed the example in the hardware acceleration tutorial using my own image and it appears to be similar in performance with my gpu reaching 25%. I have AMD Ryzen 3700 cpu and AMD 5500XT gpu.

[n9neinchnail]

Also having very poor performance. If I use a paintbrush and quickly draw some lines, I can literally sit there and watch as the lines are drawn. My graphics cards doesn't give me the option of hardware acceleration so that is disabled by default. It's almost unusable and I'm afraid I have to go back to 1.8

[MrSwitch]

I observe the same issues: Overall, Photo 1.9's whole processing pipeline and GUI feel slow and jumpy. Be it using HDR-processing or using different brushes. With the GUI I have the impression, that functions need a few seconds, before e.g. a brush stroke appears in the image. From then on it works better.

Deactivating Open-CL accelerates solves the issues.

My hardware:

Computer:      MSI MS-7B78
CPU:           AMD Ryzen 7 2700X (Pinnacle Ridge, PiR-B2)
               3700 MHz (37.00x100.0) @ 4313 MHz (43.50x99.2)
Motherboard:   MSI X470 GAMING PRO CARBON (MS-7B78)
BIOS:          2.E0, 06/10/2020
Chipset:       AMD X470 (Low-Power Promontory PROM28.A)
Memory:        32768 MBytes @ 1322 MHz, 20-19-19-43

Graphics:      Sapphire RX 5700 Pulse
               AMD Radeon RX 5700, 8192 MB GDDR6 SDRAM

OS:            Microsoft Windows 10 Professional (x64) Build 19042.804 (20H2)

 

Regards

André

[Mark Ingram]

2 minutes ago, MrSwitch said:

I observe the same issues: Overall, Photo 1.9's whole processing pipeline and GUI feel slow and jumpy.

Graphics:      Sapphire RX 5700 Pulse
               AMD Radeon RX 5700, 8192 MB GDDR6 SDRAM

Thanks for the report - we've discovered a problem with AMD Radeon RX 5000 series and later. For now we recommend disabling hardware acceleration.

[christhes]

I had the same problem. Working on a Surface Pro 5, the Intel integrated chip is allegedly compatible but in practice it doesn't work, possibly because Microsoft uses a custom version of the Intel driver which is a little bit behind the times (I've tried updating to Intel's own driver and it won't let me).

Interestingly, it only exhibited this behavior in Photo before I turned OpenCL off - in Designer it's still turned on and I haven't seen any slowness or instability.