Update.
AMD released a new beta version of the drivers (WIN 10 64 bit vers 2004), the 20.10.1 and works like version 20.8.3.
OpenCL: AMD/ATI GPU 0: AMD Radeon RX 5700 XT 50th Anniversary (driver version 3110.7 (PAL,LC), OpenCL 2.0 AMD-APP device version (3110.7), 8176MB, 8176MB available, 9370 GFLOPS peak).
Reported issue to AMD and I invite anyone, with the same problem, to do the same.
Let's see if they solve the problem to the fury of relationships.

Update.
AMD released a new beta version of the drivers (WIN 10 64 bit vers 20H2), the 20.11.1 and works like version 20.8.3.
OpenCL: AMD/ATI GPU 0: AMD Radeon RX 5700 XT 50th Anniversary (driver version 3110.7 (PAL,LC), OpenCL 2.0 AMD-APP device version (3110.7), 8176MB, 8176MB available, 9370 GFLOPS peak).
Reported issue to AMD and I invite anyone, with the same problem, to do the same.
Let's see if they solve the problem to the fury of relationships.

Update.
AMD has released a new beta version of the drivers the 20.11.2 (WIN 10 64 bit vers 20H2) and WORKS!!!
OpenCL: AMD/ATI GPU 0: AMD Radeon RX 5700 XT 50th Anniversary (driver version 3188.4 (PAL,LC), OpenCL 2.0 AMD-APP device version (3188.4), 8176MB, 8176MB available, 9370 GFLOPS peak).
We hope it will also work with the new version of the app when it is released.

Update.
AMD has released a new beta version of the drivers the 20.11.2 (WIN 10 64 bit vers 20H2) and WORKS!!!
OpenCL: AMD/ATI GPU 0: AMD Radeon RX 5700 XT 50th Anniversary (driver version 3188.4 (PAL,LC), OpenCL 2.0 AMD-APP device version (3188.4), 8176MB, 8176MB available, 9370 GFLOPS peak).
We hope it will also work with the new version of the app when it is released.

That's great news!

Could you point me to the release version notes that show what changed from the previous version? It might help to pinpoint what the problem had been.

I hope they can be useful to you, this is link of the latest version of the 20.11.2 drivers:

https://www.amd.com/en/support/kb/release-notes/rn-rad-win-20-11-2

and this is instead of the previous version 20.11.1:

https://www.amd.com/en/support/kb/release-notes/rn-rad-win-20-11-1

I hope they can be useful to you, this is link of the latest version of the 20.11.2 drivers:

https://www.amd.com/en/support/kb/release-notes/rn-rad-win-20-11-2

and this is instead of the previous version 20.11.1:

https://www.amd.com/en/support/kb/release-notes/rn-rad-win-20-11-1

Thanks!

Unfortunately, the release notes don't explicitly mention anything about a change to their openCL implementation. So we may never know what was problem was.

Update.
AMD has released a new beta version of the 20.11.3 drivers (WIN 10 64 bits vers 20H2) and IT WORKS!!!
OpenCL: AMD/ATI 0 GPU: AMD Radeon RX 5700 XT 50th Anniversary (driver version 3188.4 (PAL,LC), OpenCL 2.0 AMD-APP device version (3188.4), 8176MB, 8176MB available, peak GFLOPS 9370).
We hope it will also work with the new version of the app when it is released.

Update.
AMD has released a new beta version of the 20.12.1 drivers (WIN 10 64 bits vers 20H2) and IT WORKS!!!
OpenCL: AMD/ATI 0 GPU: AMD Radeon RX 5700 XT 50th Anniversary (driver version 3188.4 (PAL,LC), OpenCL 2.0 AMD-APP device version (3188.4), 8176MB, 8176MB available, peak GFLOPS 9370).
We hope it will also work with the new version of the app when it is released.

Update.
AMD has released a new beta version of the 21.1.1 drivers (WIN 10 64 bits vers 20H2) and IT WORKS!!!
OpenCL: AMD/ATI 0 GPU: AMD Radeon RX 5700 XT 50th Anniversary (driver version 3188.4 (PAL,LC), OpenCL 2.0 AMD-APP device version (3188.4), 8176MB, 8176MB available, peak GFLOPS 9370).
We hope it will also work with the new version of the app when it is released.

The new AMD opencl linux version has been deployed (version 4.02). I found it to be twice as fast, but I think that's because my opencl compiler does a better job with the newer stream-lined kernels. The new kernels were only 25% faster with nvidia opencl, so performance seems to be implementation specific.

The windows versions are also ready, but I will wait a few days to make sure there aren't any hiccups with the linux versions.

I only got interested in GPU computing about a year ago, so I am no means an expert. But I'll give you my opinion. I personally prefer the Nvidia cards. I think AMD is focused more on the gaming industry as opposed to scientific computing. AMD only officially supports windows and a couple linux distributions. It took me many hours to finally get my Radeon to actually work and it's still about 30% slower than the GTX 1050 when it should be a little faster.

To answer your question regarding performance, at least on this project, I believe the Nvidia cards are better. My GTX1660 is kickin butt (12x faster than a cpu core). Then comes the GTX1050 which is about 5x faster than a cpu core). Finally, the RX570 is about 4 times faster than a core. Another way to look at it, here are the average run times from the 13x271 batch:
GTX1660: 5.0 min
GTX1050: 10.5 min
RX570: 14 min

There should be no difference in cpu utilization between the app versions.

Hope that answers all your questions.

GPU app using DP (Double precision) run better on AMD GPU.
Milkyway@home is really faster on AMD GPU.

GPU app using DP (Double precision) run better on AMD GPU.
Milkyway@home is really faster on AMD GPU.

But this project uses 64bit integer arithmetic.

The new AMD opencl linux version has been deployed (version 4.02). I found it to be twice as fast, but I think that's because my opencl compiler does a better job with the newer stream-lined kernels. The new kernels were only 25% faster with nvidia opencl, so performance seems to be implementation specific.

The windows versions are also ready, but I will wait a few days to make sure there aren't any hiccups with the linux versions.

I just deployed the Nvidia opencl app for Windows. To limit the number of downloads, I will deploy the AMD version tomorrow.

I just deployed the new AMD opencl app for Windows.

I tried the new Get Decic Fields 4.02 windows_x86_64 (opencl_amd) and the old issues that DON'T WORK are back.
It has been tested with the following versions of drivers that have the same OpenCL driver version:

amd radeon driver 20.11.2 does not work with numberfield (WHQL)
amd radeon driver 21.1.1 does not work with numberfield (optional)
amd radeon driver 21.2.1 does not work with numberfield (optional)
OpenCL: AMD/ATI GPU 0: AMD Radeon RX 5700 XT 50th Anniversary (driver version 3188.4 (PAL,LC), device version OpenCL 2.0 AMD-APP (3188.4), 8176MB, 8176MB available, 9370 GFLOPS peak)

With the driver version 20.11.2:
When wu processing starts, the processing percentage remains at zero for about one minute, then suddenly advances x% and then advances and remains at x.007% blocked after 20-30 seconds, as time progresses and GPU usage driver metrics rise to 99%.
If you pause the processing and then the application, the feed rate remains stationary at x.007% and in the driver metrics the GPU usage remains stuck at 99% for just over a minute after which it returns to 0% as it should be, without system locks or restarts, even if some WU have caused system locks and restarts.
If you close the BOINC client and therefore all processing in progress and after approximately in minutes, When GPU usage driver metrics from 99% drop to 0% and the BOINC client reopens and processing starts again, the percentage and processing time of the application starts from scratch for about one minute, then suddenly advances x% and then remains at x.007% blocked after 20-30 seconds, while time advances and GPU usage driver metrics rise to 99%.
I also noticed that by opening the Microsoft Edge application (with GPU hardware acceleration set), GPU usage driver metrics unlock from 99% down from 1 to 4%, and the application's processing percentage has started to advance again until wu is completed.

With drivers version 21.1.11, 21.2.11:
When wu processing starts, the processing percentage remains at zero for about one minute, then suddenly advances x% and then remains at x.999% blocked after 15 seconds, while the time advances and the GPU usage driver metrics rise to 99%.
If you pause the processing and then the application, the feed rate remains stationary at x.999% and in the driver metrics the GPU usage remains stuck at 99% for just over a minute after which it returns to 0% as it should be, without system locks or restarts, even if some WU have caused system locks and restarts.
If you close the BOINC client and therefore all processing in progress and after approximately in minutes, When GPU usage driver metrics from 99% drop to 0% and the BOINC client reopens and processing starts again, the percentage and processing time of the application starts from scratch for about one minute, then suddenly advances x% and then remains at x.999% blocked after 15 seconds, while time advances and GPU usage driver metrics rise to 99%.
I also noticed that by opening the Microsoft Edge application (with GPU hardware acceleration set), GPU usage driver metrics unlock from 99% down from 1 to 4%, and the application's processing percentage has started to advance again until wu is completed.

And this is all and a good day.

This is interesting. I noticed from your task list you had 20 other tasks that completed successfully, so that tells me these drivers work most of the time.

I am wondering if the problem is with the restarting of the task, possibly with checkpointing, although it looks like the check point flag is not set when restarting. The error on all these tasks is ACCESS_VIOLATION which I believe means memory access (i.e. it tried to access another program's memory space). I only use Linux and have suspended my AMD gpu many times and have never seen this, so this is something inherent with the Windows version. I also looked through the Windows Nvidia results in the database and there are NO access violations, so this appears to be an AMD specific problem (Note that Nvidia and AMD use the exact same opencl code).

Of interest is this task, which although successful, took over 4 hours to complete:
https://numberfields.asu.edu/NumberFields/result.php?resultid=107400116
The RX5700 should never take that long, another hint that something is amiss.

The 20 WU i completed is because I was curating them and that is starting Microsoft Edge, it creates the checkpoint and unlocks the GPU occupied at 99%, otherwise if I do not take care of the WU continues to flow the processing time but the percentage remains blocked, like the GPU at 99%.
That WU that lasted 4 hours is because I let her stay all that time, then I started Edge and he finished it.
Whereas the 4 WU wrong is when they crashed my system.
So I wouldn't say that the new app works well.
I hope I have been clearer in explaining the problem, but if there were any doubts, I am always available for clarification.

The 20 WU i completed is because I was curating them and that is starting Microsoft Edge, it creates the checkpoint and unlocks the GPU occupied at 99%, otherwise if I do not take care of the WU continues to flow the processing time but the percentage remains blocked, like the GPU at 99%.
That WU that lasted 4 hours is because I let her stay all that time, then I started Edge and he finished it.
Whereas the 4 WU wrong is when they crashed my system.
So I wouldn't say that the new app works well.
I hope I have been clearer in explaining the problem, but if there were any doubts, I am always available for clarification.

Oh I see. But I don't understand the part about Microsoft Edge helping. Is it actually the browser that is helping or are you running a web app from within the browser?

Yes, I meant the Microsoft Edge browser, which has in the settings>system>Use hardware acceleration when available "enabled".
When I start it, having that option active, probably requiring help from the GPU, it somehow unlocks it, creates the checkpoint, and completes wu processing.
If I do not use this trick and leave the PC to process the BOINC client all day, as I normally do, without intervening, in the evening I will still find the WU to finish because the percentage of processing will be blocked, while the time, always processing, flows inexorable.
At least the system crashes and then it ends first and the WU will go wrong.
I hope I have cleared up your doubts and explained myself well.
Hello.

Yes, I meant the Microsoft Edge browser, which has in the settings>system>Use hardware acceleration when available "enabled".
When I start it, having that option active, probably requiring help from the GPU, it somehow unlocks it, creates the checkpoint, and completes wu processing.
If I do not use this trick and leave the PC to process the BOINC client all day, as I normally do, without intervening, in the evening I will still find the WU to finish because the percentage of processing will be blocked, while the time, always processing, flows inexorable.
At least the system crashes and then it ends first and the WU will go wrong.
I hope I have cleared up your doubts and explained myself well.
Hello.

This almost sounds like a Windows 10 issue, as if it's blocking computation until hardware acceleration is turned on. I'm not a Windows user anymore, but I found online that hardware acceleration can be disabled in Windows 10. Is there any chance this is disabled and that's what's blocking it until enabled by Edge?

But then again, if the old app used to work with your current Windows configuration, then I'm not sure what's going on.

When you get a chance, could you look at the stderr.txt for a WU that's stuck and paste the output here (before you unstick it). That should give me an idea where in the processing it is before it hangs.