iMac 2020, 5700XT - graphical glitch?

[Sheza]

Feel free to ask me to conduct experiments - I received my 5700XT model today but it will be sent back to Apple as soon as a replacement arrives because there is a nick in the aluminium on the back. Not sure how it passed QC....

FWIW so far I saw the white line once during set-up, haven't seen it elsewhere yet but will let you know.

EDIT: That didn't take long - spotted once now just in Safari. It wasn't as severe as the one in setup though.

 

[minimaxir]

I just got my 5700XT iMac and it did have the corners issue and a single white streak (neither of which I would have noticed if not for this thread)

However, after adding a LG Ultrafine 5k, no issues at all.

Given this, and the fact that it does not reproduce in Windows either, I’m 99% sure this is a driver issue. That would likely be fixed with Big Sur at the latest.

 

[kkh]

I just got my 5700XT iMac and it did have the corners issue and a single white streak (neither of which I would have noticed if not for this thread)

However, after adding a LG Ultrafine 5k, no issues at all.

Given this, and the fact that it does not reproduce in Windows either, I’m 99% sure this is a driver issue. That would likely be fixed with Big Sur at the latest.

That is if this is something Apple can do. The drivers are provided by AMD, afaik. People have had various issues with other AMD cards in Windows where the drivers remained unfixed for as long as 6 months. Doesn't mean it's gonna be the same here of course.

 

[filmak]

Catalina will be unforgettable...

 

[TheMelodyman]

Curious. If you decide to return your new purchase because of this glitch and your iMac came from China or Ireland, where do you return it to?

 

[shywizard]

31 people so far have voted and said they see the white line glitch.
That's a lot, considering I was led to believe that only a few people had noticed it.

 

[shywizard]

As I said previously, no horizontal line glitches for me on 2048x1152.
The Apple tech I was talking to found this interesting, and asked me to try 2880x1620 for a day or two.
Will report back my findings.....

To continue my troubleshooting, I've been testing out 2880x1620 for 2 days now and I haven't once seen the horizontal white line glitch. Seems to me that the glitch may only appear if the Display Resolution is set to DEFAULT.

Interesting isn't it?

Would love if some other people would also test these alternative display resolutions out.
I've tried 2048x1152 and 2880x1620 so far -- both had no glitches in my limited testing.

 

[PieTunes]

Curious. If you decide to return your new purchase because of this glitch and your iMac came from China or Ireland, where do you return it to?

If in the US you can return it to any Apple Store. You can also choose to ship it back when you initiate the return process online and they’ll give you a prepaid shipping label to send it back to a domestic destination.

 

[LGR]

Has anyone tried the 5700xt with an Apple XDR display attached? Any issues?

 

[filmgirl]

That is if this is something Apple can do. The drivers are provided by AMD, afaik. People have had various issues with other AMD cards in Windows where the drivers remained unfixed for as long as 6 months. Doesn't mean it's gonna be the same here of course.

This is true but with PC drivers, there are a lot more variables in place. With an iMac, every owner is going to have one of two CPUs, the same logic board, the same built-in display, etc. So in theory, isolating/solving the problem should be easier for both AMD and Apple. That doesn’t mean it will be of course, but it’s a bit different than writing a driver that has to work with a wide range of hardware components.

And AMD maintains the Boot Camp drivers now, but I would be shocked if Apple wasn’t involved in some way with the drivers for macOS.

At the risk of going down a speculative path that I don’t have any basis for going down — I wonder if the issue here is with Display Stream Compression (DSC), which is part of the 5700 and 5700XT and is in the DisplayPort 1.4 spec that allows for dual 6K monitors to be connected (despite TB3 not having the bandwidth to support that).

Apple is using DSC in the 16” MacBook Pro and the W5700X supports it too.

Just a thought...

 

[torifile]

Sigh. I suppose I did! ???

In my mind, there are no Karens who are also DCFC fans. (I was the guy who tweeted the concert recording suggestion yesterday)

 

[Shaman777]

At first I was happy, then I started opening programs working in them and comparing them one by one, and I didn't notice much difference. It was disappointing for me.

I discovered another Terrible glitch (Retina 5K, 27-inch, 2020), I opened the Spark AR Studio program

Here is a video comparison of iMac 2019 and 2020

You can download this program from the official website and test it on your 2020 Machines if you have a different configuration you will be interested in the results

Download here https://sparkar.facebook.com/ar-studio/


I work in it often, and I was shocked by what happened


Test
iMac (Retina 5K, 27-inch, 2019) 3.6 GHz 8-core Intel Core I9 processor 72GB 2667 MHz DDR4 Radeon Pro Vega 48 8GB on the

iMac (Retina 5K, 27-inch, 2020) 3.6 GHz 10-core Intel Core I9 processor 72GB 2667 MHz DDR4 AMD Radeon Pro 5700 XT 16GB

 

[Conradiminja]

/////

 

[kkh]

This is true but with PC drivers, there are a lot more variables in place. With an iMac, every owner is going to have one of two CPUs, the same logic board, the same built-in display, etc. So in theory, isolating/solving the problem should be easier for both AMD and Apple. That doesn’t mean it will be of course, but it’s a bit different than writing a driver that has to work with a wide range of hardware components.

And AMD maintains the Boot Camp drivers now, but I would be shocked if Apple wasn’t involved in some way with the drivers for macOS.

At the risk of going down a speculative path that I don’t have any basis for going down — I wonder if the issue here is with Display Stream Compression (DSC), which is part of the 5700 and 5700XT and is in the DisplayPort 1.4 spec that allows for dual 6K monitors to be connected (despite TB3 not having the bandwidth to support that).

Apple is using DSC in the 16” MacBook Pro and the W5700X supports it too.

Just a thought...

Yep. Less variables when it's the iMac, for sure. But it sounded like AMD has struggled to get the drivers right even when they have identified how and where it's going wrong. Again, certainly hoping this won't be the case here.

I guess it could be anything. Including something with the DSC. Would chroma subsampling for clusters of pixels have something to do with what we're seeing? Earlier in this thread we established that DSC is not used internally. But, interestingly, people attaching external screens have reported that having an impact. Does an external display being connected somehow make the hardware or the drivers do something correctly?

 

[joevt]

At the risk of going down a speculative path that I don’t have any basis for going down — I wonder if the issue here is with Display Stream Compression (DSC), which is part of the 5700 and 5700XT and is in the DisplayPort 1.4 spec that allows for dual 6K monitors to be connected (despite TB3 not having the bandwidth to support that).

Apple is using DSC in the 16” MacBook Pro and the W5700X supports it too.

Apple doesn't use DSC for internal displays. Check output of AGDCDiagnose command to verify. All iMac 5K displays use two DisplayPort 1.2 connections.
There are very few external displays that support DSC, such as recent high resolution displays like the XDR, or recent high refresh rate 4K+ displays like the Samsung Odyssey G9.

I guess it could be anything. Including something with the DSC. Would chroma subsampling for clusters of pixels have something to do with what we're seeing? Is DSC used for the internal display? I assume it goes 4:4:4 per pixel straight to the internal screen and thus avoids any kind of Thunderbolt bottlenecks and the need for compression. But, interestingly, people attaching external screens have reported that having an impact. Does an external display being connected somehow make the hardware or the drivers do something correctly?

Apple doesn't use chroma sub sampling except for some situations (for example, HDMI output from Mac Mini 2018 to 4K 60Hz display even though the display allows 8 bpc RGB). The XDR display doesn't support chroma sub sampling. I don't recall if iMac displays allowed chroma sub sampling (you can get the EDIDs of each half of the display using the AGDCDiagnose command, then parse the EDID with edid-decode command which you can download and install).

It seems to me that adding a little extra load to the 5700XT solves the white line glitch problem - such as having a frame buffer scaling (when the frame buffer is not a 5120x2880 based lores or HiDPI mode) - or an extra display output (when connecting an external display).

 

[kkh]

AGDCDiagnose tells me it's using the 6000 series of drivers. Instead of the 5000 ones. I assume they would not get something that basic wrong? Maybe the driver numbers have nothing to do with actual model numbers supported?

Here's a list of all the AMD drivers available on my machine:

drwxr-xr-x@ 3 root wheel 96 Jul 7 20:04 AMD10000Controller.kext
drwxr-xr-x@ 3 root wheel 96 Jul 7 20:04 AMD7000Controller.kext
drwxr-xr-x@ 3 root wheel 96 Jul 7 20:04 AMD8000Controller.kext
drwxr-xr-x@ 3 root wheel 96 Jul 7 20:04 AMD9000Controller.kext
drwxr-xr-x@ 3 root wheel 96 Jul 7 20:04 AMD9500Controller.kext

drwxr-xr-x@ 3 root wheel 96 Jul 7 20:04 AMDFramebuffer.kext

drwxr-xr-x@ 3 root wheel 96 Jul 7 20:04 AMDMTLBronzeDriver.bundle

drwxr-xr-x@ 3 root wheel 96 Jul 7 20:04 AMDRadeonServiceManager.kext

drwxr-xr-x@ 3 root wheel 96 Jul 7 20:04 AMDRadeonVADriver.bundle
drwxr-xr-x@ 3 root wheel 96 Jul 7 20:04 AMDRadeonVADriver2.bundle

drwxr-xr-x@ 3 root wheel 96 Jul 7 20:04 AMDRadeonX4000.kext
drwxr-xr-x@ 3 root wheel 96 Jul 7 20:04 AMDRadeonX4000GLDriver.bundle
drwxr-xr-x@ 3 root wheel 96 Jul 7 20:04 AMDRadeonX4000HWServices.kext

drwxr-xr-x@ 3 root wheel 96 Jul 7 20:04 AMDRadeonX5000.kext
drwxr-xr-x@ 3 root wheel 96 Jul 7 20:04 AMDRadeonX5000GLDriver.bundle
drwxr-xr-x@ 3 root wheel 96 Jul 7 20:04 AMDRadeonX5000HWServices.kext
drwxr-xr-x@ 3 root wheel 96 Jul 7 20:04 AMDRadeonX5000MTLDriver.bundle
drwxr-xr-x@ 3 root wheel 96 Jul 7 20:04 AMDRadeonX5000Shared.bundle

drwxr-xr-x@ 3 root wheel 96 Jul 7 20:04 AMDRadeonX6000.kext
drwxr-xr-x@ 3 root wheel 96 Jul 7 20:04 AMDRadeonX6000Framebuffer.kext
drwxr-xr-x@ 3 root wheel 96 Jul 7 20:04 AMDRadeonX6000GLDriver.bundle
drwxr-xr-x@ 3 root wheel 96 Jul 7 20:04 AMDRadeonX6000HWServices.kext
drwxr-xr-x@ 3 root wheel 96 Jul 7 20:04 AMDRadeonX6000MTLDriver.bundle
drwxr-xr-x@ 3 root wheel 96 Jul 7 20:04 AMDRadeonX6000Shared.bundle

drwxr-xr-x@ 3 root wheel 96 Jul 7 20:04 AMDShared.bundle

drwxr-xr-x@ 3 root wheel 96 Jul 7 20:04 AMDSupport.kext

 

[kkh]

Apple doesn't use DSC for internal displays. Check output of AGDCDiagnose command to verify. All iMac 5K displays use two DisplayPort 1.2 connections.

Remembered after posting that we already looked into this earlier in this thread. Confirmed it's off for the internal display. At least according to AGDCDiagnose.

 

[kkh]

Interesting how the one kext labeled 5700 (*HWLibs) is not loaded. It uses the *6200HWLibs instead. And then 6000 series extensions for the rest.

 

[joevt]

Interesting how the one kext labeled 5700 (*HWLibs) is not loaded. It uses the *6200HWLibs instead. And then 6000 series extensions for the rest.

The driver numbers do not use the same numbering as the GPU model numbers.
For example, my AMD RX 580 uses the three X4000 drivers, AMD9500Controller, AMDFramebuffer, AMDSupport, AMDRadeonServiceManager.
I don't know where the X4000 (Polaris?) or X5000 (Vega?) or X6000 (Navi?) numbers come from.

 

[Shaman777]

I found another terrible glitch on the 2020 model in the Spark AR Studio program. There is a video device simulator, so if you start scaling this window, the program freezes and the entire system freezes, everything is fine on my 2019 model

 

[torifile]

I found another terrible glitch on the 2020 model in the Spark AR Studio program. There is a video device simulator, so if you start scaling this window, the program freezes and the entire system freezes, everything is fine on my 2019 model

That sounds like a bug in the software and not a problem due to the 2020 iMac.

 

[Shaman777]

111

That sounds like a bug in the software and not a problem due to the 2020 iMac.

I thought so too, but I tested these models synchronously

THERE IS NO ERROR HERE
iMac (Retina 5K, 27-inch, 2019) 3.6 GHz 8-core Intel Core I9 processor 64GB 2667 MHz DDR4 Radeon Pro Vega 48 8GB
osx 10.15.6

THERE IS AN ERROR HERE
iMac (Retina 5K, 27-inch, 2020) 3.6 GHz 10-core Intel Core I9 processor 64GB 2667 MHz DDR4 AMD Radeon Pro 5700 XT 16GB
osx 10.15.6

 

[ADGrant]

111


I thought so too, but I tested these models synchronously

THERE IS NO ERROR HERE
iMac (Retina 5K, 27-inch, 2019) 3.6 GHz 8-core Intel Core I9 processor 64GB 2667 MHz DDR4 Radeon Pro Vega 48 8GB
osx 10.15.6

THERE IS AN ERROR HERE
iMac (Retina 5K, 27-inch, 2020) 3.6 GHz 10-core Intel Core I9 processor 64GB 2667 MHz DDR4 AMD Radeon Pro 5700 XT 16GB
osx 10.15.6

That doesn't mean there is not a bug in the software, but I think it is the AMD driver. Have you tried it on any other Mac with a Navi GPU?

 

[torifile]

111


I thought so too, but I tested these models synchronously

THERE IS NO ERROR HERE
iMac (Retina 5K, 27-inch, 2019) 3.6 GHz 8-core Intel Core I9 processor 64GB 2667 MHz DDR4 Radeon Pro Vega 48 8GB
osx 10.15.6

THERE IS AN ERROR HERE
iMac (Retina 5K, 27-inch, 2020) 3.6 GHz 10-core Intel Core I9 processor 64GB 2667 MHz DDR4 AMD Radeon Pro 5700 XT 16GB
osx 10.15.6

That means nothing. Sorry.

 

[Shaman777]

glitch and Crash