ASRock’s X600 DeskMini – Finally – Again

Years ago, the SFF community waited with bated breath for the ASRock DeskMini A300. With AMD APU support, it offered a better balance than Intel chips between the CPU and GPU. This SFF Gem was the ideal Deskmini, and the SFF Network community rallied around it by creating custom 3D printed cases, BIOS updates, and more.
Now, the times have changed and AMD’s AM5 platform has arrived. Again, the SFF community waited. It felt like ages but ASRocks X600 DeskMini is finally here.
Asrock was kind enough to send a sample unit to Small Form Factor Network for testing and review. No review guidance was...

Continue reading...

 

[nirvana]

Few years ago I had a problem with a laptop after replacing its memory with new (higher capacity) sticks. Memtest passed OK but shortly after I started experiencing random BSODs and other weird issues, and I knew it was due to the new memory since I didn't have a single issue before the memory change.

I tried everything I could imagine to make it more stable (disabling XMP, running low frequency speeds, etc) but no chance. I was also annoyed because I was unable to catch any errors using different combinations of memory test programs and memory settings. Despite of this, I RMAed them and after few weeks I got new ones (so I assume the manufacturer was able to detect that they were faulty).

I haven't had a single issue after installing the replacement ones I got back from the manufacturer.

What I mean is that memory may be faulty despite not showing in any memory test programs.

 

[opensesame]

Can you clarify the above statements to help us understand the cause of the problem, XMP or drivers?

Sorry for not being clear. The first post you quoted is from running Linux, where AMD drivers are preinstalled and turning XMP on caused the artifacts to appear. The second post is from running Windows, where I installed AMD Adrenalin while XMP was on, and the artifacts appeared during the installer.

Hopefully this table tells the full story (note that BallisticNG is multiplatform):

OS	Driver	RAM setting	Artifacts on desktop	Artifacts while running BallisticNG (windowed)	Artifacts while running BallisticNG (fullscreen)
Windows	Microsoft	Auto	No	No	No
Windows	Microsoft	XMP Profile 1	No	No	No
Windows	AMD	Auto	No	No	No
Windows	AMD	XMP Profile 1	Yes	Yes	No
Linux	AMD	Auto	No	No	No
Linux	AMD	XMP Profile 1	Yes	Yes	Yes

 

[dacnick]

You mentioned using XMP but also manually changing some ram timings afterwards. The artefacts are usually instable vram from past experiences.
You also mentioned the sparkles disappear in certain scenarios after you launch a game fullscreen, that could also point to drivers or bios.

 

[BaK]

Sorry for not being clear. The first post you quoted is from running Linux, where AMD drivers are preinstalled and turning XMP on caused the artifacts to appear. The second post is from running Windows, where I installed AMD Adrenalin while XMP was on, and the artifacts appeared during the installer.

Hopefully this table tells the full story (note that BallisticNG is multiplatform):

OS	Driver	RAM setting	Artifacts on desktop	Artifacts while running BallisticNG (windowed)	Artifacts while running BallisticNG (fullscreen)
Windows	Microsoft	Auto	No	No	No
Windows	Microsoft	XMP Profile 1	No	No	No
Windows	AMD	Auto	No	No	No
Windows	AMD	XMP Profile 1	Yes	Yes	No
Linux	AMD	Auto	No	No	No
Linux	AMD	XMP Profile 1	Yes	Yes	Yes

Thanks for detailed answer!
Still hard to figure something out though...

Windows / Microsoft / XMP / Profile 1 / No / No / No
This line seems to indicate that the RAM is working fine even when running at the XMP settings.

Then we see that both Windows and Linux have the issue with the AMD drivers and the XMP settings (except in full screen mode on Windows).
Which should point to a driver issue.

Maybe check on Fedora if there is something showing up on boot with dmesg or more specifically with dmesg |grep amdgpu

 

[opensesame]

Few years ago I had a problem with a laptop after replacing its memory with new (higher capacity) sticks. Memtest passed OK but shortly after I started experiencing random BSODs and other weird issues, and I knew it was due to the new memory since I didn't have a single issue before the memory change.

I tried everything I could imagine to make it more stable (disabling XMP, running low frequency speeds, etc) but no chance. I was also annoyed because I was unable to catch any errors using different combinations of memory test programs and memory settings. Despite of this, I RMAed them and after few weeks I got new ones (so I assume the manufacturer was able to detect that they were faulty).

I haven't had a single issue after installing the replacement ones I got back from the manufacturer.

What I mean is that memory may be faulty despite not showing in any memory test programs.

Your experience sounds all too familiar. I've gone ahead and submitted an RMA request to Kingston to get the ball rolling.

You mentioned using XMP but also manually changing some ram timings afterwards. The artefacts are usually instable vram from past experiences.
You also mentioned the sparkles disappear in certain scenarios after you launch a game fullscreen, that could also point to drivers or bios.

Indeed, there are multiple possibilities, but I can't pinpoint the root cause, so I'm at a stalemate.

Thanks for detailed answer!
Still hard to figure something out though...

Windows / Microsoft / XMP / Profile 1 / No / No / No
This line seems to indicate that the RAM is working fine even when running at the XMP settings.

Then we see that both Windows and Linux have the issue with the AMD drivers and the XMP settings (except in full screen mode on Windows).
Which should point to a driver issue.

Maybe check on Fedora if there is something showing up on boot with dmesg or more specifically with dmesg |grep amdgpu

I ran dmesg | grep "amd" after booting, with and without XMP, and the output (https://pastebin.com/ 7WuhkNLA) was identical both times (I removed timestamps and diff'd to confirm). I then crawled through both entire outputs. Nothing obvious pops out at me in any of the lines containing "amd", "memory", "RAM", or those around them.

More troubleshooting...

• Totally reinstalled Windows and chose the oldest version (24.1.1) of AMD Adrenalin from the 8600G support page, just in case the newest one is the problem. No difference.
• Tried a few different values for VRAM size (Advanced -> AMD CBS -> NBIO Common Options -> GFX Configuration). Auto is default. 1 GB made no difference. 2 GB and 16 GB both caused the DeskMini to hang on boot (power light on, no video output, waited 10+ mins and cut power), but I could still enter BIOS. I put it back on Auto.
• This Reddit comment suggests the 8600G doesn't fare well with speeds above DDR5-6000, but I haven't found any info on this. Not to mention, my artifacts appear on all speeds above DDR5-4800, not just DDR5-6400.

I've been at this for 3 weeks and have grown tired of it. I will try any further suggestions posted here, otherwise I'm waiting on Kingston with fingers crossed.

 

[Stoney]

Did you try not to load the XMP but the JEDEC profile and set the speed to 6400 manually? That way you should end up with 1.2v VDD/VDDIO/VDDQ

 

[opensesame]

Did you try not to load the XMP but the JEDEC profile and set the speed to 6400 manually? That way you should end up with 1.2v VDD/VDDIO/VDDQ

The default profile, called "DDR5-4800 38-38-38-77 1.10V", sets the mentioned voltages to 1.10V, even after choosing DDR5-6400 from the Memory Speed drop-down menu, so I changed them manually. I experimented with the following combinations:

Memory Speed	VDD/VDDIO/VDDQ
DDR5-6400	1.10
DDR5-6400	1.18
DDR5-6400	1.19
DDR5-6400	1.20
DDR5-6400	1.21
DDR5-6400	1.22
DDR5-6400	1.23
DDR5-6400	1.24
DDR5-6400	1.25
DDR5-6000	1.20
DDR5-6000	1.25
DDR5-6000	1.30

The first row caused the DeskMini to stop booting. I couldn't get into Windows, Linux, or BIOS. I had to reset CMOS to get it working again.

The other rows had the following outcome: artifacts everywhere, including BallisticNG fullscreen on Windows. This is interesting because previously, it was the only scenario where artifacts weren't present. Now they appear within a few seconds of launching the game, and there are also far fewer of them compared to playing on Linux (https://imgur.com/ aMiXR6b).

 

[Stoney]

does raising or lowering the SoC voltage affect the artifacts? Also does it help if you force UCLK DIV1 Mode to UCLK=MCLK/2?

 

[opensesame]

SOLVED!

It looks like it was a driver/amdgpu problem after all. The Arch Wiki explains it all: https://wiki.archlinux.org/title/AMDGPU#Screen_artifacts_and_frequency_problem

Before applying this fix, I reset all BIOS settings back to default, except for turning XMP on (I started with the 6400 profile, then tested the 6000 and 5600 ones; all good). I then wrote to the file in question, which can be:

• auto, the default
• high, which creates more artifacts
• low, which eliminates them completely

It still doesn't answer the Windows side of things, and I don't see anything in AMD Adrenalin that resembles this parameter. However, I only planned to use Linux on this machine anyway, so it's solved for me. Many thanks to everyone for helping me out: @Stoney, @dacnick, @BaK, @nirvana, @HydrAxx747, @SFFMunkee

 

[Stoney]

linux(Manjaro) is also my daily OS but i don't have that artifacts and i didn't edit anything. The archwiki mentions artifacts at 60Hz+, do you run higher than 60Hz? Maybe thats why i don't have artifacts even with my memory at 8000 with my 60Hz display.

 

[BaK]

SOLVED!

Good job finding a fix!

 

[opensesame]

linux(Manjaro) is also my daily OS but i don't have that artifacts and i didn't edit anything. The archwiki mentions artifacts at 60Hz+, do you run higher than 60Hz? Maybe thats why i don't have artifacts even with my memory at 8000 with my 60Hz display.

I'm running 75 Hz right now. However, the problem persisted with lower refresh rates (50-60 Hz) on this monitor, plus on a different 60 Hz monitor.

What desktop environment are you running with Manjaro? Another experiment I tried was Fedora Xfce, with a weird result: no artifacts running the live USB, but artifacts after installing to disk and rebooting. I then did it again, ethernet disconnected, to rule out the possibility of some kind of auto-update causing it, and got the same result. Fedora GNOME, on the other hand, always had artifacts present, whether it was a live USB or installed to disk.

Good job finding a fix!

Thanks for your help!

 

[Stoney]

i am using KDE6 with X11, didn't try Wayland on the deskmini yet.

 

[HydrAxx747]

For those who complain that the DeskMini X600 does not have an OcuLink port and that ASRock does not offer one (they should, it's true), well now a simple M.2 NVMe to "direct" OcuLink output adapter now exists, certified PCIe4.0 4x compatible and very thin and with a cable length of your choice, it can even be installed on the secondary M.2 NVMe port under the motherboard without any worries, and the OcuLink port even has fixing points(screws provided) perfectly compatible with the COM port slot with detachable metal cover on the original DeskMini case(normally intended for the ASRock accessory for 3.5mm stereo jack output for desktop speakers):​

(also compatible with other old DeskMini A300 and X300)

Link : aliexpress.com/item/1005007486714630.html