Driver issue with Ryzen APU - crashes on any load

A

a13antichrist

n00b
Joined
May 24, 2018
Messages
12
Pretty sure this is a driver issue given the findings so far.. could be wrong though, but there isn't much else in the system ;) Specs of which are in the sig.

I've recently built a new machine but am having constant crashes anytime I try to do anything graphics-related. I get a number of different blue-screen reports:

DRIVER_IRQL_NOT_LESS_OR_EQUAL
SYSTEM_SERVICE_EXCEPTION
KERNEL_something-something-I-forget-exactly
and there was another one which I forget also.

And depending on the case I have had two registers called out also:

atikmdag.sys
dxgmm2.sys

So it -seems- to point to something Radeon-driver-related. However I haven't seen similar issues from AMD discussions so I wondered if it couldn't be board-related..

I am currently using BIOS L4.51D; I had the 4.60 on before but changed it as I thought I was getting system shutdowns randomly; appears it's just the OS shutting off the CPU fan when sleeping - seemed odd to me but ok.

I was using a 17.x Radeon driver, upgraded to the 18.5.1 but that hasn't helped.

Playing Youtube in 4K it will sometimes last a fair while, an hour or so or sometimes only a few minutes; then the Chrome tab may crash or the video might start going blocky or distorted, sometimes one then the other.

Playing HEVC 265 video in VLC crashes the computer within a few minutes. Playing on MPC-HC lasted a bit longer, got up to 30mins or so.

I have tried with multiple memory sticks (have some 2666 modules here, also tried running the 3200 at only 2133) and multiple PSUs. No tangible differences in behaviour there.

I don't have anything happening on this machine at present; it's to be the HTPC so only a couple of media players installed, and nothing heavy yet.

Another thing - if I enable the HDR + WCG setting in windows, the CPU grinds to almost a halt when playing a 4K HDR video in Youtube. Weird..

Any ideas? Bad CPU/board/driver install?
 
cooling? any oc? there have been some issues with mobo software like asus suite type stuff, so try removing anything like that.
 
pendragon1 said:
cooling? any oc? there have been some issues with mobo software like asus suite type stuff, so try removing anything like that.
Click to expand...

Noctua L9, temps don't cross 40 when decoding HEVC 265; no overclock and I haven't installed any mobo tools apart from the AMD Ryzen Master software, which I'm using purely as monitoring at the moment (haven't activated any modifications to settings).
 
hmm…. maybe try manually setting your ram timings and up the voltage a bit. they can handle at least +.1v no prob.
also how did you load windows? ive had a bad usb drive image cause weird issues and recreating the usb and reloading was the only way to sort it out.
 
pendragon1 said:
hmm…. maybe try manually setting your ram timings and up the voltage a bit. they can handle at least +.1v no prob.
also how did you load windows? ive had a bad usb drive image cause weird issues and recreating the usb and reloading was the only way to sort it out.
Click to expand...

I've tried two different sets of ram, and this is 3200Mhz retail; even at 2133 the behaviour is the same.

Same USB image I've used for my current laptop and two friends' laptops and no problems there so far..

I do have a small update though: I removed all AMD drivers using the AMD CLeanup Utility. Then I re-installed the graphics drivers -only- and manually using Device Manager and not the packaged installation. Full disclosure, the driver package I used for this was the amd-chipset-drivers18.10.0418 which I guess I downloaded in April; Radeon driver version actually lists the same as when installed by the new Adrenaline package release yesterday (24.20.11016.4) so that's a bit odd. Well it also says 'Driver date 16/05/2018 so I'm not entirely sure how that happened :p But in any case the install was done manually through DM and not the package installer.

With this driver installed and none of the other AMD bits (other than Ryzen Master) I am able to watch any 4K Youtube or HEVC media without suffering the crashes - EXCEPT when I turn the HDR + WCG mode on. With that mode enabled I see the same crashes as before.
 
I would remove all AMD drivers again. Then use DDU to make sure all AMD drivers are really gone.
http://www.guru3d.com/files-details/display-driver-uninstaller-download.html
Then you need chipset drivers installed first. Get these from AMD's site directly.
Then you need to install the latest video drivers, again from AMD directly.
Both those are needed, not just 1.

I have a 2400G and I have some blue screens from time to time, but it was worse when I tried running my RAM at full speed. I had much more stability when I enabled XMP, but then manually picked a slower speed. I have 3000 RAM that will run at 2933 when XMP is selected. I have to manually lower that to 2800 for better stability.
 
This sounds more like RAM related. What version win10 are you on? Newer windows install? Old?
 
a13antichrist said:
I've tried two different sets of ram, and this is 3200Mhz retail; even at 2133 the behaviour is the same.

Same USB image I've used for my current laptop and two friends' laptops and no problems there so far..

I do have a small update though: I removed all AMD drivers using the AMD CLeanup Utility. Then I re-installed the graphics drivers -only- and manually using Device Manager and not the packaged installation. Full disclosure, the driver package I used for this was the amd-chipset-drivers18.10.0418 which I guess I downloaded in April; Radeon driver version actually lists the same as when installed by the new Adrenaline package release yesterday (24.20.11016.4) so that's a bit odd. Well it also says 'Driver date 16/05/2018 so I'm not entirely sure how that happened :p But in any case the install was done manually through DM and not the package installer.

With this driver installed and none of the other AMD bits (other than Ryzen Master) I am able to watch any 4K Youtube or HEVC media without suffering the crashes - EXCEPT when I turn the HDR + WCG mode on. With that mode enabled I see the same crashes as before.
Click to expand...
id still try a bit extra voltage, it wont hurt to try. ok prob not the image. maybe a ddu and driver reinstall like pga suggested might help. it seems like youre part way to stability...
 
pgaster said:
I would remove all AMD drivers again. Then use DDU to make sure all AMD drivers are really gone.
http://www.guru3d.com/files-details/display-driver-uninstaller-download.html
Then you need chipset drivers installed first. Get these from AMD's site directly.
Then you need to install the latest video drivers, again from AMD directly.
Both those are needed, not just 1.

I have a 2400G and I have some blue screens from time to time, but it was worse when I tried running my RAM at full speed. I had much more stability when I enabled XMP, but then manually picked a slower speed. I have 3000 RAM that will run at 2933 when XMP is selected. I have to manually lower that to 2800 for better stability.
Click to expand...

Hmm I was actually hoping to undervolt this guy a little eventually too :p I will drop it to 2933 and see how it holds, and try your driver process also. Funny thing though, when I selected the "Chipset drivers" it had first of all, "No driver currently installed" and then the options are:
18.10 (local) - downloaded 24/04
18..5 (remote/internet - recommended)

And when I choose either one, my installation options are the following:
AMD Display Driver
AMD HDMI Audio Driver
AMD Problem Report Wizard
AMD Radeon Settings

And if I de-select the Display driver, the only one I can install is the Problem Report Wizard. So doesn't even look very chipset-y. But I will try "really" removing all drivers, and try again with the lastest chipset & graphics from AMD.

THUMPer said:
This sounds more like RAM related. What version win10 are you on? Newer windows install? Old?
Click to expand...

Well, it's --just-- updpated to 1803, but it was doing it on 1710 before.. or maybe it wasn't, I wasn't testing it until just recently so I guess I can't say for surrre. But I'm --pretty-- sure the first crash was on 1710 which is what made it prompt for install of 1803 the next boot..

pendragon1 said:
id still try a bit extra voltage, it wont hurt to try. ok prob not the image. maybe a ddu and driver reinstall like pga suggested might help. it seems like youre part way to stability...
Click to expand...

OK I will set the speed down to 2933 but keep the voltage at 1.35, which is default for 3200. Or 1.25 if it drops all the way to 1.2@2933. Thanks
 
a13antichrist said:
OK I will set the speed down to 2933 but keep the voltage at 1.35, which is default for 3200. Or 1.25 if it drops all the way to 1.2@2933. Thanks
Click to expand...
that's not how ram works. that ram is rated for 1.35v dont lower it or it will become unstable. instead increase it to 1.4v and see if it helps at all.
 
I just was looking at the Asrock page for that board and the bios update options. I don't see any bios that has AGESA 1.0.0.2a yet. Other board makers have bios updates with this, and it may help solve some issues. I'd keep checking and update once that is released.
 
that's a good point. or you could email and bug them for an update. help speed up the process maybe.
 
I had all kinds of lock ups when i was using the wifi on that board until i did the newest bios update. was 100 percent stable wired, until i needed to use the Wifi then i got forced reboots and all

Edit. i see you tried bios 4.60 hmm i am not sure that was the only issue i was having with my 2200g, I do have 2667 mhz memory running the standard XMP profile so i did not tweak much. I also did a clean install of 1803 for this build on my 960evo so i do not have much to say i guess. Are you using the wifi?
 
Last edited:
Not using the wifi. I haven't disabled it in BIOS, yet, but in any case yeah I'm using the BIOS that supposedly fixes that.
Also the L4.51D BIOS I'musing now is the AGESA 1.0.0.3b release (PinnacleAM4-1.0.0.3b to be precise).

pendragon1 said:
that's not how ram works. that ram is rated for 1.35v dont lower it or it will become unstable. instead increase it to 1.4v and see if it helps at all.
Click to expand...

Well when I select the 2133 option in UEFI it (automatically) drops the (indicated) voltage to 1.2.. I will try a small speed drop & voltage bump in a second though after I give my latest update:

After first running AMD Clean util, I rebooted into Safe Mode. Before running Guru process, when switching user from dummy to admin, system crashed & rebooted again. OK :/ This in Safe Mode still.
Ok finally after booting back to safe mode, ran the Guru cleaner and back to Windows.

installed "Chipset Drivers" labelled 18.10.0418. I noted that the items being installed were:
AMD PSP
AMD SMBus
AMD GPIO
AMD Graphics Driver 24.x
Balanced Power Plan 2.0.0

A note on these 'chipset drivers' - when installing, it says I'm using 18.10, but that there's a newer recommended version 18.5 (the adrenaline). Bit curious..

Allowed to install. Rebooted - Windows failed to start after two atttempts. So I stripped off all the drivers again. This time I skipped the "chipset drivers" and went straight to the new Adrenaline 18.5.1. After install I checked that all the previous components were there & with the same or better driver versions - they were. So either the driver removal didn't touch them or the Adrenaline 18.5 installs them also.

Eventually I got the system back into its correct state. I like to tempt fate, so I went straight for HDR mode. Both Chrome & Edge crash on any YouTube video within seconds. Chrome just kills the tab; Edge actually crashes the system (although to be fair, I only tried Edge once).
Playing videos while in HDR has given me a number of different BSODs (all in the last 2 hours):
- SYSTEM_SERVICE_EXCEPTION
- SYSTEM_EXCEPTION_NOT_HANDLED: atikmdag.sys
- DRIVER_IRQL_etc_etc: dxgmm2.sys
- VIDEO_TDR_FAILURE: atikmdag.sys

When not doing anything video-related, i.e. just streaming an internet radio station through Chrome, system doesn't seem to be complaining. Been running about an hour now no complaints, this is still with HDR on.

One thing I did note: the "PTC" in the Ryzen Master is running at 95°C. Gotta look into that..

Now I will try raw 4K with no HDR before diving into any memory & voltage changes.
 
Last edited:
a13antichrist said:
Well when I select the 2133 option in UEFI it (automatically) drops the (indicated) voltage to 1.2.. I will try a small speed drop & voltage bump in a second though after I give my latest update:
Click to expand...
well that's wrong and I dont know why It would do that, maybe xmp or something. but ddr4 on amd requires minimum 1.35v no matter the speed. manually set your ram speed, timings and voltage. no not use xmp and that is supposed to be for intel systems, AMDs version is call AMP. also that 95c should be a concern, hows your case airflow?

oh and HDR is a total mess anyways so it could just be a drivers issue that needs sorting. but regardless, fix the ram settings first.
 
pendragon1 said:
well that's wrong and I dont know why It would do that, maybe xmp or something. but ddr4 on amd requires minimum 1.35v no matter the speed. manually set your ram speed, timings and voltage. no not use xmp and that is supposed to be for intel systems, AMDs version is call AMP. also that 95c should be a concern, hows your case airflow?

oh and HDR is a total mess anyways so it could just be a drivers issue that needs sorting. but regardless, fix the ram settings first.
Click to expand...

XMP is what the board gives me, I'm guessing even if theirs is technically different they're calling it XMP just the same so that people know it when they see it.

Funny story, I took a look into the HDR settings, I was on 8-bit, 4:4:4. I have an option for 4:4:2 but still only 8-bit; if I choose 4:2:0 though I get 8, 10 or 12bit. Currently running 4:2:0 10-bit with a Chrome Youtube tab and --just-- as I was typing that 'chrome' it crashed - the tab, not the whole thing. I'll try 4:2:2 8-bit now. But yeah looks like I'll take the HDR off soon and not worry about that part for now.
 
a13antichrist said:
XMP is what the board gives me, I'm guessing even if theirs is technically different they're calling it XMP just the same so that people know it when they see it.

Funny story, I took a look into the HDR settings, I was on 8-bit, 4:4:4. I have an option for 4:4:2 but still only 8-bit; if I choose 4:2:0 though I get 8, 10 or 12bit. Currently running 4:2:0 10-bit with a Chrome Youtube tab and --just-- as I was typing that 'chrome' it crashed - the tab, not the whole thing. I'll try 4:2:2 8-bit now. But yeah looks like I'll take the HDR off soon and not worry about that part for now.
Click to expand...
the board will still show that the ram has xmp profiles but lots of amd boards wont load the xmp profiles correctly because its meant for intel and doesn't configure it properly. the random errors is indicative of ram issues so try running it with manual settings and up the voltage a bit. it will not hurt to try and could sort your issue out. especially since the apu shares system ram for vram. ive had plenty of corsair ram that has needed just a bit extra voltage to stabilize properly.
 
So, I don't see how to set the speed manually without using a profile. The voltage I can do, and right now I have set it to 1.4V @ 2133 - As that's what the board sets by default if you don't use a profile.

But at least, at that settings I was able to run through an entire hour-long video @4K and then also played a VLC 4K movie on top (youtube in the background still playing) and no crash. I'm still a little dubious about the HDR settings though because depending which values I select (8b/10b/12b and 4:4:4/4:2:4/4:2:0) sometimes the HDR mode in Display Settings seems to end up switched off.

So for now I've switched back to RGB 4:4:4 @8b which I believe is the default, and I have the HDR setting on with memory running at 3200Mhz. Seems stable at least for the last 30 mins so I can --tentatively-- say that the boost in Mem voltage seems to have helped.

Will update later when I've given it more time to run and throw more things at it. Thanks for the help so far ;)
 
No prob. The has to be a way to manually set it, just need to find it. But you're getting there!
 
It actually didn't last very long, a couple of hours and it gave up once I had a youtube and a VLC player running at the same time. Detoriated from there. Am thinking now to redo the install with a fresh ISO of 1803.

I also did notice that the 3200 sticks I'm using aren't officially on the Ryzen-supported list, that I can find, although all the other Vengeance lines are, s to be extra-sure I'll swap back the 2666s that I have and retry what I've done so far..
 
Sooooo

I guess I'm pretty dumb then.

After all the drama.. I try a 2666 RAM kit I have sitting here.. running at only 2133 but at least at that point (and no XMP profile) I have no issues. So I checked it up and it turns out my 3200 I've been using actually ---isn't-- on the supported list (CMK8GX4M2B3200C16R ). This isn't to say that it's the RAM's fault, per se, but.. sure doesn't help the case. I could have sworn it was listed as compatible before. :/

Anyone know how to check if there really is a bad match between this board (Fatal1ty AB350 ITX) and/or CPU and this Mem? Or is it just "unknown" so far?
 
no I think what youre experiencing is the sign that it wont work. youll have to get something off the qvl.
 
you can probably tune in some more speed, like to 2866mhz and top out around there.
 
I'm running the 'bad' ram at 2133 now and so far (although it's only 10mins) it seems to be handling 3 things at a time that usually, individually, would either crash or BSOD after 15-30mins. Youtube is jerking though, and does that even without other things running, so that's odd, but also in line with what I saw previously. Didn't have that with the 'supported' RAM at the same speed. :/ But I'm also undervolting this RAM now to match the 'supported' one (1.25 instead of 1.35; the 2666 I was using at 2133 is 1.2 default and had that at 1.25 also) although it doesn't seem likely that RAM voltage would cause stuttering, does it? So many damned variables :/
 
id set your ram voltage to 1.35v and leave it there. the standard is 2133 at 1.2v anything over that is technically a overclock and will need more voltage. if the ram isn't getting enough voltage it could cause all sorts of goofy issues.
 
I got some goofy issues :p after about 15mins the VLC session went goofy, looked as if the video was encrypted or something; I've seen soimething similar if you try to play a DRM-encrypted video without a key. Or when you lose a keyframe somewhere but this was across the whole file. Restarting the player didn't help, no diff. Restarted, set the ram voltage up back to 1.35 and the XMP profile back on. Crashed (IRQL_etc) before I even had a chance to sign in :p Now I've got it playing a Youtbe demo only (HDR) and if/when it crashes again I'll manually set the memory to 2933 (no XMP) and 1.4V (to be sure).

I was also reading that there is some advice to not even use the HDR + WCG setting within Windows, as support was still flaky. Mind you, that was dated back in July '17; there have been two Windows releases since then. Also that was back when it wasn't even called 'WCG', but rather, 'advanced colour'. In any case that's something else I need to get to the bottom of - this grey/dimming of the screen whenever I enable HDR on anything other than 8-bit 4:2:0. I saw mentioned that 8bit 4:4:4 YCbCr already was "HDR" and that the Windoes setting wasn't needed with that. So more reading to do I guess...
 
that ram will take 1.45v no prob so dont be afraid to raise it if needed.
all of what you said about hdr is true, its have lots of issues and everyone gets that dimming effect. there are others around here talking aboot it.
 
You know it is the ram google the 2 keywords :) . And email ASrock see if they have any good tips.
 
Umm this looks supes weird to me.. I guess the 200Mhz could be due to DDR, lke CPU-Z does, but I should check the other memory. DDR3, though, really doens't look right...

UKd7SnH
Captureradeoned.png
 
just a misreading on the ddr3 and the core clock is probably because its idle. makes sure your bios is u to date and there are new drivers available too. go grab em.
 
I'm sorry to say but my gut feeling is you have some bad hardware somewhere. Tinkering for days and it still chokes on little things? Not a good sign!!
 
Yeah, the RAM thing is still a fucking mess, like it was last year. On its launch day, I bought myself a brand new 2700x with 8x2GB DDR4 3200, and an MSI x470 Gaming Plus mobo, and I cannot get the system to behave properly if the RAM is set to 3200 - the system hard locks and most of the times I have to reset the UEFI, but sometimes it resets by itself and comes to life. I can only run the RAM at 2933 at the highest, BIOS updates and all the latest drivers and all that didn't do anything to improve the situation.
The 2933 is the XMP1 profile, and the 3200 is the XMP2 profile, so it seems like at least one of the 2 XMP settings works fine on the AMD platform. but the manufacturers still didn't get their shit together to make sure ALL profiles work properly, not just a half of them.
Makes me feel being ripped off :(
 
serial__thrilla said:
Yeah, the RAM thing is still a fucking mess, like it was last year. On its launch day, I bought myself a brand new 2700x with 8x2GB DDR4 3200, and an MSI x470 Gaming Plus mobo, and I cannot get the system to behave properly if the RAM is set to 3200 - the system hard locks and most of the times I have to reset the UEFI, but sometimes it resets by itself and comes to life. I can only run the RAM at 2933 at the highest, BIOS updates and all the latest drivers and all that didn't do anything to improve the situation.
The 2933 is the XMP1 profile, and the 3200 is the XMP2 profile, so it seems like at least one of the 2 XMP settings works fine on the AMD platform. but the manufacturers still didn't get their shit together to make sure ALL profiles work properly, not just a half of them.
Makes me feel being ripped off :(
Click to expand...

Is it on the QVL?
 
serial__thrilla said:
Yeah, the RAM thing is still a fucking mess, like it was last year. On its launch day, I bought myself a brand new 2700x with 8x2GB DDR4 3200, and an MSI x470 Gaming Plus mobo, and I cannot get the system to behave properly if the RAM is set to 3200 - the system hard locks and most of the times I have to reset the UEFI, but sometimes it resets by itself and comes to life. I can only run the RAM at 2933 at the highest, BIOS updates and all the latest drivers and all that didn't do anything to improve the situation.
The 2933 is the XMP1 profile, and the 3200 is the XMP2 profile, so it seems like at least one of the 2 XMP settings works fine on the AMD platform. but the manufacturers still didn't get their shit together to make sure ALL profiles work properly, not just a half of them.
Makes me feel being ripped off :(
Click to expand...

what ram did you get?
 
You must log in or register to reply here.
Back
Top