Closed Bug 984488 Opened 12 years ago Closed 11 years ago

Blue Screen of Death Crashes with D3D11 OMTC on Windows 8.1 x64 with AMD 13.12 video drivers

Categories

(Core :: Graphics: Layers, defect)

Product:
Core
Component:
Graphics: Layers
Platform:
x86_64
Windows 8.1
Type:
defect
Priority:
Not set
Severity:
critical

Tracking

()

Status:
RESOLVED FIXED
Milestone:
mozilla32

People

(Reporter: timbugzilla, Assigned: bas.schouten)

References

Details

(Keywords: crash)

Attachments

(2 files)

Crash details from BlueScreenView App
4.32 KB, text/html
Details
Blacklist Direct2D for Catalyst 14.4
1.83 KB, patch
bjacob
: review+
Details | Diff | Splinter Review
When OMTC is enabled (D3D11) in FF nightly, I experience frequent Blue Screen of Death OS crashes while playing back video, especially if I start playback of a video file (x264 encoded) in MPC-HC (using LAV Filters (DXVA native decoding) and MadVR). I do not experience these crashes when OMTC is disabled. I am using Windows 8.1 x64, with an AMD HD7790 GPU and Catalyst 13.12 drivers (latest stable). According to the BlueScreenView App the crashes are occuring in atikmdag.sys and ntoskrnl.exe. Please see the attached file for some details.
Summary: Blue Screen of Death Crashes with D3D11 OMTC on Widows8.1 x64 with AMD 13.12 video drivers → Blue Screen of Death Crashes with D3D11 OMTC on Windows 8.1 x64 with AMD 13.12 video drivers
I was getting similar BSODs with OMTC enabled while using the latest AMD Catalyst 14.2 BETA 1.3 drivers. I went back to the 13.12(WHQL) drivers and haven't had a BSOD in days. Name: Firefox Version: 30.0a1 User Agent: Mozilla/5.0 (Windows NT 6.3; WOW64; rv:30.0) Gecko/20100101 Firefox/30.0 Graphics -------- Adapter Description: AMD Radeon HD 7900 Series Adapter Drivers: aticfx64 aticfx64 aticfx64 aticfx32 aticfx32 aticfx32 atiumd64 atidxx64 atidxx64 atiumdag atidxx32 atidxx32 atiumdva atiumd6a atitmm64 Adapter RAM: 3072 Device ID: 0x6798 Direct2D Enabled: true DirectWrite Enabled: true (6.3.9600.16384) Driver Date: 12-6-2013 Driver Version: 13.251.0.0 GPU #2 Active: false GPU Accelerated Windows: 1/1 Direct3D 11 (OMTC) Vendor ID: 0x1002 WebGL Renderer: Google Inc. -- ANGLE (AMD Radeon HD 7900 Series Direct3D9Ex vs_3_0 ps_3_0) windowLayerManagerRemote: true AzureCanvasBackend: direct2d AzureContentBackend: direct2d AzureFallbackCanvasBackend: cairo AzureSkiaAccelerated: 0
With the latest AMD drivers 14.3 beta1 BSODs are gone for me.
I installed the 14.3 beta 1 drivers and had been crash free for a week. Just got a BSOD crash with today's nightly build when playing an embedded youtube video.
I had no problems running the 13.12 CATS with OMTC enabled. However, with the latest 14.4 CATS I BSOD with OMTC enabled. Disabling OMTC there are no BSODs. Doesn't seem related to playing videos in my case. BSOD can happen just sitting on a page. The trigger seems to be OMTC turned on. Name: Firefox Version: 32.0a1 User Agent: Mozilla/5.0 (Windows NT 6.3; WOW64; rv:32.0) Gecko/20100101 Firefox/32.0 Graphics -------- Adapter Description: AMD Radeon HD 7900 Series Adapter Drivers: aticfx64 aticfx64 aticfx64 aticfx32 aticfx32 aticfx32 atiumd64 atidxx64 atidxx64 atiumdag atidxx32 atidxx32 atiumdva atiumd6a atitmm64 Adapter RAM: 3072 Device ID: 0x6798 Direct2D Enabled: true DirectWrite Enabled: true (6.3.9600.16384) Driver Date: 4-17-2014 Driver Version: 14.100.0.0 GPU #2 Active: false GPU Accelerated Windows: 1/1 Direct3D 10 Vendor ID: 0x1002 WebGL Renderer: Google Inc. -- ANGLE (AMD Radeon HD 7900 Series Direct3D9Ex vs_3_0 ps_3_0) windowLayerManagerRemote: false AzureCanvasBackend: direct2d AzureContentBackend: direct2d AzureFallbackCanvasBackend: cairo AzureSkiaAccelerated: 0
Blocks 899785 IMO
Turned OMTC back on two days ago, just had a BSOD after closing a tab with a flash plugin. Catalyst 14.4 (WHQL).
Blocks: 899785
Severity: normal → critical
Status: UNCONFIRMED → NEW
Ever confirmed: true
Keywords: crash
I'm going to suggest blacklisting the new AMD drivers for D3D11 and using D3D9 there for now as a solution for now. This is after all a driver problem.
Seems to be a little drastic. It only BSODs with OMTC enabled. Furthermore, OMTC is disabled by default so the user has to toggle it on as I did.
That's true. I have BSODs only with OMTC enabled. With disabled everything work as it should.
If you think it's the drivers, are you in contact with AMD about this? AMD would probably recommend disabling OMTC as their solution which doesn't seem acceptable. I turned on OMTC yesterday and didn't get a BSOD until this morning. Could be that it came after I cold started my machine this morning and was watching a Flash video. I since reverted back to the 13.12 drivers which I don't have any OMTC problems with. As you know there have been a few problems with different versions of AMD drivers. These have disappeared over the months. I'm not sure if new drivers or fixes in Firefox did the trick. Again, IMO Mozilla should reach out to AMD to resolve this issue. I don't think the user has very much leverage as you would.
Just wanted to add something. The 14.x drivers have the so called Mantle support. Is it possible that with OMTC enabled Firefox is using this feature, maybe unbeknownst to you? Seems like a shot in the dark but thought I'd throw it into the mix.
(In reply to Gary [:streetwolf] from comment #8) > Seems to be a little drastic. It only BSODs with OMTC enabled. > Furthermore, OMTC is disabled by default so the user has to toggle it on as > I did. We're turning on OMTC this week, and killing non-OMTC completely for acceleration with the next report :). There's no way we're using Mantle. I'm working on reaching out to AMD, but traditionally they've been very hard to reach. We have a patch for optimus which might/will hopefully fix this issue as well.
"We have a patch for optimus which might/will hopefully fix this issue as well." Is this available to test?
FYI... I ran the inbound build https://hg.mozilla.org/integration/mozilla-inbound/rev/812ba047a9af which I believe contains the optimus patch you mentioned. Sad to say I immediately got a BSOD watching a Flash video.
Not the one I meant.. can you do me a favor? I should add the blacklist entry soon, we enabled this by default now and a BSOD is a terrible experience, even for a nightly build. Does the problem go away if you set gfx.direct2d.disabled to true?
I did not encounter this bug yet in my testing. But if the problem goes away by disabling Direct2D we should land this patch before the nightly build tomorrow. If this doesn't fix the problem we should blacklist all acceleration features on the new AMD drivers. In the meanwhile I'm attempting to contact AMD.
Assignee: nobody → bas
Status: NEW → ASSIGNED
Attachment #8424408 - Flags: review?(bjacob)
Yes, there are no BSOD or driver-reset in our 1 hour test(playing flash video) with setting set gfx.direct2d.disabled to true.
I've been having luck disabling HWA using Flash settings, not Fx's HWA. So far no BSODs. I've tried this in the past with no success. Why it now appears to have fixed my BSODs (so far) is anyone's guess. As far as disabling 2D that really slows some things down and if anything should be a temporary workaround and not a solution. I filed a report with AMD but as you said they are not very responsive.
(In reply to Gary [:streetwolf] from comment #18) > I've been having luck disabling HWA using Flash settings, not Fx's HWA. So > far no BSODs. I've tried this in the past with no success. Why it now > appears to have fixed my BSODs (so far) is anyone's guess. > > As far as disabling 2D that really slows some things down and if anything > should be a temporary workaround and not a solution. I filed a report with > AMD but as you said they are not very responsive. I know, but you have to understand there's very little options for us. We need to move forward with OMTC for all kinds of reason, and the on-main-thread code for D3D10 and D3D9 should be removed next cycle for clean-up purposes. We can't hold all this progress back because of an AMD driver bug :-(. I'll do my best. Of course individual users can force-enable D2D and disable Flash acceleration.
I fully understand Bas. If disabling Flash HWA is the solution does that give you more insight into the problem? Can it be a Flash vs AMD 14.4 driver problem?
Comment on attachment 8424408 [details] [diff] [review] Blacklist Direct2D for Catalyst 14.4 Review of attachment 8424408 [details] [diff] [review]: ----------------------------------------------------------------- Assuming that driverversion >= 14.1.0.0 is the correct check for Catalyst >= 14.4 (I haven't checked) and that blacklisting is all what we can do for now here, r+.
Attachment #8424408 - Flags: review?(bjacob) → review+
Got a BSOD and I wasn't even watching a Flash video. I was looking at a slideshow on a news site. Maybe it was using Flash. I don't remember the site. Had OMTC and D2D enabled. I went back to the 13.12 CATS which have never caused me any problems with Fx.
https://hg.mozilla.org/mozilla-central/rev/1c2e758704ad
Status: ASSIGNED → RESOLVED
Closed: 11 years ago
Resolution: --- → FIXED
Target Milestone: --- → mozilla32
Shouldn't this be a "leave open" until the AMD driver issue has a full resolution?
Is this only a 32bit Nightly issue because I run the x64 version of Nightly exclusively (no other versions installed) on x64 Windows 8.1 Pro and I have never seen this issue. I have OMTC enabled (1/1 Direct3D 11 (OMTC)), HWA + DirectWrite enabled, etc. I'm running the v14.4 AMD drivers w/ a single HD7850 installed. In fact, after force enabling D2D, I'm watching a full screen flash video right now.
(In reply to timbugzilla from comment #25) > Shouldn't this be a "leave open" until the AMD driver issue has a full > resolution? Agree. This is a stop gap 'solution' not a permanent fix. Couldn't a trap be put on those getting the BSOD to possibly narrow down the cause? I'd volunteer.
(In reply to Brian Carpenter [:geeknik] from comment #26) > Is this only a 32bit Nightly issue because I run the x64 version of Nightly > exclusively (no other versions installed) on x64 Windows 8.1 Pro and I have > never seen this issue. I have OMTC enabled (1/1 Direct3D 11 (OMTC)), HWA + > DirectWrite enabled, etc. I'm running the v14.4 AMD drivers w/ a single > HD7850 installed. In fact, after force enabling D2D, I'm watching a full > screen flash video right now. I've heard varying things about what causes the BSOD. One is that you have to be using an HD79xx series as I am. Another one is that enabling HWA in Flash itself is the cause. I found out it wasn't when I disabled it and got a BSOD.
I can reproduce a similar BSOD in atikmdag.sys on http://ie.microsoft.com/testdrive/performance/fishbowl/ within minutes on the following setp: OS: Win 8.1.1 AMD Catalyst: 14.3 and 14.4 Firefox Aurora 31.0a2 (2014-05-26) Shall I throw it in a new bugreport or can people confirm it's related to this? I can supply a few minidumps if needed.
I get the feeling that any BSODs posted here won't even be looked at since the Devs decided that the problem is fixed by blacklisting the 14.4 drivers. The 14.6 beta drivers are out btw, I'll give them a try and see what happens. I never had any problems with the Fishbowl demo.
(In reply to Gary [:streetwolf] from comment #30) > I get the feeling that any BSODs posted here won't even be looked at since > the Devs decided that the problem is fixed by blacklisting the 14.4 drivers. > The 14.6 beta drivers are out btw, I'll give them a try and see what > happens. > > I never had any problems with the Fishbowl demo. I wouldn't say I'm not looking at this. I'm very interested in reducing the amount of drivers/devices we blacklist. And I'd love to hear if the new drivers fix it. But there's not much we can do about the driver bug itself :).
(In reply to AK from comment #29) > I can reproduce a similar BSOD in atikmdag.sys on > http://ie.microsoft.com/testdrive/performance/fishbowl/ within minutes on > the following setp: > > OS: Win 8.1.1 > AMD Catalyst: 14.3 and 14.4 > Firefox Aurora 31.0a2 (2014-05-26) > > Shall I throw it in a new bugreport or can people confirm it's related to > this? > > I can supply a few minidumps if needed. Can you check whether disabling OMTC makes that no longer happen? That test puts a lot of load on your card. A BSOD with a test like that could also be an example for example of overheating or other GPU bugs.
Catalyst 14.6 beta version was officially out. In the release note, there isn't any special note for stability fix on the browsers. (Of course, in the most update, AMD doesn't appeal minor fixes...) I test new driver with OMTC, D2D acceleration and Flash's HW acceleration now...
The AMD 14.6 beta drivers appear (so far) to have fixed my BSOD problems. I started a thread on Mozillazine and a few other folks also report no more BSOD's: http://forums.mozillazine.org/viewtopic.php?f=23&t=2837351
Yes, 14.6 beta is very stable(no BSOD and driver reset) for me with D2D and Flash HW acceleration.
I haven't been able to reproduce the issue either with 14.6 drivers, going back to 14.4 immediately triggered it again.
Catalyst 14.6 Beta (driver version: 14.200.0.0, date: 2014-05-22) has no BSOD with Nightly 32.0a1 (build: 20140529030207) which enables both OMTC and D2D.
I'll adjust the blacklisting range. My apologies for those of you who experienced disturbances due to the BSOD :( It is the very worst crash to have.
You need to log in before you can comment on or make changes to this bug.

Attachment

General

Creator:
Created:
Updated:
Size: