Closed Bug 1168018 Opened 10 years ago Closed 3 years ago

Firefox page element rendering issue with D2D 1.1 enabled, with NVIDIA 540M graphics card.

Categories

(Core :: Graphics, defect, P3)

Product:
Core
Component:
Graphics
Version:
38 Branch
Type:
defect

Tracking

()

Status:
RESOLVED FIXED

People

(Reporter: smartfon.reddit, Assigned: bas.schouten, NeedInfo)

References

Details

(Whiteboard: [gfx-noted][platform-rel-nVidia])

Attachments

(9 files)

GIF that displays the rendering problem
163.79 KB, image/gif
Details
Screenshot of UI artifacts.
33.02 KB, image/png
Details
Memory Measurement on time of artefacts
499.47 KB, application/gzip
Details
about-support.json
9.58 KB, application/json
Details
about:support
11.13 KB, text/plain
Details
Another example of artifacts
90.05 KB, image/png
Details
artifacts2.PNG
38.92 KB, image/png
Details
about:support nightly
6.87 KB, text/plain
Details
Windows artifacts
27.78 KB, image/jpeg
Details
User Agent: Mozilla/5.0 (Windows NT 6.1; WOW64; rv:38.0) Gecko/20100101 Firefox/38.0 Build ID: 20150513174244 Steps to reproduce: Enable Hardware Acceleration and use NVIDIA GeForce 540M graphics. The issue only happens when the HWA is enabled. Graphics drivers are up to date. Problem happens with old drivers too. Firefox stable 38.0.1. Has been happening for at least 2 years with old versions of FX. Actual results: Visual demo of the problem. Album has 4 images. First one is a GIF https://imgur.com/sRoOwTy,Io1hQeO,BDKhS34,Dz3gOLS#0 Problem 1. When I switch to a new tab, very often parts of the page are blank. I have to scroll the page up or down, or hover the mouse over the missing areas to start their rendering. Problem 2. While typing text in a text field, the letters I type turn blank and it looks like as if I hit the backspace to remove them. A second later they reappear automatically. Problem 3. If I scroll the text that I type in a text field, about 90% of the text field turns black. Problem 4. Some entries in about:config report wrong names. An entry might show as ABCD but in reality that's the EFGH entry. When I modify the entry that's when it shows its correct name. This is a big problem because it creates a risk of editing wrong entries. Problem 5. Closing tabs doesn't fully remove the tab's image. The tab is gone but the tab's image stays there for few seconds, or until I click on other tabs. Problem 6. Browser crashes often when I open certain links. The crashing links seems random at the moment, but I suspect links with some animation crash more often. The longer the browser stays open, the worse the problem gets. The longer I leave the tab open and unattended, the more likely it is to cause rendering issues when I switch to that tab. Expected results: None of the above mentioned 5 problems should happen. The page should render properly. Browser should not crash. Browser should have a proper support for NVIDIA 540M graphics card.
Please remove the previous post. The about:support graphics details are wrong. That was done when the HWA is disabled. The following is the correct info with HWA enabled: Adapter Description Intel(R) HD Graphics 3000 Adapter Description (GPU #2) NVIDIA GeForce GT 540M Adapter Drivers igdumd64 igd10umd64 igd10umd64 igdumd32 igd10umd32 igd10umd32 Adapter Drivers (GPU #2) nvd3dumx,nvwgf2umx,nvwgf2umx nvd3dum,nvwgf2um,nvwgf2um Adapter RAM Unknown Adapter RAM (GPU #2) 1024 ClearType Parameters Gamma: 2200 Pixel Structure: R ClearType Level: 100 Enhanced Contrast: 50 Device ID 0x0116 Device ID (GPU #2) 0x0df4 Direct2D Enabled true DirectWrite Enabled true (6.2.9200.17292) Driver Date 1-30-2015 Driver Date (GPU #2) 4-8-2015 Driver Version 9.17.10.4101 Driver Version (GPU #2) 9.18.13.5012 GPU #2 Active false GPU Accelerated Windows 1/1 Direct3D 11 (OMTC) Subsys ID 51401558 Subsys ID (GPU #2) 51401558 Vendor ID 0x8086 Vendor ID (GPU #2) 0x10de windowLayerManagerRemote true AzureCanvasBackend direct2d 1.1 AzureContentBackend direct2d 1.1 AzureFallbackCanvasBackend cairo AzureSkiaAccelerated 0
You mentioned that there are also crashes happening. Could you please go to about:crashes and paste the contents in a comment?
Flags: needinfo?(smartfon.reddit)
I have almost the same graphics hardware (NVIDIA GeForce GT 550M instead of 540M) and I dont get these issues.
Component: Untriaged → Graphics
Product: Firefox → Core
(In reply to alex_mayorga from comment #3) > Could you please go to about:crashes and paste the contents in a comment? What type of private information do crash reports contain ?
Nothing private.
With FX 37 and 38, crashes happened even while HWA was OFF. bp-812ec326-afc9-4282-a84b-9b5c82150525 5/25/2015 10:53 AM bp-6dc1aaba-32eb-4fc1-bf61-3d9642150525 5/25/2015 10:49 AM bp-3f0c7ce7-ec01-4e1c-98b2-aab462150525 5/25/2015 10:33 AM bp-5ab3b2db-16bd-4850-b22d-9670e2150524 5/24/2015 11:31 AM bp-4aa6db6b-1f6b-4fa0-aa92-ab8a22150524 5/23/2015 7:11 PM bp-241ab496-5a77-4678-9826-712442150522 5/22/2015 1:54 PM bp-27da3960-5384-4af3-a813-9fcff2150521 5/21/2015 11:00 AM bp-834e7960-c8f6-4ba7-ad32-a67932150521 5/21/2015 10:59 AM bp-89dc8c94-d684-44c1-9e01-194f22150519 5/18/2015 10:12 PM bp-1ecac947-ca00-4051-9967-909952150518 5/18/2015 3:12 PM bp-1b02ccf4-cda5-4e89-a687-1eac72150518 5/18/2015 3:12 PM bp-0004d1d4-c846-41b4-b78d-f9caa2150516 5/15/2015 5:17 PM bp-972dfd7a-0a66-4a55-bced-2c36d2150515 5/14/2015 6:56 PM bp-70b416f6-af53-4c99-8d07-c60ad2150515 5/14/2015 6:56 PM bp-d5258878-6782-4544-9720-2d3332150515 5/14/2015 6:55 PM bp-dc930ba0-3aa9-42c2-b57d-889232150515 5/14/2015 6:55 PM bp-321a5825-b93b-4af9-839c-490f42150514 5/14/2015 10:02 AM 3a773c77-91c4-4d28-bc35-cee30a812d46 5/2/2015 1:22 PM bp-8e369e1c-7032-4f48-9c36-a72972150502 5/1/2015 7:05 PM bp-bf16c188-ca43-4308-9451-c2b3f2150501 5/1/2015 9:15 AM bp-76a17bcc-7606-4ffc-b2ca-68e152150430 4/29/2015 11:45 PM cf2222ea-a642-4900-b481-d6d3cda2f2e1 4/29/2015 12:51 PM bp-10ac8064-1a42-4547-b6da-c996a2150429 4/28/2015 11:05 PM bp-cf468523-8a46-4376-93ae-19c642150426 4/26/2015 3:21 PM bp-9e201eba-9004-40cc-b483-9c9292150426 4/25/2015 10:46 PM bp-f2308263-2f9a-47f5-b6ce-42f1d2150426 4/25/2015 10:04 PM bp-dbe2d9a7-ed90-4035-81ac-9e82e2150426 4/25/2015 9:35 PM bp-60e03d39-5295-4618-b619-6f3512150426 4/25/2015 8:53 PM bp-5b5b6ac8-a64c-4880-9f82-2f9ad2150424 4/23/2015 6:34 PM bp-2f8d20e7-f1f3-40fd-a62a-8c24d2150423 4/23/2015 12:00 AM bp-16816010-0d7e-4e29-bcb7-3508a2150422 4/21/2015 11:34 PM bp-efb5b7ad-c4ed-4d24-8eee-89f642150421 4/20/2015 11:05 PM 0b6a38b2-0775-4abf-bca6-c769970698f0 4/20/2015 9:33 PM bp-d7ac2c71-aff5-4ad6-a0f2-2ac662150421 4/20/2015 9:21 PM bp-607bafb6-9549-443e-97ff-15d082150421 4/20/2015 7:52 PM bp-5afcd653-db8f-46a2-afa5-673be2150421 4/20/2015 7:52 PM bp-4eeb4162-af9d-489b-bbd7-035a12150419 4/19/2015 2:51 PM bp-34121ff2-aa00-4b96-ad95-8e2792150419 4/18/2015 11:10 PM bp-1f615b86-3b4a-4e47-a807-d31642150419 4/18/2015 8:17 PM bp-33612246-9474-45ef-8c51-80dc72150418 4/17/2015 6:05 PM bp-69e23dcc-38e5-41ec-a28d-838362150418 4/17/2015 6:00 PM bp-ae8102e2-c72a-4899-987f-c3cac2150418 4/17/2015 5:59 PM bp-d5bbd367-99b1-43ff-955a-3141d2150417 4/17/2015 10:54 AM bp-071359f6-d971-4d82-8efe-eef182150414 4/13/2015 10:43 PM bp-503e91ae-085a-4c0e-b6e1-f3e172150414 4/13/2015 8:01 PM bp-8e95bb95-c11b-4863-9906-12cfe2150413 4/12/2015 11:12 PM bp-7cb5bc08-a2ee-457f-923f-8c9222150412 4/12/2015 12:32 PM bp-ee86d18c-d291-4deb-8b0b-3232c2150412 4/11/2015 9:52 PM bp-d6bfde46-1337-4ff7-804e-dc8c82150412 4/11/2015 9:51 PM bp-81c6ea71-7d1b-48cb-9168-dccb92150412 4/11/2015 7:16 PM bp-8a3196ef-a2b9-4dc6-9d63-836a72150411 4/10/2015 7:40 PM 0c438864-bcbf-4277-a611-075454512a96 4/9/2015 7:37 PM bp-84cd2608-b93d-413d-b119-bdd752150408 4/7/2015 10:50 PM bp-c8e45061-4f58-4728-b2bc-7dfee2150407 4/7/2015 9:56 AM caa21e64-954e-4ada-bfd5-54c585710b64 4/7/2015 9:56 AM 05dc1b0c-1957-4c52-987c-4d6de9d25f8b 4/7/2015 9:28 AM b3ccd3c1-d5ae-40e6-95d4-58ba3e323dc5 4/6/2015 10:19 PM 0cc035c1-a9ae-407f-8f2f-fab7864e29c8 4/6/2015 4:15 PM c40468bb-a62b-4254-a9ec-0bc167a51aa4 4/5/2015 11:05 PM bp-b45640d4-8577-4167-9618-e89362150406 4/5/2015 10:58 PM bp-e301a3f2-3f5e-445e-9d52-b4c732150406 4/5/2015 10:58 PM fe247bee-c971-445b-be68-59910896de33 4/5/2015 10:57 PM 44cabc7e-1fc8-4aa6-9767-645fd40694d0 4/5/2015 6:18 PM bp-3c01e7ad-8b2f-4346-ba28-193e82150406 4/5/2015 6:13 PM bp-1ada1fe5-a0c9-46e8-aba8-8252c2150404 4/4/2015 4:42 PM bp-11df4d4a-5504-4e84-8364-8c3bd2150404 4/3/2015 8:42 PM bp-b414b0a0-f8ce-4b21-9311-e03c52150404 4/3/2015 6:17 PM bp-17fae3c0-27bb-4542-bd37-7bb8d2150322 3/22/2015 1:38 PM d314db46-7066-4cbb-92ae-9a64f63f8f30 3/15/2015 11:05 PM bd40f778-9978-42c6-ba35-ccedd052ba44 3/15/2015 11:05 PM bp-64f1b360-c7b2-4e79-8615-aa56b2150316 3/15/2015 11:04 PM 084a978c-7fb4-47c9-8766-b90843ef440d 3/15/2015 11:04 PM 0b88ceb2-699b-4b40-a807-25300887b4fe 3/15/2015 11:03 PM c93e890e-7bf1-4efd-8a8c-7b894f154b8d 3/15/2015 11:03 PM bp-ef40153c-bc9d-4342-83ec-2c7992150219 2/19/2015 2:58 PM bp-34071c9e-55cf-4b21-a658-17e362150219 2/19/2015 2:54 PM bp-7486a108-3d00-430c-8af1-a66702150219 2/19/2015 2:50 PM bp-a4d050e0-55cd-4281-bb1a-380ed2150219 2/19/2015 2:47 PM bp-5bc48f9e-aba8-41eb-905c-f8ad72150219 2/19/2015 2:43 PM bp-75d1b949-60fd-4744-b37b-97bb02150219 2/19/2015 2:40 PM bp-6cd0a8a6-3826-4dab-bfc5-8bdaa2150219 2/19/2015 2:36 PM bp-3c47e12d-a119-4d36-953a-4ad5e2150219 2/19/2015 2:33 PM bp-86a70846-663b-4776-a276-5abe22150219 2/19/2015 2:30 PM bp-a23fdcce-1ce2-4380-bf00-271782150209 2/8/2015 8:56 PM eda3e85c-7d43-4ea3-bd2a-d1dfe9c08b8b 2/6/2015 2:03 PM 1a992e96-806a-407e-b4b0-bb8db97f380d 2/6/2015 2:03 PM bp-0d894450-2920-457b-af84-913782150206 2/6/2015 2:03 PM 9483c2a3-1630-42d0-914b-27f7960b7865 2/6/2015 2:01 PM bp-fc025fa0-b18b-4478-aad2-ef9b82150206 2/6/2015 2:01 PM bee8dad8-5485-4ad5-b264-e3c6415d0a8b 2/6/2015 2:01 PM bp-2d71eb7e-ba07-4642-9e56-1d9902150202 2/2/2015 11:25 AM e739c6fa-4b0f-4a4b-95fe-559f549d2ae5 1/31/2015 1:12 PM bp-9cb9c364-bcf3-42e4-a8e8-803e62150131 1/31/2015 1:12 PM aa9e0574-24c4-4d2d-83af-a89e33acda67 1/31/2015 1:12 PM ca7fde53-0684-4fab-9a4f-433d9626b275 1/31/2015 1:12 PM 55868d15-4c01-468e-87d6-fd8fbe2d9061 1/31/2015 1:11 PM bp-f80aee01-e491-471f-8556-b33232150131 1/31/2015 1:11 PM
Flags: needinfo?(smartfon.reddit)
I was able to reproduce the crash. The crashes are caused by "Sticky Password" password manager program. Sticky Password is a program that you install on your PC. It then installs a browser plugin. If the Sticky browser plugin is enabled but Sticky's desktop process is not running, then there are no crashes. Once the desktop process is activated, that's when the browser crashes. It crashes 100% of the time when I click the mouse's scroll button to open the top image in a new tab, from this link http://www.techspot.com/news/60773-windows-10-build-10125-leak.html Here is an example of a crash report, but most of the time it doesn't show this Windows error. It just crashes and I only see Mozilla's crash reporter: ------------------------------------------- Problem signature: Problem Event Name: APPCRASH Application Name: plugin-container.exe Application Version: 38.0.1.5611 Application Timestamp: 55541a90 Fault Module Name: mozalloc.dll Fault Module Version: 38.0.1.5611 Fault Module Timestamp: 55540a1e Exception Code: 80000003 Exception Offset: 00001aa1 OS Version: 6.1.7601.2.1.0.256.48 Locale ID: 1033 Additional Information 1: 0a9e Additional Information 2: 0a9e372d3b4ad19135b953a78882e789 Additional Information 3: 0a9e Additional Information 4: 0a9e372d3b4ad19135b953a78882e789 ---------------------------------------------- Crashing is not what my original problem was. If we can return to NVIDIA rendering issue I would appreciate it. That has been happening for years. I can always use another password manager but I can't use a different GPU. I can open another BugZilla bug separately for Sticky Password if you want. I've reported another Sticky Password issue on Sticky Password's forums last year but they never replied or fixed the issue so I'm not going to bother with them anymore. The rendering issue with HWA and NVIDIA GPU still happens even without Sticky Password process.
layers.offmainthreadcomposition.enabled set to false fixes the rendering issues with HWA enabled. Am I missing any performance by disabling OMTC ? Will it be possible to disable this in future versions of Firefox?
(In reply to smartfon.reddit from comment #9) > layers.offmainthreadcomposition.enabled set to false fixes the rendering > issues with HWA enabled. > > Am I missing any performance by disabling OMTC ? Will it be possible to > disable this in future versions of Firefox? Yes, and it is a configuration that is now untested with some know issues, we will remove it entirely at some point. You said that you enabled HWA, you meant HWA was disabled by default? I am trying to see if we already run into this with the same driver and blacklisted HWA there. You can try to keep accelerated compositing but disable Dirct2d 1.1 drawing by going to about:config and setting: "layers.acceleration.force-enabled" to "true" "gfx.direct2d.use1_1" to "false" "gfx.content.azure.backends" to "direct2d,cairo" If there are still graphics issues, then please try to disable accelerated drawing (while keeping accelerated compositing) with the following settings in about:config: "layers.acceleration.force-enabled" to "true" "gfx.content.azure.backends" to "cairo" Please let us know how these configurations affect the bug.
Whiteboard: [gfx-noted]
HWA was ON be default but I disabled it last year to fix the rendering issue. Recently I noticed that the CPU was hitting 30% while watching a YouTube video and that would make the laptop really hot, so I decided to enable HWA again and find an alternative fix. The following setup works without issues: HWA = ON OMTC = enabled "layers.acceleration.force-enabled" to "true" "gfx.direct2d.use1_1" to "false" "gfx.content.azure.backends" to "direct2d,cairo"
Ok, it's D2D 1.1, then. Bas, I'm assigning this one to you so you since you are already working on D2D 1.1 stuff.
Assignee: nobody → bas
Summary: Firefox page element rendering issue with Hardware Acceleration (HWA) enabled, with NVIDIA 540M graphics card. → Firefox page element rendering issue with D2D 1.1 enabled, with NVIDIA 540M graphics card.
http://www.nvidia.com/download/driverResults.aspx/85819/en-us Can you please test the latest Nvidia 353.06 WHQL driver?
(In reply to NVD from comment #13) The problem persists . Reverted all the about:config changes, updated drivers and did a clean install, rebooted twice, still experiencing rendering issues.
Also seeing corruption issues with very similar hardware set-up: Adapter Description: Intel(R) HD Graphics 3000 Adapter Description (GPU #2): NVIDIA GeForce GT 540M Adapter Drivers: igdumd64 igd10umd64 igd10umd64 igdumd32 igd10umd32 igd10umd32 Adapter Drivers (GPU #2): nvd3dumx,nvwgf2umx,nvwgf2umx nvd3dum,nvwgf2um,nvwgf2um Adapter RAM: Unknown Adapter RAM (GPU #2): 2048 ClearType Parameters: D [ Gamma: 2200 Pixel Structure: R ClearType Level: 100 Enhanced Contrast: 200 ] D [ Gamma: 2200 Pixel Structure: R ClearType Level: 100 Enhanced Contrast: 300 ] Device ID: 0x0116 Device ID (GPU #2): 0x0df4 Direct2D Enabled: true DirectWrite Enabled: true (6.3.9600.17795) Driver Date: 3-20-2014 Driver Date (GPU #2): 4-8-2015 Driver Version: 9.17.10.3517 Driver Version (GPU #2): 9.18.13.5012 GPU #2 Active: false GPU Accelerated Windows: 1/1 Direct3D 11 (OMTC) Subsys ID: 05041025 Subsys ID (GPU #2): 05051025 Vendor ID: 0x8086 Vendor ID (GPU #2): 0x10de WebGL Renderer: Google Inc. -- ANGLE (Intel(R) HD Graphics 3000 Direct3D9Ex vs_3_0 ps_3_0) windowLayerManagerRemote: true AzureCanvasBackend: direct2d 1.1 AzureContentBackend: direct2d 1.1 AzureFallbackCanvasBackend: cairo AzureSkiaAccelerated: 0 (#0) Error: [D2D1.1] 4CreateBitmap failure Size(26,8276) Code: 0x80070057 (#1) Error: [D2D1.1] 4CreateBitmap failure Size(32,8200) Code: 0x80070057 (#2) Error: [D2D1.1] 4CreateBitmap failure Size(32,8200) Code: 0x80070057
I had this problem too, this fixes up but causes another problem. I did that and now my javascript is **** up. I dont know what caused this, but now on reddit.com I cant reply/logout (any javascript:void(0) links dont work). However I dont have this problem in private windows. Adapter Description NVIDIA GeForce GT 540M Adapter Drivers nvd3dumx,nvwgf2umx,nvwgf2umx nvd3dum,nvwgf2um,nvwgf2um Adapter RAM 2048 Device ID 0x0df4 Direct2D Enabled true DirectWrite Enabled true (6.3.9600.16384) Driver Date 5-11-2015 Driver Version 9.18.13.5286 GPU #2 Active false GPU Accelerated Windows 1/1 Direct3D 11 (OMTC) Subsys ID 16721043 Vendor ID 0x10de WebGL Renderer Google Inc. -- ANGLE (NVIDIA GeForce GT 540M Direct3D9Ex vs_3_0 ps_3_0) windowLayerManagerRemote true AzureCanvasBackend direct2d AzureContentBackend direct2d AzureFallbackCanvasBackend cairo AzureSkiaAccelerated 0
Still happening. I just upgraded to Windows 10 and it's even worse now. The rendering issue happens even with the modified D2D. I completely disabled HWA to make it somewhat more usable.
(In reply to smartfon.reddit from comment #18) > Still happening. I just upgraded to Windows 10 and it's even worse now. The > rendering issue happens even with the modified D2D. I completely disabled > HWA to make it somewhat more usable. Can you please attach your demonstration of the bug to this bug report? I know you posted an imgur link above but if that link ever expires we'll lose all context for what you're seeing. It's always preferable to attach something to the bug report whenever possible. Bas, what's needed to move this bug forward? I checked the hardware inventory and we don't have a similar device so getting a reproducible testcase internally is going to be difficult.
Flags: needinfo?(bas)
Scratch that, ignore Comment 21. Here is the current situation: Updated to FX40. All default settings and up to date NVIDIA causes the rendering issues. Now, this also causes rendering issues (it used to help): "layers.acceleration.force-enabled" to "true" "gfx.direct2d.use1_1" to "false" "gfx.content.azure.backends" to "direct2d,cairo" The next suggested config was this, but it causes the browser to crash when I type master password and hit OK in LastPass addon: "layers.acceleration.force-enabled" to "true" "gfx.content.azure.backends" to "cairo" (<---I think this line is specifically responsible for LastPass crash) Right now I have the OMTC disabled and HWA enabled, and there are no issues. OMTC=ON and HTA=OFF also fixes the problem. Some examples of crash caused by "gfx.content.azure.backends" to "cairo", after entering LastPass password. https://crash-stats.mozilla.com/report/index/28f15efb-a552-40f0-9ef7-018c62150811 https://crash-stats.mozilla.com/report/index/8675bdeb-0a0f-4713-b2b8-c2c482150811 https://crash-stats.mozilla.com/report/index/b9f5cd17-7be0-40f6-a4fa-6de1d2150811
355.60 does not solve rendering problem with default Firefox settings. It looks like with 355.60, the following suggestion solves the rendering issue again: "layers.acceleration.force-enabled" to "true" "gfx.direct2d.use1_1" to "false" "gfx.content.azure.backends" to "direct2d,cairo" However, with the new drivers, LastPass crashes the browser if OMTC is enabled. On my previous post I wrote that ONLY this "gfx.content.azure.backends" to "cairo" was causing LastPass to crash, which was true, but with latest drivers OMTC in general must be disabled for it not to crash. Here are new LastPass/OMTC crash reports https://crash-stats.mozilla.com/report/index/857ef01f-307a-4c80-ad73-d06c72150814 https://crash-stats.mozilla.com/report/index/58cb0f9d-6a67-49ad-9917-d941f2150814 https://crash-stats.mozilla.com/report/index/de2b96e1-4885-4fb5-ab20-ed0642150814 Note: mbae.dll and KeyCrypt32(2).dll are not crashing it. I've tested it without them.
Could this also be an issue with partial presents on nvidia hardware?
(In reply to Vladimir Vukicevic [:vlad] [:vladv] from comment #25) > Could this also be an issue with partial presents on nvidia hardware? I'm not sure what that means. Right now the OMTC is off so I can use the browser (everything feels more sluggish?). Once a new browser or driver update is pushed, I will let you know if the rendering and crashing issues are gone.
(In reply to Vladimir Vukicevic [:vlad] [:vladv] from comment #25) > Could this also be an issue with partial presents on nvidia hardware? It seems unlikely but it's not impossible. E_INVALIDARG is showing up in one of the about:supports but that's rather mysterious. Anthony/Kyle, can we get our hands on a machine that reproduces this? I'll really need one in order to do anything about it.
Flags: needinfo?(kfung)
Flags: needinfo?(bas)
Flags: needinfo?(anthony.s.hughes)
(In reply to Bas Schouten (:bas.schouten) from comment #27) > (In reply to Vladimir Vukicevic [:vlad] [:vladv] from comment #25) > Anthony/Kyle, can we get our hands on a machine that reproduces > this? I'll really need one in order to do anything about it. Might be best to try to track down the reporter's system first as that will most closely represent the environment where this seems to reproduce. @smartfon can you please tell us exactly what make and model of computer you are using?
Flags: needinfo?(smartfon.reddit)
Flags: needinfo?(kfung)
Flags: needinfo?(anthony.s.hughes)
As an additional note, the Dell XPS L501X in the Toronto lab has a NVidia Geforce GT 420M which is from the same chipset family as the GT 540M. Kyle, it might be worth seeing if that reproduces this.
Flags: needinfo?(kfung)
> @smartfon can you please tell us exactly what make and model of computer you are using? Sager NP5160 (aka Clevo W150HNQ). Swapped the default HDD with Crucial C300 SSD. Let me know if you need me to run any tests. Are there any alternatives to disabling OMTC to fix the LastPass crashes?
Flags: needinfo?(smartfon.reddit)
I was given a beta version of LastPass by their support team, and it works fine with OMTC enabled. The original problem with page rendering is still happening.
I don't have access to the test machines in the Toronto office anymore. You might want to look for an employee in Toronto.
Flags: needinfo?(kyle_fung) → needinfo?(anthony.s.hughes)
(In reply to Kyle Fung from comment #32) > I don't have access to the test machines in the Toronto office anymore. You > might want to look for an employee in Toronto. Okay, thanks. Milan can you take a look at this?
Flags: needinfo?(anthony.s.hughes) → needinfo?(milan)
I would like to jump in on this bug, as the same is happening in the latest Firefox version. The problem is it does not seem to be consistent, it only appears after a certain amount of time and reopening Firefox fixes the problem for a while again. But it can get so bad that you can barely tell that it was a browser window. FF Ver: 43.0.1 (32-bit) CPU: i7-3770K (HD4000) recognizzed as GPU#2 (inactive on about:support) GPU: NVIDIA GTX960 (GIGABYTE 2GB OC version) recognized as GPU#1 OS: Windows 10 64-bit Education (new install not upgraded) Resolution: 4K Intel are the latest Windows Update version (10.18.10.4276) because their driver versioning is messed up and the Intel utility wants to install older drivers. NVIDIA drivers are up to date. NVIDIA is version 359.06 (installed by Geforce experience) As soon as it acts up again I will record (ShadowPlay) some of it and try to screw it up as bad as I can. It will be art. Graphic section of about:support Adapter-RAM 2048 Adapter-RAM (GPU #2) Unknown Adapterbeschrijving NVIDIA GeForce GTX 960 Adapterbeschrijving (GPU #2) Intel(R) HD Graphics 4000 Adapterstuurprogramma’s nvd3dumx,nvwgf2umx,nvwgf2umx,nvwgf2umx nvd3dum,nvwgf2um,nvwgf2um,nvwgf2um Adapterstuurprogramma’s (GPU #2) igdumdim64 igd10iumd64 igd10iumd64 igdumdim32 igd10iumd32 igd10iumd32 Asynchroon pannen/zoomen geen Datum stuurprogramma 11-24-2015 Datum stuurprogramma (GPU #2) 7-10-2015 Device-ID 0x1401 Device-ID (GPU #2) 0x0162 Direct2D ingeschakeld true DirectWrite ingeschakeld true (10.0.10586.0) GPU #2 actief false GPU-versnelde vensters 1/1 Direct3D 11 (OMTC) Ondersteunt hardwarematige H264-decodering Yes Stuurprogrammaversie 10.18.13.5906 Stuurprogrammaversie (GPU #2) 10.18.10.4252 Subsys-ID 36901458 Subsys-ID (GPU #2) 0000000c Vendor-ID 0x10de Vendor-ID (GPU #2) 0x8086 WebGL-renderer Google Inc. -- ANGLE (NVIDIA GeForce GTX 960 Direct3D11 vs_5_0 ps_5_0) windowLayerManagerRemote true AzureCanvasBackend direct2d 1.1 AzureContentBackend direct2d 1.1 AzureFallbackCanvasBackend cairo AzureSkiaAccelerated 0 All about:config settings are left alone (except the addon signing one and a whole bunch for ff sync and some GFX ones like driver version and the like) NVIDIA 3D vision plugins (2) are installed. Just like an Acrobat plugin, Silverlight, Quicktime,Office 2015, Intel ID protection, PrimeTime, OpenH264, iTunes App detect. I would be glad to provide any other information needed.
I have a feeling at least some of this could be the Nvidia flavor of bug 1234616.
Flags: needinfo?(milan)
See Also: → 1234616
Well it is happening right now, I made a little screenshot (attached) and a video. (how do you guys want video? 89MB 4K) For now: https://youtu.be/ym67ZVQ8y7M It was doing it on the my main window and I then opened a new (private) window and then this show happened (after navigating to reddit.com and clicking the first thing there) And the about:support part about gfx (also included in the video inclusing all other stuff on the support page) Adapter-RAM 2048 Adapter-RAM (GPU #2) Unknown Adapterbeschrijving NVIDIA GeForce GTX 960 Adapterbeschrijving (GPU #2) Intel(R) HD Graphics 4000 Adapterstuurprogramma’s nvd3dumx,nvwgf2umx,nvwgf2umx,nvwgf2umx nvd3dum,nvwgf2um,nvwgf2um,nvwgf2um Adapterstuurprogramma’s (GPU #2) igdumdim64 igd10iumd64 igd10iumd64 igdumdim32 igd10iumd32 igd10iumd32 Asynchroon pannen/zoomen geen Datum stuurprogramma 12-16-2015 Datum stuurprogramma (GPU #2) 8-17-2015 Device-ID 0x1401 Device-ID (GPU #2) 0x0162 Direct2D ingeschakeld true DirectWrite ingeschakeld true (10.0.10586.0) GPU #2 actief false GPU-versnelde vensters 1/1 Direct3D 11 (OMTC) Ondersteunt hardwarematige H264-decodering No; DXVA2D3D9 crashes detected in the past Stuurprogrammaversie 10.18.13.6143 Stuurprogrammaversie (GPU #2) 10.18.10.4276 Subsys-ID 36901458 Subsys-ID (GPU #2) 0000000c Vendor-ID 0x10de Vendor-ID (GPU #2) 0x8086 WebGL-renderer Google Inc. -- ANGLE (NVIDIA GeForce GTX 960 Direct3D11 vs_5_0 ps_5_0) windowLayerManagerRemote true AzureCanvasBackend direct2d 1.1 AzureContentBackend direct2d 1.1 AzureFallbackCanvasBackend cairo AzureSkiaAccelerated 0 (#0) Error DXVA2D3D9 just crashed; hardware video will be disabled. (#132) Error DXVA2D3D9 video decoding is disabled due to a previous crash. (#133) Error DXVA2D3D9 video decoding is disabled due to a previous crash. (#134) Error DXVA2D3D9 video decoding is disabled due to a previous crash. (#135) Error DXVA2D3D9 video decoding is disabled due to a previous crash. (#136) Error DXVA2D3D9 video decoding is disabled due to a previous crash.
Let us set up a "special" Firefox build that could give us more information about these failures, and hopefully you could run it and tell us what messages show up when you run into these problems.
(In reply to Milan Sreckovic [:milan] from comment #36) > I have a feeling at least some of this could be the Nvidia flavor of bug > 1234616. Looking at the video, I don't think there's any relation between the two. (In reply to Milan Sreckovic [:milan] from comment #39) > Let us set up a "special" Firefox build that could give us more information > about these failures, and hopefully you could run it and tell us what > messages show up when you run into these problems. I have absolutely no idea what to put in that build, the fact they're not seeing any special errors in about:support suggests nothing wildly unexpected is happening, those things have all been annotated with gfxCriticalErrors by now.. The video suggests some serious presentation issues. I'd be curious to know if this is happening with latest nightly where we've disabled partial presentation when using NVidia Hardware. I suspect this might be a partial present issue.
Any chance you could test this on Nightly, with HWA enabled?
Flags: needinfo?(smartfon.reddit)
Sure! 64-bit build or 32-bit build?
(In reply to EraYaN from comment #42) > Sure! 64-bit build or 32-bit build? Either would do just fine! Whatever you're normally experiencing the problem on.
The current nightly does not appear to have the same issue, but it's hard to say. It is hard to prove something does not exist. The issue is I can't run it long enough without crashing on me to really get an idea. Will try some more.
I am on that driver version as of right now, I wasn't at my computer for long enough this week to see anything yet/again. Maybe tomorrow, (I can leave the browser on without quitting something like 14 hours of active usage).
This is a verbose memory measurement done on about:memory at the time the artifact started (about 13 hours in since start). It seems firefox is a little "leaky".
Attached file about-support.json —
It still happens in the latest stable: http://i.imgur.com/OzXDSJ0.png (This is after 11 hours of use. The RAM climbed from a decent level: 400 to about 1.6GB. That seems to be strongly related. The drivers have not fixed it, sadly. I can't run the nightly long enough, before it **** out and corrupts some of the SQLite files used to store addon settings and such. So I have stopped using it for now. I have attached the memory measurement and now the json exported from about:support.
(In reply to Milan Sreckovic [:milan] (PTO 1/29) from comment #41) > Any chance you could test this on Nightly, with HWA enabled? Hello, OP here. The problem still happens with stable and Nighly versions (both 32 and 64 bit). Latest NVIDIA drivers 361.75 WHQL. It appears to be worse on Nightly version. On Nightly, I see a lot of small black boxes scattered around the page but they disappear once I scroll. I haven't seen those in stable version. In stable version, the whole page or most of it can turn white until I scroll the page.
Flags: needinfo?(smartfon.reddit)
Having the same problem since Firefox 46 with a GTX 970 driver version 361.91. H264 decoding is intentionally disabled due to problem with playback of some videos. No crashes unlike the OP but I get the rectangle blocks on pages that pop in/out on mouseover or scroll. It only occur when the browser has been up for hours. Its rather inconsistent and I could go for a whole day without the bug occurring. Once the problem occurs in a tab it seems to affect newly opened tabs. My old tabs continue working just fine for some reason. Adapter Description: NVIDIA GeForce GTX 970 Adapter Drivers: nvd3dumx,nvwgf2umx,nvwgf2umx,nvwgf2umx nvd3dum,nvwgf2um,nvwgf2um,nvwgf2um Adapter RAM: 4095 Asynchronous Pan/Zoom: wheel input enabled ClearType Parameters: Gamma: 2200 Pixel Structure: R ClearType Level: 100 Enhanced Contrast: 50 Device ID: 0x13c2 Direct2D Enabled: true DirectWrite Enabled: true (10.0.10586.0) Driver Date: 2-8-2016 Driver Version: 10.18.13.6191 GPU #2 Active: false GPU Accelerated Windows: 1/1 Direct3D 11 (OMTC) Subsys ID: 00000000 Supports Hardware H264 Decoding: No; Hardware video decoding disabled or blacklisted Vendor ID: 0x10de WebGL Renderer: Google Inc. -- ANGLE (NVIDIA GeForce GTX 970 Direct3D11 vs_5_0 ps_5_0) windowLayerManagerRemote: true AzureCanvasBackend: direct2d 1.1 AzureContentBackend: direct2d 1.1 AzureFallbackCanvasBackend: cairo AzureSkiaAccelerated: 0
When the problems do happen, assuming you don't crash, are there any error messages in the graphics section of about:support?
No error message in about:support. No crashes. However I do not use any password managers which looks to be the cause of crash for smartfon. Also I do not get any UI artifacts but that might just be due to using the devedition theme instead of australis.
I tried setting "layers.async-pan-zoom.enabled" to false for a few days since this is new in Firefox 46. I have not encountered this bug since then.
It's happening right now (I'm posting this in a bugged browser) There are no errors in about:support. The browser also still runs. Although the browser is hogging 1.3 GB of ram for 18 active tabs. (this seems to be due to long run time) Adapter-RAM 2048 Adapter-RAM (GPU #2) Unknown Adapterbeschrijving NVIDIA GeForce GTX 960 Adapterbeschrijving (GPU #2) Intel(R) HD Graphics 4000 Adapterstuurprogramma’s nvd3dumx,nvwgf2umx,nvwgf2umx,nvwgf2umx nvd3dum,nvwgf2um,nvwgf2um,nvwgf2um Adapterstuurprogramma’s (GPU #2) igdumdim64 igd10iumd64 igd10iumd64 igdumdim32 igd10iumd32 igd10iumd32 Asynchroon pannen/zoomen geen Datum stuurprogramma 2-8-2016 Datum stuurprogramma (GPU #2) 8-17-2015 Device-ID 0x1401 Device-ID (GPU #2) 0x0162 Direct2D ingeschakeld true DirectWrite ingeschakeld true (10.0.10586.0) GPU #2 actief false GPU-versnelde vensters 1/1 Direct3D 11 (OMTC) Ondersteunt hardwarematige H264-decodering Yes Stuurprogrammaversie 10.18.13.6191 Stuurprogrammaversie (GPU #2) 10.18.10.4276 Subsys-ID 36901458 Subsys-ID (GPU #2) 0000000c Vendor-ID 0x10de Vendor-ID (GPU #2) 0x8086 WebGL-renderer Google Inc. -- ANGLE (NVIDIA GeForce GTX 960 Direct3D11 vs_5_0 ps_5_0) windowLayerManagerRemote true AzureCanvasBackend direct2d 1.1 AzureContentBackend direct2d 1.1 AzureFallbackCanvasBackend cairo AzureSkiaAccelerated 0
Kats, comment 54.
Flags: needinfo?(bugmail.mozilla)
(In reply to anonbugreport from comment #54) > I tried setting "layers.async-pan-zoom.enabled" to false for a few days > since this is new in Firefox 46. I have not encountered this bug since then. Can you (a) post a screenshot of the browser when it happens, and (b) turn layers.async-pan-zoom.enabled back to true and change the following preferences: apz.x_skate_size_multiplier -> 0.5 apz.x_stationary_size_multiplier -> 0.5 apz.y_skate_size_multiplier -> 0.5 apz.y_stationary_size_multiplier -> 0.5 and see if it still happens? Note that with these prefs set you will probably see more checkerboarding (i.e. content that's initially unpainted when you scroll it into view) but it should get painted without you having to mouse over it or interact with it in any way.
Flags: needinfo?(anonbugreport)
Flags: needinfo?(bugmail.mozilla)
(In reply to Kartikaya Gupta (email:kats@mozilla.com) from comment #58) > (b) turn layers.async-pan-zoom.enabled back to true and change the following > preferences: > apz.x_skate_size_multiplier -> 0.5 > apz.x_stationary_size_multiplier -> 0.5 > apz.y_skate_size_multiplier -> 0.5 > apz.y_stationary_size_multiplier -> 0.5 I do not get any rendering issues with these settings. I only seem to get temporary checkerboarding when using Firefox after waking up the PC from a long sleep. > (a) post a screenshot of the browser when it happens With the default apz size multipliers and the recent updates to either Firefox or Nvidia drivers the rendering issue still occurs but fixes itself in less then a second. Its like a single corrupted frame. I don't think I can get that screenshot. The previous behavior is similar to comment 0 and bug 1249080 examples. It can be triggered on the initial page load, scroll, and mouseovers. However unlike all both their report I do not get UI rendering issues and no about:support errors. All reporters seems to have Nvidia cards.
Flags: needinfo?(anonbugreport)
Thanks for the info. Reading through this bug again, I think the issue you (anonbugreport@mailinator) were seeing is different from previous issues in this bug. In particular you found that disabling APZ fixed the problem (per comment 54) but the reporter in comment 55 had APZ off ("Asynchroon pannen/zoomen geen") already and was still seeing the issue. Given that yours is a different issue and that you're not seeing it in recent Firefox any more (at least not longer than a single corrupted frame) I don't think there's much APZ-related going on here. However, you also mentioned that reducing the displayport multipliers helped, which means that the problem is likely to be related to large displayport sizes which can result in large texture uploads on Nvidia cards. I don't know if that helps Bas at all.
Kats, if large displays are one of the problem areas, I'm using a 3860 by 2160 screen. And I run my browser fullscreen lots of the time. And I have noticed that it only happens after a long time of having the browser opened. So the RAM usage climb constantly, so until it reaches a certain threshold it seems to be fine. It might be something RAM management related. Besides a browser without memory leaks would be cool as well.
Attached file about:support —
It seems I'm affected by the same bug as OP. After using Firefox for a few hours, I get occasional arifacts that look like this: http://i.imgur.com/rCiy8jL.png Setting "layers.async-pan-zoom.enabled" to false doesn't help. However, the following fixes the issue completely: "layers.acceleration.force-enabled" to "true" "gfx.direct2d.use1_1" to "false" "gfx.content.azure.backends" to "direct2d,cairo" I've attached my about:support page. Let me know if any other information would be useful.
By the way, my screen resolution is 1920x1080 so I doubt that it has anything to do with screen size. However, I do use 125% scaling. Perhaps this might have something to do with the issue?
Nikolai, do you see the same problems with Firefox Nightly, and a clean, new, profile?
Attached image artifacts2.PNG —
After updating to the latest NVIDIA drivers (365.19) and switching to Nightly with a clean profile (49.0a1 (2016-05-13)), I'm still getting occasional artifacts.
Attached file about:support nightly —
However, the about:support page suggests that Firefox is using Intel(R) HD Graphics 530 instead of NVIDIA GeForce GTX 960M, so perhaps the NVIDIA driver version is irrelevant?
Can you force the Nvidia graphics to be used?
I reset the three d2d settings to their defaults, forced Firefox to use Nvidia and restarted it. No artifacts after a few days. To make doubly sure, I changed it back to using integrated graphics and got artifacts after about an hour. So apparently the problem is with Intel graphics.
Attached image Windows artifacts —
After updating my Intel graphics drivers from 20.19.15.4390 to 20.19.15.4424, I stopped getting artifacts in Firefox. However, I started getting occasional artifacts in other Windows applications, as seen in the attached screenshot. Also, for some reason Windows doesn't consider it the latest and clicking "Update driver" in Device Manager reinstalls 20.19.15.4390 (after which I go back to having artifacts in Firefox but not anywhere else). I'm also quite sure that https://bugzilla.mozilla.org/show_bug.cgi?id=1260258 is the same issue.
Comment 71 - would we consider blocking 20.19.15.4390 in favour of 20.19.15.4424?
Flags: needinfo?(jmuizelaar)
What exactly would "blocking" entail?
Right - the blocking would stop us from using D2D when you have the "bad" driver. Things may be somewhat slower, because we're using more of the CPU and less of the GPU, but the users with those "bad drivers" would at least get the correct results. And once they update to a better version, they would get D2D and correctness.
That seems appropriate.
platform-rel: --- → ?
Whiteboard: [gfx-noted] → [gfx-noted][platform-rel-nVidia]
platform-rel: ? → ---
This bug still exists with the latest drivers. This issue isn't present while running Firefox with Intel iGPU, although that one now has an issue with playing GIFV videos https://bugzilla.mozilla.org/show_bug.cgi?id=1346857
Status: UNCONFIRMED → ASSIGNED
Ever confirmed: true
Status: ASSIGNED → RESOLVED
Closed: 3 years ago
Resolution: --- → FIXED
You need to log in before you can comment on or make changes to this bug.

Attachment

General

Creator:
Created:
Updated:
Size: