I'm experiencing similar problems: I'm using a clean install of Win10 Pro x64 with the latest Intel display and chipset drivers. About 5 times so far, (within a week), when playing a YouTube video in Firefox 39 (with Cisco h.264 plug-in disabled), after 30-40 minutes of video playback (a single long video), the machine may permanently freeze. The display would still have a picture, but the mouse would not respond and keyboard lights would not even respond when trying to toggle "caps lock" or "num lock" (a state requiring a hard reset).
I have made several experiments at trying to reproduce the issue. I managed to reproduce it once on the same day (with the same video in Firefox), and today (with the same video in Firefox) I managed to get not a hang but a permanent black out that would respond to the mouse for several seconds (over a total black background) and then freeze for several seconds and then repeat indefinitely (requiring a hard reset). Trying to open the OSD menu in the monitor (Dell P2414h) did work but it would close by itself when the total blank occurred. A subsequent check showed my system event log was filled with "Display driver igfx stopped responding and has successfully recovered." warnings with intervals of 5-20 seconds apart. The original freeze problem did not leave any message on the event viewer so there's the possibility this may be a different problem (or not). I will report if I get this particular problem again.
I've also tried downloading the same video and trying to play it locally with MPC-HC using both DXVA2 and Intel Quick Sync rendering options (through the built-in LAV filters) for several hours without any effect. I've tried to load 6 instances of MPC-HC and play video in parallel for an hour (DXVA2), but it didn't reproduce the issue. I tried playing 6 instances of the same YouTube video in Firefox for an hour without success so far (it doesn't seems to happen very consistently). I tried playing the video while running a video card stability test app and monitor temperatures but it did not help (the CPU temperatures didn't go over 60c even at maximum load, 35c idle).
An interesting observation in my first reproduction was that the video froze first but for 4-5 seconds the mouse was still responding. Then the mouse froze as well and the machine had to go through hard shutdown.
I have used the same machine for a year and a half now with Windows 7 without any problems whatsoever, it is only after installing windows 10 that the problem started. I did have an intermittent 1 second black-out issues with the current Dell monitor (that I bought a month ago) in Windows 7 as well, that happened randomly about once a day, but not with the previous monitor (Samsung 2333T), and that is a separate problem that I'm still investigating (doesn't seem to be related, and I believe I found a way around it, we'll see about that).
I'm strongly interested in helping solving this issue and willing to give any assistance needed. It is not clear yet how to reproduce reliably and will continue making more experiments to find out how to make the reproduction more consistent.
My test video (45 minute 1080p 50fps video) that got a freeze 3 times now.
Intel Pentium G3220 with Intel HD Graphics.
ASUS H81M-C Motherboard.
This issue is still happening for me as well on the newly released driver (8/5/15) and as of yet I still haven't found a way to fix this issue completely. I'm still getting around this issue by having my Nvidia GPU handle these programs instead of the HD 4600. I haven't tested Youtube myself, but I'm fairly certain that we are experiencing the same issue and that is quite concerning because HTML5 is being used more and more these days on everything, and if all sort of HTML5 content is affecting the HD 4600 this same way, well...that's just really bad. Hopefully this gets resolved by Intel in future drivers or maybe even by Microsoft, since it could be on their end too after all.
For a minimal test case (and to save on bandwidth), I downloaded the test video (same as the one I linked, but without audio) and playing it via a very simple dedicated HTML5 player page I've made (the video file is hosted on a local IIS web server). I'll be running it continuously, 24/7 in a loop with the latest Firefox, Chrome, IE and Edge all in parallel (at least for now, I'll try to simplify that in the future).
Using this setup I have successfully reproduced the freeze issue again today. Surprisingly it didn't happen by itself, even after up to 6 hours of continuous playback (though the videos were not visible on screen and running in a different virtual desktop). However, the second I opened a website (in FireFox) that launched a flash video it immediately froze (the mouse immediately stopped responding this time and keyboard lights were not responding) and a hard reset was required again. There was no error or warning information in the event viewer.
By now I can begin speculating that freezes may only occur when a video is actually visible on screen, and may be somehow related to web browsers. (Overnight I ran 6 instances of MPC-HC in a loop with http://freestone-group.com/video-card-stability-test.htm (video card stability test) in the background for about 7 hours and it did not reproduce the issue.).
I will continue to try simplifying the method and time taking for the reproduction of the issue.
(attached the minimal HTML5 player page I'm using)
index.html.zip 392 bytes
Some updates (all tests were done with the reduced HTML5 test case and the same test video):
Successful reproductions (within the last 2 days):
- after 12 hours running 1 instance in Firefox.
- after 1 hour running 4 instances in Chrome (multiple windows, not tabs).
- after 5 hours running 4 instances in Chrome (multiple windows, not tabs). Test was started immediately following the previous one.
- 7 hours running 10 instances in Edge (multiple windows, not tabs) - overnight, where the browsers were visible on screen and the computer was not in use (was stopped midway to start the test in Chrome instead).
In all 3 of the successful cases the videos were not visible on screen (running in a different virtual desktop) and in 2 of them the computer was unattended when the freeze happened (though in all the computer was in active use during some duration of the tests). This contradicts my speculations that video has to be visible on screen or even that any moving picture must be rendered on screen for this to happen.
I'm eager to simplify this even more, and also make sure this is actually a graphic driver problem and not a browser/other software/other hardware problem. I will try running it in a desktop video player again for more extended period of time then I tried before.
I'm also considering installing the Windows Hardware Lab Kit (HLK) for Windows 10 which has a DXVA compatibility test suites and stress test suites. The problem is that this would require downloading more than 10GB of data, has some learning curve, and I'm not sure if it will show anything useful that Intel hasn't tested themselves.
This information is intended to all Windows 10 users that are experiencing issues with a supported Intel® Graphics Controller (supported graphics)
See post # 93
I realize by now Intel are probably already looking into the problem (and may even solved it) but I thought I'd give a short update:
11/8 18:30 - 13/8 11:54 - 4 instances in chrome did not reproduce the problem. The test was stopped in the middle and the computer was restarted.
13/8 11:54 - 15/8 23:32 - 2 instances of MPC-HC with DXVA2 ran until video driver crashed and was reloaded twice (with a task bar and event viewer notifications). The running videos appeared to have frozen and did not response to start, stop and seek controls. I tried to reload the videos again in MPC-HC, several seconds afterwards, the whole system froze.
15/8 23:45 - 16/8 18:00 Reproduced system freeze again with 2 instances of MPC-HC. and did not renew the test.
16/8 21:20 - A sequence of blackouts and then a freeze while casually watching a video in Firefox (no intentional processes were run in the background).
Since 16/8 I did not run any tests. I wanted to see if I would get any freezes without running continuous videos. So far there have been none (though I did watch occasional YouTube videos).
These results seem to give a stronger evidence this might be a general DXVA issue (but still not 100%). I will renew the tests when a new driver is released.
Please post your results to the thread below.
Re: -NEW- Intel® Iris™, Iris™ Pro, and HD Graphics Production Driver for Windows* 10 64-bit 18.104.22.168.4256
See post # 93