1. This site uses cookies. By continuing to use this site, you are agreeing to our use of cookies. Learn More.

Adobe Flash HW Acceleration / AMD Overdrive driver bug - who else can reproduce this?

Discussion in 'Videocards - AMD Radeon Drivers Section' started by freibooter, Aug 2, 2013.

  1. CPC_RedDawn

    CPC_RedDawn Ancient Guru

    Messages:
    7,682
    Likes Received:
    200
    GPU:
    Zotac GTX1080Ti AMP
    I don't run CCC.exe, MOM.exe, or AMD Events Utility. I have them installed just in case I need to use CCC but otherwise I just run RadeonPro. Never had this bug.
     
  2. freibooter

    freibooter Active Member

    Messages:
    58
    Likes Received:
    0
    GPU:
    PowerColor R9 280X 3GB OC
    Have you adjusted your maximum clock speeds from their defaults in RadeonPro?
    Do you still have functional HW video acceleration?
     
  3. Offler2

    Offler2 Member Guru

    Messages:
    185
    Likes Received:
    0
    GPU:
    Gigabyte HD 7970OC, 3gb
    never happened to me. I always saw 1000Mhz (stock for my vbios and 3d frequency) if 3d and HW accelerated video/flash running together.

    If you see flickering if you already running Vram at lets say 1700Mhz, 3d engine is running well, but video kills the system then:

    a) your overclock is bad, but game is not taxing GPU/Vram enough to trigger any flicker of crash.

    b) other latent system issue.
    In this case however even SW issue.

    this belongs to "Idle clock bug" group. GPU/Vram is using different frequency as expected from user. Definitely not to HW Flash acceleration bug. That is completely different thing.
     
  4. Falkentyne

    Falkentyne Master Guru

    Messages:
    418
    Likes Received:
    2
    GPU:
    Sapphire HD 7970 Ghz Ed.
    Offler2:

    It has *****nothing***** to do with an unstable overclock.

    This issue has been reported to AMD ages ago.
    You can trigger this bug by setting the vram frequency ONE MHZ HIGHER than the bios defaults when running at 144 hz monitor (does not happen at 120 hz or lower).

    On ghz edition, if you set overdrive video ram at 1501 MHz(instead of 1500), it triggers the scrambled screen. All it takes is ONE MHZ over default for it to happen. I haven't tested it recently because I keep flash HW acceleration off or I don't use 144 hz (I use 100 or 120hz now)
     

  5. DrunkenDonkey

    DrunkenDonkey Master Guru

    Messages:
    204
    Likes Received:
    2
    GPU:
    2xPC 290 PCS+
    Yeah I caved in and turned off flash and any video acceleration. With new amd cards on the horizon I'm both tempted to upgrade and turned away from such issues :/
     
  6. freibooter

    freibooter Active Member

    Messages:
    58
    Likes Received:
    0
    GPU:
    PowerColor R9 280X 3GB OC
    :stewpid:

    Offler2, why do you keep posting in a thread that you are not reading?

    The bug has been described in great detail, the root cause has been identified and pretty much fully confirmed. This is a serious problem in just about any current AMD card and 100% reproducible in certain configurations and not present in any Nvidia or Intel GPU. It appears to be broken "by design".

    I learned a huge amount on this issue since the creation of this thread. Just about everyone else was very helpful in pinning this down and identifying the issue.
    But the only thing I learned from you, is that you are full of crap:

    You refuse to follow any instruction to recreate this bug. You refuse to add anything even remotely constructive to this discussion. And then you claim that it doesn't exist and that it must be a user error or cause by "bad software".

    You have done nothing in this thread other than to prove your utter ignorance and blind fandom over and over and over again.

    Please, just go away and stop derailing this thread any further.
     
  7. Pill Monster

    Pill Monster Banned

    Messages:
    25,216
    Likes Received:
    7
    GPU:
    7950 Vapor-X 1100/1500
    ^There is no bug.....and if you don't like flash player - don't use it. Pretty simple really.




    Personally I prefer VLC for web content:

    [​IMG]








    Btw this is MPC-HC playing a HW accelerated .MP4 file.

    [​IMG]
     
    Last edited: Aug 12, 2013
  8. Offler2

    Offler2 Member Guru

    Messages:
    185
    Likes Received:
    0
    GPU:
    Gigabyte HD 7970OC, 3gb
    To Freibooter:
    What you are describing is a mixture of two different bugs.

    1. The "clock bug". This also is not a single bug, but a group of bugs...
    2. Inconsistent behavior of video software when using HW acceleration.
    This affects mostly flash player, and some browsers. Mostly SW induced bugs, causing flickering.

    You dont need to have 144Hz display for that to reproduce it.
     
  9. Falkentyne

    Falkentyne Master Guru

    Messages:
    418
    Likes Received:
    2
    GPU:
    Sapphire HD 7970 Ghz Ed.
    To Pill Monster:
    You can't trigger the bug (at least with single monitor) if your IDLE clocks are 300/150. The bug won't trigger.

    You NEED idle clocks at 500/(overdrive ram speed) for this bug to happen.
    And although I've seen this on a Sammy 2233rz at 120hz, I was UNABLE to reproduce this AT ALL on 1) a CRT connected with adapter, or 2) a 144 hz LCD (VG248QE) monitor UNLESS the LCD was running at 144 hz. 120hz would not trigger it.

    At 144 hz, the idle clocks were 500/(overdrive ram speed) in this case, 1700. Although 1501 MHz would also trigger it (1500 was stock). When I played a flash video, the clocks would go to 501/1500, and this downclocking of the RAM would trigger the scrambled screen.

    Your example of course won't show any bug, because you are at 300/150. you won't get the bug like that.

    The poster above had the scrambled screen when he tried running youtube while playing a game. That would fit the bug triggering, because the RAM clocks would go from overdrive speeds to 1500 (or 1375). For some reason, this downclocking like this would scramble the screen.

    Going from 300/150 to UVD clocks would not.
    Guys, both of you, stop trolling. If you don't want to believe it's a bug, and you can't even REPRODUCE it correctly by having the IDLE clocks at your overdrive speeds (like at a 144 hz desktop), then stop posting in this thread. Because you have nothing to contribute except "hey there's no bug because I CANT MAKE IT HAPPEN TO ME." ...go grab a 144 hz monitor and THEN let's talk.
     
  10. freibooter

    freibooter Active Member

    Messages:
    58
    Likes Received:
    0
    GPU:
    PowerColor R9 280X 3GB OC
    No, I'm not, read this thread. This is one, single bug: This is a UVD clock problem.

    Unless you are willing to actually try to reproduce this behavior as instructed please go away and stop trolling!

    One more time just for you because I can't get rid of you:

    1. Go to AMD Overdrive and adjust at the very least your GPU memory clock and optionally also your GPU core clock! (In any direction, higher or lower, by any amount other than zero - all up to you. But if you don't do this, go away, leave this thread, you are not helping!)

    2. Launch Furmark and leave it running! (This is just to keep taxing the GPU and to keep the clock rates at maximum, you're welcome to use something else that maxes out your card contantly and reliably even while running in the background).

    3. Keep an eye on your actual clock rate, e.g. with GPU-Z.

    4. Launch anything that triggers AMD's idiotic, separate UVD clocks. Playing a video in Flash Player with HW acceleration in Firefox and Intener Explorer does trigger these clocks. If in doubt, do that.

    => On a standard, 60Hz single monitor setup you will now notice incorrect or at the very least unwanted maximum clock speeds during playback. On a multi-monitor or 144Hz or similar setup you will be able to produce flickering, corruption and even crashes if you repeat 4. often enough due to the unexpected and unnecessary memory clock change when starting and stopping playback.

    That's the bug!


    So, unless you have done that: go away and stop talking out of your ass.


    @Pill Monster

    The same pretty much goes for you, just imagine I had said it a lot nicer because, unlike Offler2, you have not tried to derail this thread with useless ramblings since its creation. This is not a bug in Adobe Flash Player, and it's not unique to it. Adobe Flash player just triggers it just like anything else that uses AMD's UVD for HW video playback.

    You also didn't follow instructions, your screenshots both show idle clocks, those are basically useless in the context of this bug.

    This is a problem with maximum clock speeds after they have been manually adjusted (e.g. in Overdrive) and while watching UVD accelerated video and simultaneously putting other load on the GPU (e.g. a benchmark or a game).

    Instead you show idle clocks that have not been adjusted (especially not the memory clock) and your answer appear to be to avoid triggering AMD's (not Adobe's) UVD clocks with is useless and borderline idiotic advice. And then you come to the utterly ignorant conclusion that "there is no bug".
     
    Last edited: Aug 12, 2013

  11. Pill Monster

    Pill Monster Banned

    Messages:
    25,216
    Likes Received:
    7
    GPU:
    7950 Vapor-X 1100/1500
    You're the troll and talking out your arrogant ass as usual. No surprises there. It is NOT a bug, it's a conflict. Two different things.

    I don't need to use Adobe flash because UVD clocks are triggered in MPC which is why I posted that screenshot, so people could actually see the clocks working. The first screenshot was simply to demonstrate another option aside from flash player.


    Look at my second screenshot again or would you like me to blow it up 200% for you so you can see the clock speed history.

    I'm also running two 100hz CRT's, that's 200Hz if you can't count.

    Oh and let's not forget the 500-1000 LCD's we use for tertiary clients at Fujitsu (where I work as an Engineer), no trouble there either.

    Maybe that's because our end users don't do silly things like watch flash videos and game at the same time.? Just a thought.

    If you're going to act all high and mighty and start insulting people at least make sure you have some idea what you're talking about.
     
    Last edited: Aug 12, 2013
  12. freibooter

    freibooter Active Member

    Messages:
    58
    Likes Received:
    0
    GPU:
    PowerColor R9 280X 3GB OC
    @Pill Monster

    I don't want to fight with you.

    "A conflict" should not causes corruption and crashes. This bug does. It has been confirmed by several people in this thread and a quick Google search reveals tons more. I would love for you to recreate it as well so you can see what we are talking about.

    I just want you to reproduce the bug as described and instructed and not show us what you think we are talking about, when you are indeed not showing us anything we are complaining about.

    Your GPU Memory Clock history shows zero changes throughout. Which means that:
    a) You did not follow instructions and did not adjust it manually
    b) You are at no point triggering AMD's UVD clocks
    c) You are somehow not affected by the bug that we are describing

    a) and b) would mean you did not follow the steps to recreate the bug. c) would be interesting. Sadly, from what you have written, I'm almost certain it's a) and b) not c)

    As a Fujitsu engineer I'm sure you can follow four simple steps to reproduce this issue and you probably even have a second monitor. So, please, do ... and the get back to us.

    I am quite certain by now, that at least now I'm confident in what I am talking about in this area. But I do seem to be unable to communicate it to those people (like Offler2 and yourself) who have read little more than the headline already think they know better without ever properly reading up on the real issue.
     
    Last edited: Aug 12, 2013
  13. DrunkenDonkey

    DrunkenDonkey Master Guru

    Messages:
    204
    Likes Received:
    2
    GPU:
    2xPC 290 PCS+
    Don't think freibooter is a troll, just annoyed person that can't use his videocard properly (definition of properly: watching videos while playing games at non-default memory clocks). And he is annoyed, because several people, totally ignoring the conditions that lead to the problem have ran flash and said "see? no problem".
    I can reproduce it with 2 7950s with different driver versions (from 13.1 to 13.8) and even different bios versions on different computers (work and home), so it should be no problem for anyone to see the fancy display corruption and crashes, if prerequirements are done.
    Sure, it is not a problem everyone faces, still it is one.
     
  14. freibooter

    freibooter Active Member

    Messages:
    58
    Likes Received:
    0
    GPU:
    PowerColor R9 280X 3GB OC
    It's just so painfully and incredibly clear and obvious, that this is a design/driver error on AMD's site.

    Yet people seem to read the word "Flash" and won't read any further. Instead they start screaming: "Ooooh, Adobe Flash, evil, it's clearly your fault for using it!"

    It's not exclusive to Flash but anything triggering AMD's stupid UVD clocks. And even if it were only Flash triggering this (it isn't), it would still be AMD's fault for allowing a simple userland process to change clock speeds like this and to ultimately crash and lock up the entire GPU (and only AMD's GPUs, not Nvidia's or Intel's).

    It's easy to reproduce if one follows the instructions to reproduce the error ... but that's apparently too much to ask, even from a fancy "Fujitsu engineer".
     
  15. Offler2

    Offler2 Member Guru

    Messages:
    185
    Likes Received:
    0
    GPU:
    Gigabyte HD 7970OC, 3gb
    I was talking about changing clock from 775Mhz (in my case) to 501Mhz in time when I had HD 5850.

    At that time when you used any video, including HW accelerated flash in browser GPU downclocked, decreased the performance of 3d regardless of running 3D engine.

    A well Known conflict between 3D engines and any SW which used UVD.

    On HD 7870 and HD 7970 the behavior changed. When I play 3D engine with HW accelerated video/flash GPU clocks down from 1050Mhz (My slight but safe overclock) to 1000Mhz (3D frequency in bios).

    This Happens ALWAYS regardless whan other steps you added there guys. Regarless of flash, 144Mhz bug or anything else.

    So once again. You are rediscovering the wheel.

    The higher GPU/Vram frequency you use, and higher voltage you use, the bigger Vdroop, and in such case you may cause VDROOP effect, and that may result if flickering, driver or system crash...

    So:

    Running OCCT/Furmark + Flash/any video with UVD
    Will always decrease clock and thus performance. Its normal behavior. I agree that running video and playing game together is more cool.

    Higher refresh rate, more displays mean higher requrements on Vram, thus higher possibility of flicker if you overclock your Vram, run furmark, and them play video.

    Once again = this effect is called Vdroop and its caused when you are downclocking, and later downvolting CPU, GPU or any kind of memory.

    Its true that your GPU and CPU and memory may be stable on some overclocked frequency when running 3D, but will cause flicker when you lower the frequency, lets say because you started video using UVD. However, it will NOT trigger when you use stock frequencies of your graphic card (or cards), regardless of number of displays and their refresh rates...

    And that is the point where any customer support will say, that you configuration is wrong.

    it works like this:

    More displays = Higher requirements or graphical memory.
    Higher display frequency = higher requirements on graphical memory

    Start 3d engine, tax your GPU and Vram = everything is fine.
    Play video = everything is still fine.

    Overclock your memory/GPU = agree that you have done something on your own risk.

    Start 3d engine, Tax your GPU and Vram.

    Now your memory is much more used as before, as frequency allows more tasks to be processed.

    Play video = GPU, VRAM downclocks. VDROOP occurs due overclocking (It does not when you keep your graphic card on stock). And voilá. you have grahphical glitches...

    The "downclocking" is surely unwanted behavior. I understand it.

    However, i have to agree with techsupport (trust me I already considered whether to report it or not), but ANY overclocking applied, while stock graphic frequencies are fine, automatically voided any such "bugreport" as invalid.

    All the method you described is situation, where you are taxing Vram a lot with more displays and higher refresh rates. That is all OK if you can trigger the flickering with furmark+video on STOCK frequencies and voltages.
     

  16. freibooter

    freibooter Active Member

    Messages:
    58
    Likes Received:
    0
    GPU:
    PowerColor R9 280X 3GB OC
    Offler2, at least you finally recreated it, even though you are still rambling like a lunatic. And, yes, of causes this happens always - it just happens to only cause severe problems/crashes with certain HW configurations like dual monitors.

    Facts are:

    Only AMD uses the crutch of separate, locked and non-adjustable maximum UVD clocks. Nvidia and Intel do HW video acceleration without problems and without this crutch.

    Settings in AMD's official over-/underclocking tool, AMD Overdrive, are ignored when these clocks are triggered.

    This causes switching between different memory clocks on multi-monitor and 144Hz displays etc. where otherwise the clock would never change.

    This eventually causes crashes on these systems, with or without additional load.

    So, if you want to adjust your memory clock in any direction (over/underclock) and use more than one monitor and occasionally watch HW accelerated video - then you can't do that with an AMD card.

    But for an Nvidia card this is child's play and I never even thought this could be a problem before switching to AMD.

    I understand that this is "by design". What I'm saying is that this design is wrong, unnecessary and causes severe problems exclusive to AMD hardware.

    I believe this could be fixed in a driver update, but most likely won't be - especially with people like you defending AMD's practice.


    You are saying that any manual clock changes are entirely unsupported. Then why is AMD even offering the option?

    Why allow manual clock changes when your drivers/hardware are designed in a way that actually makes it impossible to run them in combination with something as mundane as HW video?

    Why not just apply the manually adjusted maximums to both 3D and UVD clocks to solve the problem entirely?

    Your answer is: If you want to over- oder underclock - don't buy AMD! Buy a card that doesn't use such an idiotic design ...

    ... or, you know, maybe we could actually try to convince AMD to fix this, instead. Would that be an option?
     
    Last edited: Aug 12, 2013
  17. Pill Monster

    Pill Monster Banned

    Messages:
    25,216
    Likes Received:
    7
    GPU:
    7950 Vapor-X 1100/1500
    It's NOT A BUG. It is a FEATURE.

    UVD has been implemented as a power saving feature by AMD since at least Cayman, and they don't expect people to be running games and videos and whatever all at once.

    Seriously I get frustrated with the few guys like you who try and use hardware in a way it wasn't designed for and then cry when it doesn't work perfectly.
    If I listen to my radio and my TV at the same time would you expect me to blame Sony because I can't hear the movie properly? See the connection??? without issues

    Furthermore with the power of CPU's today HW accelerated flash video is not necessary. Any modern CPU will play a YouTube or VOD file from the internet EASILY. Hell I can even watch youtube videos with an Athlon XP 2800+.

    Btw HW rendering is only used in fullscreen, not windowed mode, and your comment regarding Nvidia GPU's is void since UVD is propriety to AMD.


    If you don't like the way AMD does things you have two choices: either deal with it or go buy an Nvidia card.

    The choice is yours.
     
    Last edited: Aug 12, 2013
  18. DrunkenDonkey

    DrunkenDonkey Master Guru

    Messages:
    204
    Likes Received:
    2
    GPU:
    2xPC 290 PCS+
    Your comparison is incorrect, as both TV and radio sound overlap each other and physics makes it so you can't hear perfect both, not some company decision. Better comparison will be if it is not possible to listen to music while browsing, because microsoft decides to turn off the sound card to save energy while the browser is active and there is no way to disable it.
    Sure, call it a feature, call it the one thing selling the cards, whatever, but provide an option to turn it off, thats all we want.

    On a side note, current UVD implementation is puzzling me:
    - when video is in idle clock, lower than what is required for a video to function, it is okay to temp bump it
    - when the video is at max, you are obviously doing something with it, why lowering the clocks, whats the gain? Going from 1000 to 950 and making problems, scrambled screens, freezes, crashes?

    If first part is left untouched, but only 2nd part, simply checks for memory frequency, if it is different than the stock - don't touch it, should be 2 lines of code (one for getting said freq, one for checking it). Then everyone will be happy - they retain their UVD, we get the ability to operate with 2 things at the same time.
     
  19. freibooter

    freibooter Active Member

    Messages:
    58
    Likes Received:
    0
    GPU:
    PowerColor R9 280X 3GB OC
    That expectation is wrong. Watching a review or walkthrough video on Youtube while gaming has become a perfectly normal thing to do.
    Do you think Valve is using PCs wrong for introducing Steam Overlay? Microsoft for introducing multitasking and "alt-tab"?

    The design doesn't match modern use cases. That doesn't make the usage wrong, that makes the design wrong. If your hardware and driver design doesn't support multitasking while the operating system does and the user expects it, then it's not the user's fault for using multitasking.

    Completely wrong analogy. If turning on both would cause them both to short circuit and trip the breaker for my living room ... yeah, then I would sure as hell blame the manufacturer.

    And using both is not an uncommon use case, either. Playing music while a news channel is on TV is a bloody common occurrence, both in private homes as well as public places. Even Burger King does it here ... and surprisingly, neither their TVs nor their sound systems crash while doing so.

    But video HW acceleration is not unique to AMD, far from it.
    AMD calls it UVD, Nvidia calls it "PureVideo HD", Intel calls it ... I don't even know if they have a fancy name for it. They all serve the same purpose but only AMD's design doesn't allow over/underclocking or multitasking.

    Yeah, that's not the only purpose of HW video and CPU decoding and especially upscaling looks bloody terrible in Flash -> http://forums.guru3d.com/showpost.php?p=4631410&postcount=17

    The alternative is to file a bug report and hope this gets addressed. Trust me, had I known this was an issue I would have never, ever bought an AMD card. This is something so bafflingly stupid, I never even considered the possibility that AMD would screw it up.

    tl;dr: You are a terrible, terrible engineer if you are one and a terrible representative of your company.
     
    Last edited: Aug 12, 2013
  20. DrunkenDonkey

    DrunkenDonkey Master Guru

    Messages:
    204
    Likes Received:
    2
    GPU:
    2xPC 290 PCS+
    I'm software developer. When a customer approaches me with an idea, much more complex than the simple 2 lines amd needs to make that useful, but explains me how it can use my products better I'm doing it. Not only that makes for satisfied customer (that in turns communicates with other customers, speaking of enterprise and there is a stream of new ones as a result), but it ups the usability and many more people enjoy the new feature.
    I'm also a guy many consult when about to buy hardware (I also have the connections to import cheaper gadgets from USA). Now imagine how I feel about AMD now, about my inability to watch any movie/flash while using my card (games/computing) not because it is unable to do so, but because someone took a bad call, do you think I will recommend AMD to the next person? I don't know what AMD's take is on this, but I consider watching movies in any situation a baseline any gfx card should support (granted it is healthy and not pushed beyound operational limits) and if this does not change, it leaves pretty bad taste about company's policy.
     

Share This Page