none
Wrong monitor enumeration RRS feed

  • General discussion

  • Hello,

    I'm here (still not sure if that's correct place) after long discuss with nVidia support and developers, at few stages. I'm experiencing a bug that is really annoying, and which causes many other issues - so for me it's something like critical bug.

    Thing that I'm writing about - is wrong monitor enumeration (that's what I've been told by nVidia developers). The symptom visible by end-user (me) is simple: my primary monitor is numbered as #2, whether secondary TV is set as #1. No, no, no - it's not so simple as switching cables in the back of my PC - if it were so, I wouldn't seek any help here.

    My config is typical multimonitor setup:

    * primary display: LCD monitor, FullHD resolution, "desktopped" (icons on there), DVI-D connection, always on,

    * secondary display: HD-Ready LCD TV, FullHD-based resolution (1824x1026 due to TV's overscan), DVI->HDMI connection, not always turned om (just for watching movies @ night),

    * Windows 7 Ultimate x64 (original, sent by Microsoft once I revealed it was a fake copy; anyways it's clean install now), updated,

    * GeForce 660Ti graphic card from KFA2 @ primary slot, newest nVidia drivers (335.23 WHQL),

    * MSI P67A-GD65 motherboard, newest drivers for Sandy Bridge CPU, and BIOS updated.

    Let me just quote conversation with nVidia - it will be quicker for me to introduce my bug further, and to show nVidia's point of view (I still believe this bug might be fixed by two sides: Microsoft - as drivers "handler", and nVidia - as drivers executor).

    Chat Transcript Rahul via Chat 01/08/2014 02:09 AM

    [01:44:09 AM] Hi, my name is Rahul. How may I help you?
    [01:44:17 AM] Martin : Hello,
    [01:44:29 AM] Martin : Can I report nVidia drivers bugs here ?
    [01:45:15 AM] Rahul: Sure,
    [01:45:24 AM] Rahul: What is the exact issue you are facing?
    [01:45:34 AM] Martin : I am facing two problems
    [01:46:16 AM] Martin : First arose few revisions back, and it regards to monitor numners in system
    [01:46:39 AM] Martin : Both issues regard to multimonitor setup (I mean monitor and TV, for example).
    [01:47:53 AM] Martin : I see it has not been ammended, and it was not so "crucial" until yesterday's new bug in 332.21 revision
    [01:48:47 AM] Martin : So bug number 1, please :) is monitor number change in system. My primary screen is numbered as "2" whereas it was (and it should've been) numbered as "1".
    [01:49:39 AM] Martin : It causes issues like icons position change, or newly open windows bad placing - in not correct screen.
    [01:51:19 AM] Martin : Please report developers to change those numbers back, as supposed: primary screen should be always #1. Never #2. It was good - as I said - few driver revisions back, but three or four klatest revisions suffer this bug.
    [cut]
    [01:58:59 AM] Martin : My setup is: FullHD monitor as primary and HD-Ready TV as second screen. Maybe it will help ypu I am using scalling on TV using 1826x1024 (based on FullHD).
    [01:59:51 AM] Martin : Yup, I choosed nVidia cards due to.. drivers which are better than AMD's in my opinion, but this bug is really irritating.
    [01:59:53 AM] Martin : :)
    [02:00:31 AM] Martin : Is everything clear with info that I gave you, or you need more clarification with any detail ?
    [02:02:33 AM] Rahul: This should be fine
    [02:02:53 AM] Rahul: Let me know the Operating System (Windows XP/ Vista/ 7 / 8 ) with the version (32 or 64 bit).
    [02:03:03 AM] Martin : It's Windows 7 x64
    [02:03:18 AM] Rahul: a) Click on the Start button. b) Select Search, c) In the Open field, type in "msinfo32" and then press the OK button. d) This will bring up the Microsoft System Information Utility. Click on File and then save to save this information to your hard drive. You may give it any name you choose. e) Once the file has been saved on your hard drive, attach it to this support request so that we could review your system configuration.
    [02:03:37 AM] Martin : But bug probably is present everywhere as you've got similar source code :)
    [02:05:00 AM] Rahul: Yes, in order to go ahead and escalate this case from my end I will also need you system information If you can help me with the system information I can go ahead and escalate this case to the concerned team
    [02:05:24 AM] Martin : Yes, it's in progress :)
    [02:05:37 AM] Rahul: Sure, Thank you
    [02:07:07 AM] Martin : It is being set now
    [02:07:10 AM] Martin : sent
    [02:07:50 AM] Rahul: Thank you so much for the information
    [02:08:30 AM] Martin : OK, I hope you'll pass this report to developers and they ammend those bugs as it's really annoying for multimonitor setups.
    [02:08:42 AM] Martin : Thanks for ypu help
    [02:08:49 AM] Martin : Bye bye ! Have a nice day.
    [02:09:23 AM] 'Martin ' disconnected ('Concluded by End-user').

    -----

    Response Rashid via Email 01/17/2014 11:46 AM

    Hello Martin,

    Kindly update me with the following information to check further on this:

    1. The system information shows the latest 332.21 driver, please let me know the driver version that is working fine.
    2. Kindly update me with the monitor model and the numbering.
    3. Please provide me screenshot of the following Windows in NVIDIA Control Panel.

    Set PhysX Configuration
    Set Up Multiple displays

    http://www.wikihow.com/Take-a-Screenshot-in-Microsoft-Windows

    The resolution generally depends on the EDID report built within the monitor and graphics card shouldn't generate any mixed resolution unless that is available in the monitor's EDID.

    Regards,
    NVIDIA Customer Care

    -----

    Customer Martin via CSS Email 01/18/2014 07:25 AM

    Hello,

    Thanks for reply.
    I can see you can't reproduce those bugs, so I'm going to give you as
    much details as possible incuding what I was asked for.

    I'm referencing to 2 bugs:
    #1 regards to monitor number change in multimonitor setup,
    [cut]

    Ad.1.
    * for bug #1 = 327.23 (and 3 previous works fine, too - I've checked).
    So 331.82 is buggy.
    [cut]

    Ad.2.
    Monitor models are:

    * primary - LCD monitor Dell ST2210 (FullHD) connected via DVI-DVI cable
    * secondary - LCD TV Sharp LC-26SA1E (HDReady) connected via DVI-HDMI cable

    Primary desktop is 1920x1080@60Hz with no size extension.
    Secondary desktop is 1824x1026 based od 1920x1080@60Hz with desktop size
    scalling (screenshot "nv_Multiple2" and "nv_Multiple3") due to HDReady
    overscan

    Ad.3.
    Screenshots:
    * PhysX - file "nv_PhysX"
    * Multiple monitor setup - file "nv_Multiple1"

    Hopefully I gave you clear information, just in case = ask for more
    details :)
    Cheers,
    Martin

    -----

    Note from me - screenshots mentioned above shows that my primary monitor is numbered as "2" and it is located on RIGHT-hand side. It's because my TV is located (secondary display) on LEFT-hand side, and as such, is easier for me to drag windows to left-hand-side display. Maybe it's a clue of whole issue.

    -----

    Response Rashid via Email 03/13/2014 07:02 AM

    Hello,

    Your case is being escalated to our Level 2 Technical Support group for further attention. The Level 2 agents will review the case notes to troubleshoot the issue and find a solution or workaround. As this process may take some time and require a good deal of testing and research, we ask that you be patient. A Level 2 tech will contact you as soon they can to assist or point you in the right direction.

    Best Regards,

    NVIDIA Customer Care

    -----

    Response Ray via Email 03/17/2014 11:23 AM

    Hi Martin,

    The case was escalated to me. I've review the notes and if I understand correctly the first issue is actually not something we can control. Monitoring numbering, or the monitor identity is completely own and assigned by the OS. As far as I know the OS assigned the monitor number based on which display is enumerated first. This can be different depending on the order the OS enumerate the displays if both displays are connected on startup. The OS rely on the boot-table from the GPU BIOS, to designate which of the outputs (HDMI, DVI, DP) to initialize first when Windows boot. This can vary depending on the GPU card, some have the boot order as HDMI > DVI > DP, while others may have boot order as DVI > HDMI > DP. The monitor numbering assigned by the OS has nothing to do with primary vs. secondary in terms of display ordering, these two are completely different from each other. So you can have a monitor labeled as # 2 and that can still be the primary display. Again this is entirely control by the OS and shouldn't cause any multi-display problems regardless of the numbering.

    If you prefer the numbering to match with #1 to primary and #2 to secondary, then you must boot into Windows with one monitor at a time. Another word boot into Windows with only the primary monitor first, the OS will automatically assigned this monitor as #1 since it's the only display detected. Once in Windows you can then hot-plug the secondary display, the OS will automatically assign this as the next enumerated, in this case monitor #2. This should help avoid any confusion that may be associated with the numbering. Once that is set the OS should preserved this numbering, at least until you reinstall the driver. If you update the driver it should preserved the same numbering, but if you do a clean install then you'll have to repeat the same steps.

    As for the second issue resolution gets mixed, I'm not certain if that has anything to do with the above, but again if you follow the steps above and sync #1 with primary and #2 with secondary than that shouldn't be a problem. If problem still persist after above then this may be a separate issue not related to the numbering. Try following the steps above and see if that helps.

    Another possible alternative is to swap the display output since you're using DVI connections for both displays. If you have both displays attached while booting into Windows, which ever display output is assigned #1 is likely the first enumerated output. With that information just make sure the display you want as primary is connected to that DVI output. This will help insure the display connected to that DVI output will be #1.

    Best regards,
    Ray 

    -----

    Customer Martin via CSS Web 03/19/2014 04:56 AM

    Hello,

    Thank you for reply.
    Unfortunately, it's everything I can thank for, because you are probably a person I was looking for: full of incompetency, with amateurish attitude, big “holes” in general knowledge.

    Let me introduce myself: I'm a computer troubleshooter, with hundreds (logged) or thousands (unlogged) resolved PC problems. On daily basis I help people to solve their problems with both: hardware and software. I am also a programmer with knowledge of few languages: starting from assembler, ending with PHP and web programming. I also sell customs PC sets basing on my configuration recommendations, and do some repairs, or electronic projects.
    As you can see – I have extensive (over 27 years) knowledge and if I seek any help – it means problem is really significant; or at least – it exists.

    During all my 3 (THREE) bug reports, I couldn't resolve my problem due to invisible wall – I could convince all nVidia on-line-chat operators (Mr Vaseem for example) there we have some serious BUG in drivers, but all reports crashed against that wall located “one person higher” in your job/competency organisation – probably it was/is you.

    Now, once I reached some “level 2 technical support”, some person (you) tells me rubbish (I repeat: RUBBISH) that I couldn't even imagine I would PERSONALLY advice to ANYONE SEEKING HELP.
    In other words – level of your advice is ridiculous – and looking at your position, it's UNPROFESSIONAL and seems to be some mistake. Like “hey, who hired you there, your uncle ? If I were your boss, I'd fire you and would do your job 100x better and for 33% of your wages, with my knowledge”. You even didn't make an effort to read all my reports, and watch screenshots I have attached.

    So – when you understand my frustration (you're welcome, finally it's your fault) – we can talk about the core: THE BUG. So – please – change you attitude (I will prove you were wrong below), or if you are not able to – escalate my reports to “level 3 technical support” or pass my case to your manager, please. All I need to – is to contact a person who understands what I am writing and is able to give appropriate feedback. Moreover, if there will be a need – I WILL GO HIGHER AND HIGHER WITH MY ISSUE, UNTIL I REACH MR JEN-HSUN HUANG, NVIDIA CEO, informing everyone on the way and by the way about eventual lower level incompetency. Somebody must know where and what for his money are being spent by a mistake.

    You wrote (it's really insult to the intelligence to quote you, so just a part of it):
    “With that information just make sure the display you want as primary is connected to that DVI output. This will help insure the display connected to that DVI output will be #1.”

    No, this will NOT help (except drivers 327.23 and previous revisions*). Regardless the DVI port my primary screen is connected to, it will be ALWAYS #2 (which you should've known, if you *REALLY read my reports*).
    Effect of connection to port DVI-2 = well, graphic card's BIOS shown on TV (it's my secondary display), motherboard POST SCREEN (and BIOS entrance, if any) shown on TV, any operating system booting theme shown on TV during drivers init … tadaa... and desktop shown on main screen (whch is enumerated as #2 by the way). I don't say about DVI-1, because it's default port which works as described many times... Completely useless advice....
    Except drivers 327.23 and previous revisions.


    * now – I explain why drivers 327.23 matters. It's why: I'm sitting on 335.23, main screen is enumerated as #2. Secondary screen is theoretically enabled (in drivers), but physically TV is turned off; drivers point this TV as display #1. As shown on screenshot:
    http://i740.photobucket.com/albums/xx42/biuro74/331_xx_zps073e6cf7.png
    I underline, all drivers setrtings are default. Installation is “clean” one.
    And then, I install 327.23 drivers. Clean install as well. Of course, not everything is being installed in older revisions – PhysX and GeForce Experience stays untouched in newer revisions, but core drivers and HDMI audio drivers are “downgraded” to 327.23.
    Again, it's how screen looks like – with default settings (nothing changed):
    http://i740.photobucket.com/albums/xx42/biuro74/327_xx_zps739b06e3.png
    You see – it looks like IT SUPPOSED TO LOOK and what's more important – IT LOOKS LIKE IT LOOKED BEFORE IN PREVIOUS DRIVERS REVISIONS !
    So I try to install newest 335.23 one more time – with clean install option again. We've got main screen as #2 again.
    So it's a PURE PROOF that in 331.xx+ drivers code changed somehow – the way that enumerates displays WRONG WAY. And one more important information – it's nothing to do with GeForce Experience (unlike another bug that I reported regarding to incorrect resolutions that have been mixed up like taking data from two displays = it's not present now, but I was confirmed it was connected with GFE).
    Summing this annotation up – all drivers 331.82 WHQL, 332.21 WHQL, 334.89 WHQL and 335.23 WHQL have THE SAME bug which enumerates displays WRONG WAY. Unlike 327.23 WHQL, 320.49 WHQL, 320.18 WHQL and 314.22 WHQL which WORK FINE. One more important thing – I've checked ALL of those drivers with clean install option (not only, but it's probably required).


    Let's make things clear: I've reported few bugs. I numbered them somehow, so for future reference I organize them one more time:
    Bug #1 – wrong displays enumerations (see Incident 140108-000023 for details, I've attached few screenshots which you can update with above mentioned Photobucket links)
    [cut]


    AD. BUG#1 – WHAT IT MIGHT CAUSE (another strange behaviours) – so it's why I treat this bug as MAJOR:
    1. Enumerating displays = no matter which video out socket is populated (in single or multimonitor configs), main screen should've appeared always as primary and numbered as #1. Otherwise: wrong enumerating gives blank screen (once drivers initialized) due to enumerated screen is "out of display range" (not main screen). Try to change cable with primary display in SINGLE setup (not DVI-DVI – as usual, but DVI-HDMI) and it will give BLANK SCREEN until you connect anything to second socket which simple is sick as second display needs to be connected to give signal to first – already connected socket (!) Typical scenario of such actions is: disconnect main monitor, and connect PC to another display, like TV in salon. Even using Windows shortcut WINDOWS key + P it's impossible to get screen on primary display once you pick “Computer only”.
    2. Wrong icon positions. Newly added icons, or just after 2nd screen disabling, some icons move sometimes out of range of MAIN display. Cause: they show up at screen #1 which is badly enumerated.
    3. Wrongly opened windows. Sometimes window just open NOT at main screen. Typical programs like TheGimp which opens multiple sub-windows (toolbars, picture windows etc) are most irritating as ANY windows needs to be moved to correct window (all the time they are being open). Cause: they open at screen #1 which is badly enumerated.
    SO you see how MAJOR bug #1 messes up with other things that NORMALLY should work.

    At the end, there's proof it's NOT ONLY my delusion:
    http://forums.overclockers.co.uk/showthread.php?t=18553531&highlight=nVidia+drivers+username_biuro74
    (thread author: mushtafa, my answer: biuro74). Read it carefully, as it gives ideally my point. Easy to reproduce.

    I don't mind you contact Microsoft. I don't mind ANYTHING YOU DO, until it fixes this bug. Of course, you might blame Microsoft for bad enumerating displays, but in such case I ask you one simple question: what “wonderful” is in 327.23 drivers that allows operating system to enumerate displays CORRECT ? Huh ?

    My config:
    Motherboard: MSI P67A-GD65(B3), latest BIOS 1.J0.
    Graphic card: KFA2 (Galaxy) GeForce 660Ti EX OC, BIOS revision: 80.04.4B.00.0F (P2004-0001), device ID: 10DE-1183.
    Windows 7 x64 original, latest patches, no viruses, malware and generally mess in OS.
    Linux drivers 331.49 for 32-bit systems don't have such errors (but .run file for 64-bit systems is probably compiled with bad CR/LF notation and all messages are overlayed – tested on Mandriva 2010.3 x64).

    -----

    Response Ray via Email 03/19/2014 08:48 AM

    Hi Martin,

    I am sorry if I've offended you, that was not my intention. We discussed display identity numbering with development many times and we're always told that this is control by the OS and not by our drivers. I'm referring to display identity numbering in general so lets put aside the potential bug for a moment. There are plenty posting of this from various forums with users trying to figure out a way to change the display identity numbering. Physically swapping the display connection was the only solution that had worked in the past. Here is just one of many postings: http://social.technet.microsoft.com/Forums/forefront/en-US/30b16266-cbc6-4759-8c30-8d10b46fb7c6/changing-monitor-identities?forum=w7itproui.

    Having said that, I apologize for jumping to conclusion and generalizing your report as yet another OS display identity report. Clearly something in the new graphics driver has changed the display identity behavior. I missed that part and that was my bad, I am sorry. Your test results between the 327.23 and the 335.23 drivers does suggest a change in our drivers that is impacting the display identity numbering. This is potentially a bug or a change in our driver behavior that will need to be investigated by development. Let me submit a bug to our quality lab so they can try to replicate the behavior. Once the lab is able to replicate this behavior then we can escalate the bug to development to get their feedback. This is Bug #1. Bug #3 and Bug #4 have already been submitted to our development team as enhancement requests. Is there anything I should follow-up with Bug #2?

    Again let me sincerely apologize for jumping the gun Martin. Let me get a bug submitted so we can look into Bug #1.

    Best regards,
    Ray

    -----

    Response Ray via Email 03/19/2014 10:28 AM

    Hi Martin,

    The lab was able to replicate this behavior and has escalated the bug to development. And if the lab follow my steps below to boot with just the monitor connected first, and then hot-plug the TV once in Windows, that does workaround the problem and TV does get enumerated as display #2. But this only works for that Windows session, on restart the TV is re-enumerated as display #1 again. This behavior is definitely not consistent. With the 327.23 drivers the TV is always enumerated as display #2, even if we hot-plug. Will have to wait for development input on this new behavior and see if this is expected or a bug. Will update as soon as I hear back from development.


    Best regards,
    Ray

    -----

    Customer Martin via CSS Web 03/21/2014 02:59 AM

    Thank you for response.

    Sorry for my frustration show. Now, it's all as expected - with engagement which makes me happy as it gives hope to fix a problem, or at least having a knowledge it might be treated as a bug (the worst scenario is when devs would say it'd be "expected" just because nobody wanted to spend more time to find out the cause). It shouldn't be too hard to find out which change or group of changes is responsible for wrong screen enumerating - developers surely have "changes.list" that corresponds to 327.23 successor (331.81 WHQL, or some other betas), so I wish them plenty of power, Coke, RedBull, carrot juice or whatever makes their brains clearer :)

    Everything I am aware and what I believe in - is this must be treated as a bug due to.. the nearest future. There will be games using two, or three monitors (not "split screen", but "split displays") because hardware nowadays is strong enough to drive such solutions. I can only imagine that mess and frustration of many players that can't set up such multi-display game properly. This should drive nVidia to solve issue right now, at least in my personal opinion, because later - it will be much harder.

    Thanks for your help, so I'm waiting for what developers would say as well.
    Cheers,
    Martin

    -----

    Response Ray via Email 03/26/2014 09:27 AM

    Hi Martin,

    Development review and updated the bug I submitted. As suspected they reaffirmed that the display enumerations is entirely OS decision where display driver has nothing to do with the display number. I question the behavior differences between the two drivers and they indicated that different driver version may have a different interaction with the OS but that ultimately the OS makes the decision on how the display number is reported. They further add that this OS numbering only serves to identify the display and should not impact multi-display functionality in terms of primary vs. slave or arrangement since all that can be configured independent of the OS numbering. The one potential problem they did point out is that some applications may falsely use this OS numbering to identify the primary display and that could result in some pop-up dialog showing up in the #1 display regardless which is designated as main display. The only way to resolve this is for Microsoft to provide an option in the OS to allow users to change the numbering. They mentioned that there have been many requests from users for such a feature but that has not been made available in Vista, Win7, or Win8. According to development we couldn't change this OS numbering even if we wanted to. I realize this is not the answer you're expecting but my hands are tied if the development manager tell me this is not a bug. Let me know if you have any further questions.

    Best regards,
    Ray

    -----

    So - as you can see, I couldn't make nVidia developers to fix that bug in their own code, despite previous releases of those drivers (WHQL-ed by you, Microsoft... well... it doesn't speak well of you) worked FINE. Personally for me it looks like someone in nVidia changed way (maybe an order) to detecting displays - and now they cannot reverse the changes.

    But.. but... it's not so bad, finally. Why ? Because I can see similar issues, but with Radeon users. So maybe it would be better idea to set up properly a way to enumerating displays, no matter which graphic driver is being used (in short: one patch seems to be smarter, quicker and more effective than forcing drivers makers to do it it their own way - each of them separately).

    So is there a chance to fix this bug by Microsoft developers team ?

    Wednesday, April 23, 2014 10:32 AM

All replies

  • Sorry, but this isn't a support forum for issues you have with drivers or the is, it is for driver developers writing their software. I appreciate how difficult this is for you, but this is not the spot to get a fix

    d -- This posting is provided "AS IS" with no warranties, and confers no rights.

    Wednesday, April 23, 2014 11:44 PM
  • Sorry, forgot to mention two thigs in a rush :/

    I was diverted here by Microsoft Chat support operator, so I wrote "quick" post without double check if this is a correct place.

    Apparently it seems to be incorrect one, and Microsoft Chat support operator doesn't recognize "Microsoft developer" and "developer". All I wanted was to contact Microsoft Windows developer, any of them (or at least anybody who knows where to report Windows bugs), who might discuss the enumeration subject. I know you people are very helpful (otherwise you wouldn't read posts here) - and I pray any Windows developer would be as much helpful as you are.

    Thanks for any help - if you know correct person/place/link/whatever, I'd be grateful.

    Martin

    Saturday, April 26, 2014 6:31 PM
  • Sorry, I don't want to be unkind, but this doesn't explain anything. Looks like Windows mechanism called "enumeration" is some kind of BIG BLACK HOLE - everybody knows it exists, but nobody knows how it works.

    Ironically, I have to convince myself and my logic that Microsoft is able to fix that mechanism, which enumerates monitor as #1 once, and as #2 another day with new drivers installed. Irony is because of explanation given by nVidia developers claiming that is all about operation system, not drivers. Huh, I'm old enough to know ALL ISSUES are to be fixed from TWO sides; in this particular case - from OS and drivers sides. But nobody wants to admit his side is wrong - and this is a clue, probably.

    I will probably push nVidia developers to contact their counterparts in Microsoft - to fix this issue definately. It's just ridiculous that device can't be configured as supposed and nobody wants to do anything with that fact. It's not Windows 95 !

    Sunday, May 18, 2014 1:17 PM