Topic Title: My display doesn't 'wake' during the morning so to speak
Topic Summary:
Created On: 05/31/2014 02:44 AM
Status: Post and Reply
Linear : Threading : Single : Branch
<< 1 2 3 4 5 6 Previous Next Last unread
Search Topic Search Topic
Topic Tools Topic Tools
View similar topics View similar topics
View topic in raw text format. Print this topic.
 06/02/2014 05:48 PM
User is offline View Users Profile Print this message

Author Icon
Emopunk
Peon

Posts: 38
Joined: 04/21/2014

This bug is really difficult to track. Today I got it even with 13.12, which I considered  most trustful driver release in the last months. At the moment, I switched off auto power-off on monitor and set up black screen screen saver in windows. Until I hear that the problem is solved, at least.



-------------------------
We need to concentrate on more than meets the eye
 06/03/2014 09:34 PM
User is offline View Users Profile Print this message

Author Icon
Solid State Brain
Peon

Posts: 32
Joined: 05/31/2014

Just wanted to report that this problem occurred again (though only once so far), after I performed a Windows 8.1 "System Refresh" and using the latest AMD Catalyst drivers from Windows update (version 13.251, which are a bit outdated) and built-in / WU drivers for all other system components.

 06/04/2014 07:35 AM
User is offline View Users Profile Print this message

Author Icon
Solid State Brain
Peon

Posts: 32
Joined: 05/31/2014

Do people having this problem have a second monitor physically connected to the GPU, whether enabled or not?

After attempting several things without success (the list is growing quite a bit), all seems to point either to that, or the fact that my unit under Linux when using the open source display driver is reported to have an "invalid EDID" (as far as I remember, it was a while ago). Or, that connecting more than one monitor to the GPU is triggering this problem.

I will first test by physically connecting only the "good", primary monitor first, though (a Dell U2412M). When the second monitor was connected together with it, the problem eventually occurred whether I used a DVI or DisplayPort connection on the main one.

 

 06/04/2014 10:40 AM
User is offline View Users Profile Print this message

Author Icon
Solid State Brain
Peon

Posts: 32
Joined: 05/31/2014

It looks like I might have talked too soon. Physically connecting only the main monitor on the GPU doesn't solve the problem either.

However, I think I found out something interesting. It looks like upon monitor sleep, the GPU will enter the ULPS mode and switch off its fan completely when the screen doesn't get updated with new graphics. Even having something updating (for example a blinking cursor or notification) will either prevent the GPU from entering such mode or waking it up from it, if it entered it (although the display will remain turned off).

This can be made very noticeable by toggling on Graphics OverDrive from the Catalyst settings panel and setting up a manual, fixed fan speed, like for example 70%. When the GPU enters the ultra low power state (ULPS) it will switch the fan off completely. Opening the computer chassis will reveal that the GPU fan is actually not moving.

So, this wake up problem appears to happen primarily (if not solely, but I can't confirm this yet) when the GPU wakes up from this low power state and turns the fan on. Occasionally, after turning the fan on, it also fails to turn the display on and won't perform any further attempt doing it. OS control with keyboard and mouse is functional, but the video output can only be restored after forcing the GPU to attempt display detection again (for example by either disconnecting/connecting a display or switching the multimonitor configuration).

Interestingly, after increasing fan speed to a noisy setting, I noticed that the regedit fix for turning ULPS off apparently doesn't perform anything useful. The fan still keeps getting turned off completely (either with older 13.251 drivers or the newest beta drivers), indicating that the GPU is also being switched off completely. The only way the avoid this is making the GPU avoid entering such state. This can be achieved by having something constantly (not intermittently) updating on screen, OR by enabling a second display on the integrated Intel GPU if you have one. It looks like if any monitor is connected and enabled on the IGP, the discrete video card won't enter the ULPS state ie its fan won't be switched off. This occurs even if the second display isn't getting video updates. It looks like ULPS works only if all video outputs come from the discrete GPU.

This isn't an optimal solution (it's not even a fix to be fair, rather a workaround), but the power consumption of the discrete AMD GPU even without entering the ultra low power state, appears to be relatively low when all video outputs are in sleep mode (turned off via software, not just with the monitor's power button), according to the control panel of my uninterruptible power supply.

This will need further testing to make sure that it does fix (partially) this issue. Even if it doesn't, it's still "interesting" that setting to 0 all "EnableULPS" registry keys doesn't seem to do what it's supposed to.

 06/05/2014 05:54 AM
User is offline View Users Profile Print this message

Author Icon
TheIcon
Peon

Posts: 31
Joined: 05/20/2014

Appareciate you are trying to solve the problem. A few days back I resintalled my Windows 8.1, updated all the drivers, windows updates, etc. But still have this issue. The monintor light keeps blinking, doesn't seem that it takes any input at that time fro mmy keyboard or mouse.

 

Originally posted by: Solid State Brain It looks like I might have talked too soon. Physically connecting only the main monitor on the GPU doesn't solve the problem either. However, I think I found out something interesting. It looks like upon monitor sleep, the GPU will enter the ULPS mode and switch off its fan completely when the screen doesn't get updated with new graphics. Even having something updating (for example a blinking cursor or notification) will either prevent the GPU from entering such mode or waking it up from it, if it entered it (although the display will remain turned off). This can be made very noticeable by toggling on Graphics OverDrive from the Catalyst settings panel and setting up a manual, fixed fan speed, like for example 70%. When the GPU enters the ultra low power state (ULPS) it will switch the fan off completely. Opening the computer chassis will reveal that the GPU fan is actually not moving. So, this wake up problem appears to happen primarily (if not solely, but I can't confirm this yet) when the GPU wakes up from this low power state and turns the fan on. Occasionally, after turning the fan on, it also fails to turn the display on and won't perform any further attempt doing it. OS control with keyboard and mouse is functional, but the video output can only be restored after forcing the GPU to attempt display detection again (for example by either disconnecting/connecting a display or switching the multimonitor configuration). Interestingly, after increasing fan speed to a noisy setting, I noticed that the regedit fix for turning ULPS off apparently doesn't perform anything useful. The fan still keeps getting turned off completely (either with older 13.251 drivers or the newest beta drivers), indicating that the GPU is also being switched off completely. The only way the avoid this is making the GPU avoid entering such state. This can be achieved by having something constantly (not intermittently) updating on screen, OR by enabling a second display on the integrated Intel GPU if you have one. It looks like if any monitor is connected and enabled on the IGP, the discrete video card won't enter the ULPS state ie its fan won't be switched off. This occurs even if the second display isn't getting video updates. It looks like ULPS works only if all video outputs come from the discrete GPU. This isn't an optimal solution (it's not even a fix to be fair, rather a workaround), but the power consumption of the discrete AMD GPU even without entering the ultra low power state, appears to be relatively low when all video outputs are in sleep mode (turned off via software, not just with the monitor's power button), according to the control panel of my uninterruptible power supply. This will need further testing to make sure that it does fix (partially) this issue. Even if it doesn't, it's still "interesting" that setting to 0 all "EnableULPS" registry keys doesn't seem to do what it's supposed to.

 

 06/05/2014 06:37 AM
User is offline View Users Profile Print this message

Author Icon
Solid State Brain
Peon

Posts: 32
Joined: 05/31/2014

TheIcon: Can you confirm that this happens when the GPU switches on and turns on its fan? You can check out this better by setting a significantly higher manual fan speed in the Catalyst drivers by enabling Overdrive (overclocking) settings. Just increase the fan speed, leave all other settings at default.

 06/05/2014 07:03 AM
User is offline View Users Profile Print this message

Author Icon
TheIcon
Peon

Posts: 31
Joined: 05/20/2014

Ok I'll let you know 

 06/05/2014 08:31 AM
User is offline View Users Profile Print this message

Author Icon
Solid State Brain
Peon

Posts: 32
Joined: 05/31/2014

Originally posted by: TheIcon Ok I'll let you know

You can accelerate the process by either setting display sleep to 1 minute, or by downloading NirCmd here:

http://www.nirsoft.net/utils/nircmd.html
Actual download link: http://www.nirsoft.net/utils/nircmd.zip

Then, if you unpack it in a directory, through the command prompt you can type "nircmd monitor off" to make the system immediately trigger its display sleep state, without delays.

If the screen isn't receiving any video update (which includes blinking cursors, so you might want to create a shortcut icon for that instead of opening the command prompt), the GPU will switch off, together with its fan. If you click "Show Desktop" before doing this, minimizing all applications (shortcut: Win+D), this will have a greater chance of happening.

Keep in mind though that there doesn't seem to be an immediate correlation of the monitor sleep bug occurring with this, although it's a prerequisite.

 

 06/05/2014 10:56 AM
User is offline View Users Profile Print this message

Author Icon
Eydee
Ninja Zombie Killer

Posts: 4905
Joined: 12/27/2008

I wonder what happens if

- you disable EDID in CCC (by remote access for example)

- you install a custom monitor driver that overrides the EDID

- you use a broken d-sub cable that cannot carry ddc info

 



-------------------------

CPU: AMD Phenom II X4 810 @ 3120MHz | RAM: Kingmax 2x2GB DDR2 800 @ 833MHz| MoBo: MSI K9A2 CF v1.0 (BIOS: 1.D)| GPU: Asus HD 6850 1024MB (DirectCu) @ 835/1135MHz | Display: L24FHD | PSU: PC Power & Cooling Silencer 750 Quad | OS: MS Windows 3.11 Pro x64

 06/05/2014 11:12 AM
User is offline View Users Profile Print this message

Author Icon
Solid State Brain
Peon

Posts: 32
Joined: 05/31/2014

Eydee: DDC/CI can be disabled on several monitors, including both of those I'm using. However, the secondary monitor I thought was the reason for the issues I have (in the end it turns out it wasn't), had it disabled. Turning it on didn't fix anything.

I'm beginning to have a different hypothesis on this problem, however. Sometimes instead of not getting my displays waking out from sleep, I get a graphics corruption on certain applications, or application graphics UI gets first rendered slowly (I get black colored windows for longer than I think it would normally be expected).
When the GPU gets switched off, GPU video ram occupation also drops to zero, therefore its content must be getting dumped somewhere (either the system RAM or the system paging file).

What I'm getting at is: what if there is some problem or chipset incompatibility with Catalyst video drivers preventing the process of restoring video ram content from being performed correctly? If it's the case, that could explain why these problems seem to occur only when the GPU gets switched off.

I'm now trying a couple related fix attempts. Notably:
- I have set a minimum paging file size to the amount of RAM in my system (8GB), in case it's important that it can be readily and easily mapped to the physical memory content (there shouldn't be any problem here, but you never know).
- I disabled the XMP-1600 RAM memory speed profile (which is non-standard, although both the memory sticks and the motherboard are supposed to support it) and now running with the default (and within JEDEC specifications) 1333 MHz memory speed profile.

I will let you know if there will be improvements. To tell the truth, so far (about 6 hours of testing) I haven't encountered any yet, but I guess it's still too early to tell.

 06/05/2014 11:32 AM
User is offline View Users Profile Print this message

Author Icon
TheIcon
Peon

Posts: 31
Joined: 05/20/2014

i set the max fan speed as you asked, went to power, additional settings and set turn off display to 1 minute . After it turned off, i could still hear my gpu fan roaring, after 2 minutes i clicked a random keyboard button and it woke up.

 06/05/2014 11:42 AM
User is offline View Users Profile Print this message

Author Icon
Solid State Brain
Peon

Posts: 32
Joined: 05/31/2014

TheIcon: remember that for the ultra low power state (ULPS) to activate, which basically switches the GPU off almost completely, including the fan, you have to preferably (it's not exactly needed, but it helps big time and prevents the GPU from temporarily turning on when something happens on screen within those applications) minimize all applications in order to avoid any screen update. This means that only the desktop should be showing.

Also keep in mind that certain applications can cause the ULPS to not turn on, even if they are apparently not updating the screen.

The display wake up problems I had so far only happened (as far as I remember... it would be a bad surprise if this wasn't the case) when the GPU fan turned off completely, indicating that the ULPS has engaged.

 

Can you remember what's in the user interface and which programs do you keep opened when you leave the PC in the evening before realizing that the display(s) won't wake up in the next morning?

 

Eydee: by the way, it appears that setting minimum pagefile size equal to physical RAM size or disabling XMP-1600 memory profile from BIOS settings does not solve my display wake up problem. So, my hypothesis got debunked soon after I wrote about it.

I have now attempted disabling DDC/CI on my primary display, although I doubt it will solve any problem.



Edited: 06/05/2014 at 12:02 PM by Solid State Brain
 06/05/2014 12:48 PM
User is offline View Users Profile Print this message

Author Icon
Eydee
Ninja Zombie Killer

Posts: 4905
Joined: 12/27/2008

One more thing I can add, I have a non-GCN card which does not have such extreme low power states and have no issues. However, I still wonder why doesn't everyone have it, who have GCN cards.



-------------------------

CPU: AMD Phenom II X4 810 @ 3120MHz | RAM: Kingmax 2x2GB DDR2 800 @ 833MHz| MoBo: MSI K9A2 CF v1.0 (BIOS: 1.D)| GPU: Asus HD 6850 1024MB (DirectCu) @ 835/1135MHz | Display: L24FHD | PSU: PC Power & Cooling Silencer 750 Quad | OS: MS Windows 3.11 Pro x64

 06/05/2014 01:44 PM
User is offline View Users Profile Print this message

Author Icon
Solid State Brain
Peon

Posts: 32
Joined: 05/31/2014

It could be an incompatibility issue with GCN AMD videocards, Windows 8/8.1 and certain chipsets + other prerequisites (Primary display model?) involving the ULPS. I have an Intel H77 chipset-based motherboard. Tried on a different similar one, with the same results. I tried changing pretty much everything else in my system except the GPU and the OS, without being able to solve the problem so far.

It would be nice to have an option or suggestion to disable the GPU's ULPS more or less permanently and see if it's really the cause (I suspect it is). Registry hacks (setting to 0 "EnableULPS" entries) don't seem to work as I thought but I only realized this recently after manually setting the fans to a noisy threshold.

 06/10/2014 07:52 PM
User is offline View Users Profile Print this message

Author Icon
Solid State Brain
Peon

Posts: 32
Joined: 05/31/2014

@TheIcon: how do you actually wake up your pc/display in the morning? Is there any specific keyboard key that you're pressing every time which incidentally (and perhaps indirectly) is causing this problem?

Are you by any chance using modifier keyboard keys? (shift, control, alt) I know they can be handy in such situations because they usually avoid causing any unwanted change to the interface when waking up the display(s).

By the way, here's a list of things I tried which didn't work:


DisplayPort / DVI on primary monitor -- not relevant
Only main display physically connected to the AMD GPU -- not relevant
Only one display connected at all -- not relevant
Monitor input signal auto detect -- not relevant
AMD Catalyst driver version -- not relevant (happens with 13.251 too)
Intel Rapid Storage drivers installed -- not relevant
Intel IGP enabled/disabled -- not relevant
Intel Graphics drivers installed -- not relevant
Windows System Refresh performed -- not relevant
APC UPS monitoring software installed -- not relevant
BIOS settings on PCIe link speed or processor C1E state -- not relevant
BIOS VT-d disabled/enabled -- not relevant
Disabling XMP-1600 memory profile from BIOS settings -- not relevant
GPU triggers to/from ULPS (fan off) several times when display is sleeping -- not relevant
Catalyst Graphics Overdrive enabled -- not relevant
Wacom pen tablet driver installed -- not relevant
Setting minimum pagefile size equal to physical RAM size -- not relevant

 

Also, as of 2014-06-13:

Different keyboard on different USB port -- not relevant
NOT using a keyboard modifier key to wake up the display (shift, ctrl, alt or windows key) -- not relevant
Not using a keyboard at all to wake up the display -- not relevant



Edited: 06/13/2014 at 08:24 AM by Solid State Brain
 06/13/2014 07:48 AM
User is offline View Users Profile Print this message

Author Icon
Solid State Brain
Peon

Posts: 32
Joined: 05/31/2014

Today I discovered that when the problem occurs - in addition of switching the multimonitor configuration or disconnecting the video cable - turning the monitor off with its power button, waiting a few seconds then turning it on again restores the video output.

It's as if sometimes the discrete AMD GPU can't detect the display's power state and won't attempt detecting it again if it misses the right timeframe for that.

I don't have this problem with the Intel IGP.

My primary display is a Dell U2412M, the operating system is Windows 8.1 Update.

 

EDIT: I thought the culprit could have been a hidden power setting of my primary display, but in the end it turned out it isn't.



Edited: 06/13/2014 at 05:51 PM by Solid State Brain
 06/13/2014 09:35 PM
User is offline View Users Profile Print this message

Author Icon
Emopunk
Peon

Posts: 38
Joined: 04/21/2014

Originally posted by: Solid State Brain Today I discovered that when the problem occurs - in addition of switching the multimonitor configuration or disconnecting the video cable - turning the monitor off with its power button, waiting a few seconds then turning it on again restores the video output.

 

It's as if sometimes the discrete AMD GPU can't detect the display's power state and won't attempt detecting it again if it misses the right timeframe for that.

 

I don't have this problem with the Intel IGP.

 

My primary display is a Dell U2412M, the operating system is Windows 8.1 Update.

 

 

 

EDIT: I thought the culprit could have been a hidden power setting of my primary display, but in the end it turned out it isn't.

 

Maybe I wrote it in some previous post, I can't remember at the moment. I have the same monitor and the problem is happening both on Win 8.1 and 7. It seems to happen with every driver release I have tested, too.



-------------------------
We need to concentrate on more than meets the eye
 06/13/2014 09:55 PM
User is offline View Users Profile Print this message

Author Icon
Solid State Brain
Peon

Posts: 32
Joined: 05/31/2014

That's interesting to know. Maybe not a coincidence. It could be that my Sapphire HD7770 GPU and perhaps other GCN-based AMD video cards manufactured in about the same period do not like much this display or similar ones for some reason.

It's probably not just a monitor's problem, as it seems to work fine with the integrated Intel GPU I have. Perhaps a GPU BIOS upgrade could fix that; only problem is that Sapphire won't provide any officially. Besides, I was already trying out the latest 100% compatible one from Sapphire I could find (unofficially) with apparently no improvement over the original one.

So, I tried flashing an even newer BIOS image from an XFX HD7770 instead of a Sapphire one. It works but I have one less DisplayPort video output than with proper ones. I will report if there will be any improvement. One thing I immediately noticed is that the latest Catalyst drivers now correctly detect my video card as an HD7770, not an "R7 Series".

Here are the BIOS images I tried so far:

015.014.000.001.000746 (original BIOS for my GPU) -- bug occurs
015.017.000.000.000000 (last 100% compatible Sapphire HD7770 BIOS) -- bug occurs
015.030.000.001.000000 (most recent working HD7770 BIOS I could find on techpowerup.com - 1 DisplayPort output not working) -- still testing

 

@Emopunk: I don't know what videocard you have, but speaking of BIOS version, what's yours?

EDIT:

Ah I see you reported your GPU model here a while back.

I confirm the problem is present for both 14.4 and 14.6 on my R9 270

Being the R9 270 a rather recent model, this makes me skeptical whether having flashed an updated BIOS version on my HD7770 will improve anything or not.

 



Edited: 06/13/2014 at 10:23 PM by Solid State Brain
 06/14/2014 08:12 AM
User is offline View Users Profile Print this message

Author Icon
Solid State Brain
Peon

Posts: 32
Joined: 05/31/2014

@Emopunk: unfortunately, as I started suspecting from your post (this happens on an R9 270 as well), flashing the newest working firmware I could find for my Sapphire HD7770 didn't solve anything. This morning the issue occurred again. Merely physically turning the monitor off, then turning it on again restored video signal, but that's still annoying nevertheless.

I wonder if this happens on GPUs that are not from Sapphire as well.

 06/14/2014 11:50 PM
User is offline View Users Profile Print this message

Author Icon
Eydee
Ninja Zombie Killer

Posts: 4905
Joined: 12/27/2008

Take a note, people, you won't find bios updates for video cards. Manufacturers stopped providing them for everyone because people flashed anything they could find. Sapphire was probably the first to introduce this new policy.

You need to contact Sapphire tech support, explain your problem and request a bios update. This is the only way to get a proper one.



-------------------------

CPU: AMD Phenom II X4 810 @ 3120MHz | RAM: Kingmax 2x2GB DDR2 800 @ 833MHz| MoBo: MSI K9A2 CF v1.0 (BIOS: 1.D)| GPU: Asus HD 6850 1024MB (DirectCu) @ 835/1135MHz | Display: L24FHD | PSU: PC Power & Cooling Silencer 750 Quad | OS: MS Windows 3.11 Pro x64

Statistics
86401 users are registered to the AMD Support and Game forum.
There are currently 4 users logged in.

FuseTalk Hosting Executive Plan v3.2 - © 1999-2014 FuseTalk Inc. All rights reserved.