Register - Login
Views: 85600327
Main - Memberlist - Active users - Calendar - Wiki - IRC Chat - Online users
Ranks - Rules/FAQ - JCS - Stats - Latest Posts - Color Chart - Smilies
09-20-17 08:31:37 PM

Jul - Computers - The 900MB VRAM Problem New poll - New thread - New reply
Next newer thread | Next older thread
RanAS
Member
Level: 29


Posts: 27/272
EXP: 147172
For next: 713

Since: 10-10-14
From: Brazil

Since last post: 9 days
Last activity: 4 days

Posted on 02-28-15 06:21:55 PM Link | Quote
This is a very complicated problem that will probably never get a real solution. This is going to be a long post. Here's the thing:
My computer freezes when more than 900MB of VRAM (video RAM) is used. The most normal answer would be that this is a bad video card, but I don't think so. You'll see why.

Before I continue, here are some specifications of my computer:
20.0 inch Samsung SyncMaster B2030N
Intel Core 2 Duo E7500 @ 2.93GHz
4GB Unbuffered DIMM DDR3-1333 667MHz
640GB 7200 RPM SATA-II Hard Drive
nVidia GeForce GT610 PCIE 1.1 2GB (manufactured by PointOfView Graphics)
Windows 7 Service Pack 1


Now, this problem happens with any program while running Windows (will get in detail about that in a little bit). If any program/application uses 900MB of video RAM or more, it'll freeze the computer. Windows will display some artifacts and after a while, this: "The NVIDIA Windows Kernel Mode Driver has stopped working and recovered succesfully". And everything goes back to normal...except sometimes one of the programs may stop functioning properly.

I was a little curious of why this problem was happening, so I decided to use VMST (Video Memory Stress Test) while on Windows, and it was then that I noticed that when it reached/surpassed 900MB of VRAM, it freezes, but VMST doesn't notice anything unusual and will simply say "Test passed!" and countinue on with the testing, which will freeze the computer once again. The only way out is to use Ctrl+Alt+Delete.

Here's the "fun" part of the problem:
-This problem does not happen on Linux with the nVidia drivers installed, although I'm not sure if these were made by nVidia.
-This problem also doesn't happen on the VMST bootable USB disk (you basically install VMST in your USB drive for it to boot up directly after the BIOS). I even tested memtest86+ with a bootable USB, and again nothing happened. All tests were successfull without any weirdness.
-I tried different two different driver versions on Windows to see if the problem went away. First one being the one in the box with the Video Card. Didn't work (in fact, the problem was even worse in this version, permanently freezing Windows!). The most recent driver also didn't solve the issue.
-I thought that the problem would be older drivers that are still being used by Windows somewhere, so I used a program called DriverSweeper, or something of the sort, to completely remove the old drivers (I did some other minor things too). It did remove the drivers, but it didn't solve the problem.

So, for this problem, I can only think of a few answers. None of them have an easy solution.

1. The problem is Windows itself either because of damaged DLL or because of "Windows Rot", so it would be necessary a reinstall.
2. The problem is that Linux and the USB test can't use every resource from the GFX card, while Windows can. This proves the issue is one of the features of the Video Card.
3. The problem is in the drivers, specifically the ones that nVidia provides, which is why this problem doesn't happen on Linux and on the USB drives.

I heavily think that answer 2 is wrong, because Linux still uses more than 900MB regardless. It shouldn't matter what "features are on". The USB thing is even more absurd for obvious reasons.

I don't know if answer 1 would solve this problem, but I'm certain that if it somehow did do it, it would be some obscure DLL/registry conflict no one knows about.

I'm slightly more confident with answer 3. nVidia says that the drivers they provide are generic, and while they should work fine, the recommended version is a personalized one by the manufacturer. The manufacturer's website (PointOfView Graphics), on the other hand, redirects me to the nVidia website when I try to download the drivers. We have a problem here. There is no personalized driver for this video card.

I quite sincerely don't know if there's anything else I should've written here, and I don't know myself if what I'm writing is BS and the reason this is happening is something completely different, but I know one thing: No one I contacted so far was able to even understand what the problem could be. I was only able to say a few thing in IRC here and there. This makes me think that I have quite a...unique problem.

This is the entire problem so far, and the end of this post. Hopefully, everyone understands the horrible phrasing I tend to make in these posts, and hopefully, someone can help me in this weird situation.
Joe
Common spammer
🗿
Level: 103


Posts: 3184/3286
EXP: 11462183
For next: 9223

Since: 08-02-07
From: Pororoca

Since last post: 46 days
Last activity: 1 day

Posted on 02-28-15 07:45:24 PM Link | Quote
What does event viewer have to say when the driver crashes? It might be able to provide more useful information than the error message you've seen so far.

Which components of the graphics drivers are you installing? Does the issue occur when you do a minimal installation with just the graphics driver and none of the additional components?

Do you have up-to-date drivers for your motherboard's chipset?

Does underclocking your RAM (through the BIOS) change anything?

Does underclocking your video card (using MSI Afterburner or similar) change anything?

Have you tried taking it out and putting it back in?
RanAS
Member
Level: 29


Posts: 28/272
EXP: 147172
For next: 713

Since: 10-10-14
From: Brazil

Since last post: 9 days
Last activity: 4 days

Posted on 03-01-15 06:19:52 AM Link | Quote
Originally posted by Joe
What does event viewer have to say when the driver crashes? It might be able to provide more useful information than the error message you've seen so far.

The event viewer has quite a few things about this issue. These are the events that happened when Windows froze:
ERROR 1
- System
- Provider
[ Name] Display
- EventID 4101
[ Qualifiers] 0
Level 3
Task 0
Keywords 0x80000000000000
- TimeCreated
[ SystemTime] 2015-03-01T08:28:26.000000000Z
EventRecordID 218124
Channel System
Computer E7500
Security
- EventData
nvlddmkm

Description of the event:
The video driver nvlddmkm stop responding and recovered successfully.

ERROR 2
- System
- Provider
[ Name] Windows Error Reporting
- EventID 1001
[ Qualifiers] 0
Level 4
Task 0
Keywords 0x80000000000000
- TimeCreated
[ SystemTime] 2015-03-01T08:28:33.000000000Z
EventRecordID 47673
Channel Application
Computer E7500
Security
- EventData

0
LiveKernelEvent
Unavailable
0

C:\Windows\LiveKernelReports\WATCHDOG\WD-20150301-0528.dmp C:\Windows\Temp\WER-1985362-0.sysdata.xml C:\Windows\Temp\WER67C7.tmp.WERInternalMetadata.xml
C:\ProgramData\Microsoft\Windows\WER\ReportQueue\Kernel_0_0_cab_02be6825

0
eda83763-bfec-11e4-a407-c83a35dc9052
4

Description of the event (I translated this into english):

Failure in shared memory, type 0
Event Name: LiveKernelEvent
Answer: Unavailable
Id of the CAB file: 0

Signature:
P1:
P2:
P3:
P4:
P5:
P6:
P7:
P8:
P9:
P10:

Attached Files:
C:\Windows\LiveKernelReports\WATCHDOG\WD-20150301-0528.dmp
C:\Windows\Temp\WER-1985362-0.sysdata.xml
C:\Windows\Temp\WER67C7.tmp.WERInternalMetadata.xml

These files are available at:
C:\ProgramData\Microsoft\Windows\WER\ReportQueue\Kernel_0_0_cab_02be6825

Analisys Symbol:
Verifying solution again: 0
Report Id: eda83763-bfec-11e4-a407-c83a35dc9052
Report Status: 4


I would like to bring up that a event happened right before this one, that was basically asking the Desktop Window Management service to start, but it couldn't because of WDDM.

I also found these strange errors with no description. These are from a while ago.

ERROR 1
- System
- Provider
[ Name] nvlddmkm
- EventID 13
[ Qualifiers] 49322
Level 2
Task 0
Keywords 0x80000000000000
- TimeCreated
[ SystemTime] 2014-11-11T00:22:21.313501900Z
EventRecordID 166337
Channel System
Computer E7500
Security
- EventData
\Device\Video5
Graphics Exception: Shader Program Header 11 Error
0000000002003000000000000D00AAC0000000000000000000000000000000000000000000000000

ERROR 2
- System
- Provider
[ Name] NVIDIA OpenGL Driver
- EventID 1
[ Qualifiers] 49152
Level 2
Task 0
Keywords 0x80000000000000
- TimeCreated
[ SystemTime] 2014-10-27T15:16:08.000000000Z
EventRecordID 32460
Channel Application
Computer E7500
Security
- EventData
Too many errors occurred, which indicates a serious problem from which we cannot recover. The application must close. Error code: 1 Visit http://nvidia.custhelp.com/cgi-bin/nvidia.cfg/php/enduser/std_adp.php?p_faqid=3006 for more information.

I did visit the website and it seems like whatever program was using OpenGL at that time was able to crash the video driver.

Which components of the graphics drivers are you installing? Does the issue occur when you do a minimal installation with just the graphics driver and none of the additional components?

I generally install the normal drivers, audio drivers (for the built-in HD audio thing that I'll probably never use) and GeForce Experience (yes, that stupid thing). Just now, I installed a clean version (clean install) of the most recent drivers for this GFX card without any of that "extra" they make you install. The problem is still happening.


Do you have up-to-date drivers for your motherboard's chipset?

I unistalled them once I bought a external GFX card, about a year ago, but I still did download the newest drivers recently just in case.


Does underclocking your RAM (through the BIOS) change anything?

My BIOS doesn't support underclocking/overclocking of anything whatsoever. It is quite old, actually (Made in 2005). Maybe I should update it at some point, but it doesn't seem to be causing any issues, at least it didn't when I had my previous video card.
Either way, the BIOS clearly says that the video card has 3585MB (obviously more than 2GB because of the "little" bit of VRAM that gets shared)


Does underclocking your video card (using MSI Afterburner or similar) change anything?

I'll have to try that later...but meanwhile, have some specifications via GPU-Z:


I did not overclock or underclock anything yet. I also just installed CPU-Z in the case you need more information.


Have you tried taking it out and putting it back in?

Yes, a few times. That didn't change much.



I should probably mention, this is what normally happens when I use VMST:
Changing video mode to 640x480x16...OK
[01-03-2015 05:27:13] Test started for "NVIDIA GeForce GT 610"...
Trying 16bpp RGB:565 mode...OK
TEST FAIL (Code: 887601C2)
Trying 16bpp RGB:555 mode...NOT SUPPORTED (Code: 8876024E)
Trying 16bpp BGR:565 mode...NOT SUPPORTED (Code: 8876024E)
Trying 32bpp RGB:888 mode...NOT SUPPORTED (Code: 8876024E)
Trying 32bpp BGR:888 mode...NOT SUPPORTED (Code: 8876024E)
NO PASS COMPLETED

What is happening here is that, before the first test completes, I press Ctrl+Alt+Del which cancels every single test so it'll not keep going with the test, because that'll crash Windows again.

Second thing is, when I enabled the option to enable onscreen buffer ...well, things got weird, I'll put this in a paste instead.
VMST Log 1

Yeah...the screen also got a lot more glitchy this time, but it didn't freeze the computer.
I discovered via GPU-Z that the point where everything glitches is exactly at 992MB, which is bad considering that's half of my VRAM that is now unusable.

I would also like to bring up the point that this is a WINDOWS ONLY problem. VMST outside of Windows has zero issues testing the 2GB of VRAM that my GFX card has. While on Windows it does say that, for safety reasons, it'll only test 1024MB of VRAM because the driver says that there's only 1998MB of VRAM available. That doesn't happen outside of Windows.
Just so you know, here is a paste of the log files for VMST while on the USB disk:
VMST Log 2
Joe
Common spammer
🗿
Level: 103


Posts: 3185/3286
EXP: 11462183
For next: 9223

Since: 08-02-07
From: Pororoca

Since last post: 46 days
Last activity: 1 day

Posted on 03-01-15 10:53:06 AM Link | Quote

The video driver nvlddmkm stop responding and recovered successfully.
"Your graphics driver was taking a long time to do something so I killed it and restarted it. I hope you don't mind."
Originally posted by RanAS

Do you have up-to-date drivers for your motherboard's chipset?
I unistalled them once I bought a external GFX card, about a year ago, but I still did download the newest drivers recently just in case.
You do need chipset drivers. You probably don't need the drivers for the onboard graphics, but it's not always possible to install them separately.
Originally posted by RanAS
My BIOS doesn't support underclocking/overclocking of anything whatsoever. It is quite old, actually (Made in 2005). Maybe I should update it at some point, but it doesn't seem to be causing any issues, at least it didn't when I had my previous video card.
A BIOS update probably won't add features like that. It's also pretty unlikely a BIOS update would help, but I wouldn't say it's impossible.
Originally posted by RanAS
Either way, the BIOS clearly says that the video card has 3585MB (obviously more than 2GB because of the "little" bit of VRAM that gets shared)
Can you change the amount of shared VRAM? If so, try the lowest possible value.
Originally posted by RanAS
I did not overclock or underclock anything yet.
According to GPU-Z, your graphics card is using automatic throttling to save power. Is there an option somewhere to disable that? (Or set the power/performance balance to maximum performance?) If successful, you should see the GPU clocks listed in the Sensors tab of GPU-Z matching those in the Graphics Card tab even when your GPU is idle.
RanAS
Member
Level: 29


Posts: 29/272
EXP: 147172
For next: 713

Since: 10-10-14
From: Brazil

Since last post: 9 days
Last activity: 4 days

Posted on 03-02-15 03:51:53 AM (last edited by RanAS at 03-02-15 12:35:43 PM) Link | Quote
Originally posted by Joe
You do need chipset drivers. You probably don't need the drivers for the onboard graphics, but it's not always possible to install them separately.

Oh. I didn't know that. Unfortunately for me, these are only one driver for both the chipset and the onboard graphics. I'll install these drivers later today.

EDIT: Evidently, they are separate drivers, one for the chipset and one for the video. Either way, I have the chipset drivers installed, same error, although this time Windows said that the NVIDIA Windows Kernel Mode Driver, Version 347.52 stopped working instead of the nvlddmkm driver (they're probably the same thing, though).

Originally posted by Joe
Can you change the amount of shared VRAM? If so, try the lowest possible value.

Only through the BIOS, and, you know it already, the BIOS has almost nothing about voltages/frequencies.
I found one option in there that lets me change the frequency of the DRAM frequency from 800Hz to 1600Hz or something like that. It is currently set to Automatic.

Originally posted by Joe
According to GPU-Z, your graphics card is using automatic throttling to save power. Is there an option somewhere to disable that? (Or set the power/performance balance to maximum performance?) If successful, you should see the GPU clocks listed in the Sensors tab of GPU-Z matching those in the Graphics Card tab even when your GPU is idle.

From what I've seen, nVidia does not let you disable that feature, it is required to install another program not from nVidia to change it. I believe the name is Nvidia Inspector?

I can change the 3D options from Balanced to Quality or Performace, but they didn't change the throttling whatsoever (maybe because that option was only for the 3D graphics?). GPU-Z says that the VDDC changes from 0.9 (Idle) to 1.1 becasue of the throttling.


...I was wondering about something. Is it normal that, when the computer is off, but my monitor is on I get the message "Verify the cables" as if the cables are not connected? Before I had this GFX card it just displayed a black screen with nothing in it, although this is also the first GPU card I have that displays the GPU BIOS before the actual BIOS, so I don't know, it's just out of curiosity.
Joe
Common spammer
🗿
Level: 103


Posts: 3186/3286
EXP: 11462183
For next: 9223

Since: 08-02-07
From: Pororoca

Since last post: 46 days
Last activity: 1 day

Posted on 03-02-15 04:22:11 PM Link | Quote
Originally posted by RanAS
[...] this time Windows said that the NVIDIA Windows Kernel Mode Driver, Version 347.52 stopped working instead of the nvlddmkm driver (they're probably the same thing, though).
Yes, they are the same thing. (The version number doesn't match your GPU-Z screenshot from earlier. You've updated your drivers since then, right?)

Originally posted by RanAS
Originally posted by Joe
Can you change the amount of shared VRAM? If so, try the lowest possible value.

Only through the BIOS, and, you know it already, the BIOS has almost nothing about voltages/frequencies.
That sounds about right. Try setting it to the lowest value, if you haven't already, and see if it makes a difference.

Originally posted by RanAS
I found one option in there that lets me change the frequency of the DRAM frequency from 800Hz to 1600Hz or something like that. It is currently set to Automatic.
Try setting it to 800Hz and see if it makes things better. If it does, keep turning it up to higher values until it starts crashing again. Don't go above 1333Hz. (If 800Hz doesn't improve things, leave it on automatic.)

Originally posted by RanAS
I can change the 3D options from Balanced to Quality or Performace, but they didn't change the throttling whatsoever (maybe because that option was only for the 3D graphics?).
Yeah, that sounds like it only affects 3D graphics. It couldn't hurt to try changing it and see if it makes a difference.

Originally posted by RanAS
...I was wondering about something. Is it normal that, when the computer is off, but my monitor is on I get the message "Verify the cables" as if the cables are not connected?
Yes.
RanAS
Member
Level: 29


Posts: 30/272
EXP: 147172
For next: 713

Since: 10-10-14
From: Brazil

Since last post: 9 days
Last activity: 4 days

Posted on 03-03-15 03:31:06 AM (last edited by RanAS at 03-03-15 12:36:11 PM) Link | Quote
1. Yes, I have updated my drivers.
2. I can't change the amount of shared memory.

Originally posted by Joe
Try setting it to 800Hz and see if it makes things better. If it does, keep turning it up to higher values until it starts crashing again. Don't go above 1333Hz. (If 800Hz doesn't improve things, leave it on automatic.)

So, the values were 800MHz and 1600MHz, I guess I didn't pay enough attention to that, and me being somewhat foolish, I changed the value either way. Then, the BIOS stopped working. Good times. I'll try to reset it later (removing the PC battery). I'm currently using my phone to post this.

EDIT: BIOS reset is successfull! Hooray! Unfortunately...I lost the custom configuration for my cooling system. But fortunately, yet unexpectedly, Shared Memory is now at zero, somehow!



I really think the problem is not hardware here, instead, that the problem is somewhere between the hardware and the user interface.
Next newer thread | Next older thread
Jul - Computers - The 900MB VRAM Problem New poll - New thread - New reply




Rusted Logic

Acmlmboard - commit 2f1bc75 [2017-08-27]
©2000-2017 Acmlm, Xkeeper, Inuyasha, et al.

27 database queries, 5 query cache hits.
Query execution time: 0.150684 seconds
Script execution time: 0.011449 seconds
Total render time: 0.162133 seconds