BUG detected (Inconsistency with Drivers)
[color="#FF0000"][u][b]Pretty Important for the Developer Drivers, a potential bug.[/b][/u][/color]

I'd like to inform the Driver Developers of some odd behavior I've recently discovered.

I've observed some strange behavior in correlation with your Nvidia Desktop/Mobile drivers.

I'm using a Macbook Pro 6.2 with a 330M GT (256MB VRAM) booted into Native EFI (Same inconsistency also happens in Bootcamp with BIOS emulation mode, so this shouldn't matter at all).

Since there are so few, or nobody who reported this yet, I can only assume it happens on this specific model or some specific models.

DeviceID: DEV_0A29
Subsys: 00C7106B&REV_A2

Description of the bug:
-----------------------

Newer Drivers (275.33 and up) report lower total VRAM allocated in Windows itself, and all independent external tools. Older drivers (pre 275.33) seems not to have been affected at all. Tested with:

- Process Manager (procxp).
- GPU-Z.
- Windows Experience Score for Gaming/Graphics increased/decreased score 5.6 versus 6.6
- Video Memory Stress Test.


Results:

[img]http://farm8.staticflickr.com/7143/6767270019_f722e19585_b.jpg[/img]



Results:

[img]http://farm8.staticflickr.com/7026/6767270025_23706432db_b.jpg[/img]




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


So my question is, what is causing this, and since it also happens in the so called recent beta verda driver I hope the Graphics Driver developers aren't accidentally overlooking this potential problem for their future drivers. I'm pretty sure they arent even aware of it.

[color="#FF0000"][b]At this point I'm trying to pinpointing at which driver this sudden VRAM decrease begins to happen. I'll update when I have the results. I dont think any new function or feature can explain a sudden decrease in available/mapped/allocated VRAM and even so, such function should have the ability to be disabled.

I doubt that running an older Driver should give me better results in all benchmarks, and that Windows Experience Score shows me 1.0 points higher in both Gaming and Graphics. But that is what is going on at the moment.
[/b][/color]


EDIT: I've discovered that the sudden VRAM decrease happens when installing driver; [u][b]275.33[/b][/u]

Every driver never than that also has the same strange behavior.

Anything below this driver version seems to work fine and gives me the full 256MB and higher Windows Experience Scores/3DMark benchmarks. Also noticeable in Games which run close or a bit over 256MB. Stuttering from loading textures from Windows Shared Video Memory to Dedicated Memory are gone.

Also 275.27 [u]Beta[/u] still works fine and fully detects my VRAM. I do not see any noticeable changes in the .inf files either.

[i][b]Gmanuel[/b][/i], if you are able to, perhaps you can relay this to the Developer team and keep me updated on what could possibly be wrong here? Seems like a bug to me affecting a very small amount of (Mobile?) GPU's.

I'm pretty sure this shouldn't be happening, and I'd assume they would like to have it fixed before they release the new 'Verda' drivers.
Pretty Important for the Developer Drivers, a potential bug.



I'd like to inform the Driver Developers of some odd behavior I've recently discovered.



I've observed some strange behavior in correlation with your Nvidia Desktop/Mobile drivers.



I'm using a Macbook Pro 6.2 with a 330M GT (256MB VRAM) booted into Native EFI (Same inconsistency also happens in Bootcamp with BIOS emulation mode, so this shouldn't matter at all).



Since there are so few, or nobody who reported this yet, I can only assume it happens on this specific model or some specific models.



DeviceID: DEV_0A29

Subsys: 00C7106B&REV_A2



Description of the bug:

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



Newer Drivers (275.33 and up) report lower total VRAM allocated in Windows itself, and all independent external tools. Older drivers (pre 275.33) seems not to have been affected at all. Tested with:



- Process Manager (procxp).

- GPU-Z.

- Windows Experience Score for Gaming/Graphics increased/decreased score 5.6 versus 6.6

- Video Memory Stress Test.





Results:



Image







Results:



Image









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





So my question is, what is causing this, and since it also happens in the so called recent beta verda driver I hope the Graphics Driver developers aren't accidentally overlooking this potential problem for their future drivers. I'm pretty sure they arent even aware of it.



At this point I'm trying to pinpointing at which driver this sudden VRAM decrease begins to happen. I'll update when I have the results. I dont think any new function or feature can explain a sudden decrease in available/mapped/allocated VRAM and even so, such function should have the ability to be disabled.



I doubt that running an older Driver should give me better results in all benchmarks, and that Windows Experience Score shows me 1.0 points higher in both Gaming and Graphics. But that is what is going on at the moment.







EDIT: I've discovered that the sudden VRAM decrease happens when installing driver; 275.33



Every driver never than that also has the same strange behavior.



Anything below this driver version seems to work fine and gives me the full 256MB and higher Windows Experience Scores/3DMark benchmarks. Also noticeable in Games which run close or a bit over 256MB. Stuttering from loading textures from Windows Shared Video Memory to Dedicated Memory are gone.



Also 275.27 Beta still works fine and fully detects my VRAM. I do not see any noticeable changes in the .inf files either.



Gmanuel, if you are able to, perhaps you can relay this to the Developer team and keep me updated on what could possibly be wrong here? Seems like a bug to me affecting a very small amount of (Mobile?) GPU's.



I'm pretty sure this shouldn't be happening, and I'd assume they would like to have it fixed before they release the new 'Verda' drivers.

#1
Posted 01/26/2012 06:47 PM   
Bump
Bump

#2
Posted 01/27/2012 03:50 PM   
275x to 285x are garbage drivers for the Macbook pro's. Upgrade to the 290x on a clean install of the drivers and leave it at that. No issues on my Macbook Pro 330m Win 7 x64. I get the impression they rewrote large parts of the drivers and ended up releasing with a lot of bugs. I.E mini displayport external display not working for awhile on the 275x series drivers. Good luck sir and stick to the 260x series if you are still having issues. It's what I did until they released the 290's.
275x to 285x are garbage drivers for the Macbook pro's. Upgrade to the 290x on a clean install of the drivers and leave it at that. No issues on my Macbook Pro 330m Win 7 x64. I get the impression they rewrote large parts of the drivers and ended up releasing with a lot of bugs. I.E mini displayport external display not working for awhile on the 275x series drivers. Good luck sir and stick to the 260x series if you are still having issues. It's what I did until they released the 290's.

#3
Posted 01/28/2012 12:16 AM   
[quote name='Chosin137' date='28 January 2012 - 01:16 AM' timestamp='1327709768' post='1361594']
275x to 285x are garbage drivers for the Macbook pro's. Upgrade to the 290x on a clean install of the drivers and leave it at that. No issues on my Macbook Pro 330m Win 7 x64. I get the impression they rewrote large parts of the drivers and ended up releasing with a lot of bugs. I.E mini displayport external display not working for awhile on the 275x series drivers. Good luck sir and stick to the 260x series if you are still having issues. It's what I did until they released the 290's.
[/quote]

The 290 series still give me less dedicated memory then it should, perhaps you can confirm its also happening on your 330M with the latest drivers? Because this is not right. :(

I'd like to get some support for this.

The Macbook Pro's 330M is no different then any other 330M laptop GPU, the only difference is that it has a slightly modified BIOS with Apple carved into it and some edited Clock speed values. Like I said I'm not even 'using' bootcamp which would imply having a Compatibility BIOS layer which could explain this, i'm running native EFI. Since it only happens after the 275 drivers I can only assume its a bug since then. If you could check your system and confirm this odd memory gap behavior It would certainly help my case.
[quote name='Chosin137' date='28 January 2012 - 01:16 AM' timestamp='1327709768' post='1361594']

275x to 285x are garbage drivers for the Macbook pro's. Upgrade to the 290x on a clean install of the drivers and leave it at that. No issues on my Macbook Pro 330m Win 7 x64. I get the impression they rewrote large parts of the drivers and ended up releasing with a lot of bugs. I.E mini displayport external display not working for awhile on the 275x series drivers. Good luck sir and stick to the 260x series if you are still having issues. It's what I did until they released the 290's.





The 290 series still give me less dedicated memory then it should, perhaps you can confirm its also happening on your 330M with the latest drivers? Because this is not right. :(



I'd like to get some support for this.



The Macbook Pro's 330M is no different then any other 330M laptop GPU, the only difference is that it has a slightly modified BIOS with Apple carved into it and some edited Clock speed values. Like I said I'm not even 'using' bootcamp which would imply having a Compatibility BIOS layer which could explain this, i'm running native EFI. Since it only happens after the 275 drivers I can only assume its a bug since then. If you could check your system and confirm this odd memory gap behavior It would certainly help my case.

#4
Posted 01/28/2012 03:31 PM   
Download Speccy from piroform and post a picture of the graphics tab/ (section)Specifically the Geforce GT 330M section.

GeForce GT 330M
GPU GT216
Device ID 10DE-0A29
Revision A3
Subvendor Apple Computer (106B)
Current Performance Level Level 1
Current GPU Clock 135 MHz
Current Memory Clock 135 MHz
Current Shader Clock 270 MHz
Voltage 0.800 V
Technology 41 nm
Die Size 100 nm²
Release Date 2009
DirectX Support 10.1
OpenGL Support 4.1
Bus Interface PCI Express x16
Temperature 78 °C
ForceWare version 290.53
BIOS Version 70.16.58.0a.00
ROPs 8
Shaders 48 unified
Physical Memory 512 MB
Virtual Memory 224 MB
Download Speccy from piroform and post a picture of the graphics tab/ (section)Specifically the Geforce GT 330M section.



GeForce GT 330M

GPU GT216

Device ID 10DE-0A29

Revision A3

Subvendor Apple Computer (106B)

Current Performance Level Level 1

Current GPU Clock 135 MHz

Current Memory Clock 135 MHz

Current Shader Clock 270 MHz

Voltage 0.800 V

Technology 41 nm

Die Size 100 nm²

Release Date 2009

DirectX Support 10.1

OpenGL Support 4.1

Bus Interface PCI Express x16

Temperature 78 °C

ForceWare version 290.53

BIOS Version 70.16.58.0a.00

ROPs 8

Shaders 48 unified

Physical Memory 512 MB

Virtual Memory 224 MB

#5
Posted 01/29/2012 01:46 AM   
You sure the onboard chip which the macbooks switch to constantly when not in 3D programs isn't sapping the available VRAM? Update the intel onboard GPU drivers and see what happens.
You sure the onboard chip which the macbooks switch to constantly when not in 3D programs isn't sapping the available VRAM? Update the intel onboard GPU drivers and see what happens.

i7 3930k - Asus Rampage IV Formula - Kingston 16gb 2400mhz - Samsung 128gb 840 PRO - Seagate 2tb hdd - SLI Geforce TITANs - BenQ 2411T - Corsair 600T silver - Corsair H100 - Seasonic 1250w - Logitech G500 - Edifier 1900TII - Cherry G80-3000

#6
Posted 01/29/2012 05:26 AM   
[quote name='oblivion20' date='29 January 2012 - 06:26 AM' timestamp='1327814796' post='1362098']
You sure the onboard chip which the macbooks switch to constantly when not in 3D programs isn't sapping the available VRAM? Update the intel onboard GPU drivers and see what happens.
[/quote]

No, In bootcamp (Bios emulation layer) the Intel IGP gets disabled just like an ordinary PC with BIOS where you disable your integrated GPU. In EFI the IGP is detected but disabled and it cannot utilize any resources. And since it only happens after a specific driver version and up, it doesn't have anything to do with the IGP.


[quote name='Chosin137' date='29 January 2012 - 02:46 AM' timestamp='1327801601' post='1362064']
Download Speccy from piroform and post a picture of the graphics tab/ (section)Specifically the Geforce GT 330M section.
[/quote]

The problem is; Some applications detect the total dedicated memory by reading it directly from the card or BIOS, but this is not always the accurate total memory that is allocated by Windows or the Driver.

Exactly the same as that Windows or dxdiag reports you have a 2.4GHz processor and its only running at 1.2GHz when you look in CPU-Z which reports real-time clock frequencies. (due to wrong bios setting or powersaving)

Like I shown before, real-time VRAM memory usage monitors consequently show me lowered total VRAM allocated with the newer drivers, this is a bug.

To strengthen my case, I've made additional screenshots and added my dxdiag in both cases.
If this is not enough proof, I don't know what will.

Drivers:
Verde Notebook Release 260 BETA
Verde Notebook Release 260 WHQL
Verde Notebook Release 265 BETA
Verde Driver v266.58 WHQL
Verde Driver v267.24 BETA
Verde Driver v267.76 WHQL
Verde Driver v270.51 BETA
Verde Driver v270.61 WHQL
Verde Driver v275.27 BETA

256MB allocated:
[img]http://farm8.staticflickr.com/7151/6783231403_07db6d12dd_b.jpg[/img]

And now comes the culprit:

Verde 275.33 Driver WHQL
Verde 275.50 Driver BETA
Verde 280.19 Driver BETA
Verde 280.26 Driver WHQL
Verde 285.27 Driver BETA
Verde 285.38 Driver BETA
Verde 285.62 Driver WHQL
Verde 285.79 Driver BETA
Verde 290.36 Driver BETA
Verde 290.53 Driver BETA

All of these perform less vram memory allocation;

224MB:
[img]http://farm8.staticflickr.com/7171/6783231405_16796e07b3_b.jpg[/img]

Notice the GPU-Z, Nvidia Inspector and FurMark Memory load increase, these three program show their memory usage based on the total dedicated memory reported by the card itself, but since there isn't 256MB in reality allocated by windows/drivers, the programs think its being utilized. It's nicely shown in the screenshots above.

But when you look at programs like procxp you clearly see the memory utilization hasnt increased, but the limit has been reduced. Tada! I doubt one will blame this on a visual bug mhh?

I'd like Nvidia to look into this because this is not normal.
[quote name='oblivion20' date='29 January 2012 - 06:26 AM' timestamp='1327814796' post='1362098']

You sure the onboard chip which the macbooks switch to constantly when not in 3D programs isn't sapping the available VRAM? Update the intel onboard GPU drivers and see what happens.





No, In bootcamp (Bios emulation layer) the Intel IGP gets disabled just like an ordinary PC with BIOS where you disable your integrated GPU. In EFI the IGP is detected but disabled and it cannot utilize any resources. And since it only happens after a specific driver version and up, it doesn't have anything to do with the IGP.





[quote name='Chosin137' date='29 January 2012 - 02:46 AM' timestamp='1327801601' post='1362064']

Download Speccy from piroform and post a picture of the graphics tab/ (section)Specifically the Geforce GT 330M section.





The problem is; Some applications detect the total dedicated memory by reading it directly from the card or BIOS, but this is not always the accurate total memory that is allocated by Windows or the Driver.



Exactly the same as that Windows or dxdiag reports you have a 2.4GHz processor and its only running at 1.2GHz when you look in CPU-Z which reports real-time clock frequencies. (due to wrong bios setting or powersaving)



Like I shown before, real-time VRAM memory usage monitors consequently show me lowered total VRAM allocated with the newer drivers, this is a bug.



To strengthen my case, I've made additional screenshots and added my dxdiag in both cases.

If this is not enough proof, I don't know what will.



Drivers:

Verde Notebook Release 260 BETA

Verde Notebook Release 260 WHQL

Verde Notebook Release 265 BETA

Verde Driver v266.58 WHQL

Verde Driver v267.24 BETA

Verde Driver v267.76 WHQL

Verde Driver v270.51 BETA

Verde Driver v270.61 WHQL

Verde Driver v275.27 BETA



256MB allocated:

Image



And now comes the culprit:



Verde 275.33 Driver WHQL

Verde 275.50 Driver BETA

Verde 280.19 Driver BETA

Verde 280.26 Driver WHQL

Verde 285.27 Driver BETA

Verde 285.38 Driver BETA

Verde 285.62 Driver WHQL

Verde 285.79 Driver BETA

Verde 290.36 Driver BETA

Verde 290.53 Driver BETA



All of these perform less vram memory allocation;



224MB:

Image



Notice the GPU-Z, Nvidia Inspector and FurMark Memory load increase, these three program show their memory usage based on the total dedicated memory reported by the card itself, but since there isn't 256MB in reality allocated by windows/drivers, the programs think its being utilized. It's nicely shown in the screenshots above.



But when you look at programs like procxp you clearly see the memory utilization hasnt increased, but the limit has been reduced. Tada! I doubt one will blame this on a visual bug mhh?



I'd like Nvidia to look into this because this is not normal.

#7
Posted 01/29/2012 03:44 PM   
Added Dxdiag reports who knows what Nvidia can do with it;
Added Dxdiag reports who knows what Nvidia can do with it;

#8
Posted 01/29/2012 05:41 PM   
Bump, I expect someone from Nvidia to at least look into this and make me somewhat more comfortable /blarg.gif' class='bbc_emoticon' alt=':/' />

Please.
Bump, I expect someone from Nvidia to at least look into this and make me somewhat more comfortable /blarg.gif' class='bbc_emoticon' alt=':/' />



Please.

#9
Posted 01/30/2012 09:04 AM   
[quote name='Marctraider' date='30 January 2012 - 10:04 AM' timestamp='1327914260' post='1362480']
Bump, I expect someone from Nvidia to at least look into this and make me somewhat more comfortable /blarg.gif' class='bbc_emoticon' alt=':/' />

Please.
[/quote]
they won't, this is a user-to-user forum. go [url="http://nvidia-submit.custhelp.com/cgi-bin/nvidia_submit.cfg/php/enduser/std_alp.php"]here[/url] with your support request.
[quote name='Marctraider' date='30 January 2012 - 10:04 AM' timestamp='1327914260' post='1362480']

Bump, I expect someone from Nvidia to at least look into this and make me somewhat more comfortable /blarg.gif' class='bbc_emoticon' alt=':/' />



Please.



they won't, this is a user-to-user forum. go here with your support request.
#10
Posted 01/30/2012 12:18 PM   
Scroll To Top