Can TCC mode be enabled with Quadro 2000M/4000M
Hello.
I'm new to CUDA computing and learning it for realtime video processing.
I'm working with Windows7 64bit.

I learned about the WDDM problem in this forum and wanted to use my Quadro 2000M/4000M in TCC mode.
(My PCs are DELL Mobile Precision M4600 and M6600. Each of them has a built-in Intel Grahics chip.
I want to use Quadro only for computing in TCC mode, while using Intel chip for GUI.)

I tried to set TCC mode with:
.\nvidia-smi.exe -i 0 -dm 1 (at C:\Program Files\NVIDIA Corporation\NVSMI with administrator privilege)

It didn't work with an error message "Unable to determine display mode for GPU 0:1:0: N/A".

With the following command:
.\nvidia-smi.exe -i 0 -fdm 1

It failed with a different error message "Unable to set driver model for GPU 0:1:0: N/A".

Is it because Quadro 2000M/4000M can't be used in TCC mode?
".\nvidia-smi.exe -h" tells me that the tool can be used only for Tesla and Quadro 4000/5000/6000/7000.
I expected that TCC mode can be enabled with Quadro 4000M at least because it was a mobile version of Quadro 4000.
Was it a wrong assumption?

I understand that GeForce can't be used in TCC mode from the following topic in this forum.
"Cannot set up CUDA toolkit for Quadro FX 3800 under win7"
http://forums.nvidia.com/index.php?showtopic=212458

The driver versions are:
Quadro 2000M : 275.33
Quadro 4000M : 276.05

Unfortunately, I don't have any Quadro4000/5000/6000 for comparison.
Your help will be greatly appreciated.
Hello.

I'm new to CUDA computing and learning it for realtime video processing.

I'm working with Windows7 64bit.



I learned about the WDDM problem in this forum and wanted to use my Quadro 2000M/4000M in TCC mode.

(My PCs are DELL Mobile Precision M4600 and M6600. Each of them has a built-in Intel Grahics chip.

I want to use Quadro only for computing in TCC mode, while using Intel chip for GUI.)



I tried to set TCC mode with:

.\nvidia-smi.exe -i 0 -dm 1 (at C:\Program Files\NVIDIA Corporation\NVSMI with administrator privilege)



It didn't work with an error message "Unable to determine display mode for GPU 0:1:0: N/A".



With the following command:

.\nvidia-smi.exe -i 0 -fdm 1



It failed with a different error message "Unable to set driver model for GPU 0:1:0: N/A".



Is it because Quadro 2000M/4000M can't be used in TCC mode?

".\nvidia-smi.exe -h" tells me that the tool can be used only for Tesla and Quadro 4000/5000/6000/7000.

I expected that TCC mode can be enabled with Quadro 4000M at least because it was a mobile version of Quadro 4000.

Was it a wrong assumption?



I understand that GeForce can't be used in TCC mode from the following topic in this forum.

"Cannot set up CUDA toolkit for Quadro FX 3800 under win7"

http://forums.nvidia.com/index.php?showtopic=212458



The driver versions are:

Quadro 2000M : 275.33

Quadro 4000M : 276.05



Unfortunately, I don't have any Quadro4000/5000/6000 for comparison.

Your help will be greatly appreciated.

#1
Posted 02/14/2012 01:25 PM   
Hi Heffalump,

Unfortunately Quadro 4000[b]M[/b] is not NVSMI supported product. The name is very similar but Quadro 4000M uses a different version of the chip - GF104. Quadro 4000 uses GF100 (see [url="http://en.wikipedia.org/wiki/Comparison_of_NVIDIA_graphics_processing_units"]this link[/url]).
NVSMI allows changing driver model only on devices listed in NVSMI help message.

Hope this explains your question. Let me know you if you have any other questions.

Regards,
Przemyslaw Zych
Hi Heffalump,



Unfortunately Quadro 4000M is not NVSMI supported product. The name is very similar but Quadro 4000M uses a different version of the chip - GF104. Quadro 4000 uses GF100 (see this link).

NVSMI allows changing driver model only on devices listed in NVSMI help message.



Hope this explains your question. Let me know you if you have any other questions.



Regards,

Przemyslaw Zych

#2
Posted 02/16/2012 12:47 AM   
Dear Mr. Przemyslaw Zych,

Thank you for your reply to my question.

> Unfortunately Quadro 4000M is not NVSMI supported product.
> The name is very similar but Quadro 4000M uses a different version of the chip - GF104.
> Quadro 4000 uses GF100 (see this link).
> NVSMI allows changing driver model only on devices listed in NVSMI help message.

Thank you for the link to the helpful comparison chart.

According to the chart, Quadro 5000M uses GF100.
Does NVSMI allow changing driver model on Quadro 5000M (if I could afford it)?

I want to have a mobile CUDA solution with notebook PC, dedicating NVIDIA GPU to calculation while using Intel GPU for GUI.
Coincidentally, the poster of another topic "[url="http://forums.nvidia.com/index.php?showtopic=222961"]"Intel GPU + NVIDIA GPU"[/url] seems to have a similar idea.

I want to use NVIDIA mobile GPU in TCC mode because the WDDM problem on Windows7 degrades performance of CUDA applications by giving extra overhead on CUDA kernel launching; besides the TDR problem.
[url="http://forums.nvidia.com/index.php?showtopic=150236"]"CUDA slower in Windows 7 than in Windows XP"[/url]

Tesla and Quadro 4000/5000/6000 are too large in volume for my purpose.

I will appreciate both positive and negative inputs.
Dear Mr. Przemyslaw Zych,



Thank you for your reply to my question.



> Unfortunately Quadro 4000M is not NVSMI supported product.

> The name is very similar but Quadro 4000M uses a different version of the chip - GF104.

> Quadro 4000 uses GF100 (see this link).

> NVSMI allows changing driver model only on devices listed in NVSMI help message.



Thank you for the link to the helpful comparison chart.



According to the chart, Quadro 5000M uses GF100.

Does NVSMI allow changing driver model on Quadro 5000M (if I could afford it)?



I want to have a mobile CUDA solution with notebook PC, dedicating NVIDIA GPU to calculation while using Intel GPU for GUI.

Coincidentally, the poster of another topic ""Intel GPU + NVIDIA GPU" seems to have a similar idea.



I want to use NVIDIA mobile GPU in TCC mode because the WDDM problem on Windows7 degrades performance of CUDA applications by giving extra overhead on CUDA kernel launching; besides the TDR problem.

"CUDA slower in Windows 7 than in Windows XP"



Tesla and Quadro 4000/5000/6000 are too large in volume for my purpose.



I will appreciate both positive and negative inputs.

#3
Posted 02/16/2012 03:12 AM   
Hi,

Because of the high QA costs some features (like TCC) are supported only on high end platforms.

To get full list of all supported platforms you can run "nvidia-smi -h"
[code]
Supported products:

Tesla: S1070, S2050, C1060, C2050/70/75, M2050/70/75/90, X2070/90
Quadro: 4000, 5000, 6000, 7000 and M2070-Q
Other: All other products are unsupported
[/code]
As far as I know Quadro 5000M is not going to support TCC but I'll try to confirm this information internally.

If this is an option for you CUDA on Linux doesn't have any of the WDDM Windows drawbacks.

Let me know if you have other questions.
Hi,



Because of the high QA costs some features (like TCC) are supported only on high end platforms.



To get full list of all supported platforms you can run "nvidia-smi -h"



Supported products:



Tesla: S1070, S2050, C1060, C2050/70/75, M2050/70/75/90, X2070/90

Quadro: 4000, 5000, 6000, 7000 and M2070-Q

Other: All other products are unsupported


As far as I know Quadro 5000M is not going to support TCC but I'll try to confirm this information internally.



If this is an option for you CUDA on Linux doesn't have any of the WDDM Windows drawbacks.



Let me know if you have other questions.

#4
Posted 02/18/2012 01:36 AM   
Dear Mr. Przemyslaw Zych,

Thank you very much for clarification.

I'll start investigation about the Linux option.
Dear Mr. Przemyslaw Zych,



Thank you very much for clarification.



I'll start investigation about the Linux option.

#5
Posted 02/21/2012 02:39 PM   
I've double checked. Quadro 5000[b]M[/b] will [b]not[/b] support TCC.
I've double checked. Quadro 5000M will not support TCC.

#6
Posted 02/21/2012 10:29 PM   
Scroll To Top