Tesla Compute Cluster driver released non-display driver for 64-bit Windows Server 08/08 R2

Accelerated Computing CUDA CUDA Programming and Performance
1

We’ve released the Tesla Compute Cluster driver for 64-bit Windows, which solves a lot of issues people have encountered with trying to deploy Teslas on Windows.

Note that this is not a display driver. You won’t get any 2D, D3D, or GL acceleration from this driver. Only CUDA is supported in this release. However, you do get exclusive mode, Remote Desktop support, and services support.

The driver is available here:

[url=“Official Drivers | NVIDIA”]Official Drivers | NVIDIA Tesla → 1U System → S1070 → Windows Server 2008 x64

2

Hi,

some questions:

I have tested on a GTX 275 (adding device id to the driver inf) with AMD 5850 as display good work…

(hmm I hope by saying that please don’t block this possibilty similar to how are you going to block double prec potential by slowing down 4x on geforce fermi cards…)

By “Only CUDA is supported in this release” you mean OpenCL is supported right now?

I have tested OpenCL ocldeviceQuery and fails to search platform ID and is clear the driver doesn’t include opencl.dll nor nvcompiler.dll…

Using that dlls from 196.34 don’t work altough nvcuda.dll seems to have Khronos ICD entry points…

are you going to support OpenCL soon on Tesla driver?

I have not tested but are CUDA programs using textures mean to work? I assume yes altough a graphics feature…

Also what about for DirectCompute? i.e for DirectCompute apps not using graphics are you going to support it?

And finally Nexus, I have access to Nexus beta which supports by using two Nvidia cards in one PC debugging in one computer… is this supported by using as display device an ATI card now one Nvidia card has not to have extended desktop

I hope you add support for Nexus for Tesla Computing driver in case it isn’t supported right now…

an ATI card…

3

I’ve tried this driver on Server 2008R2 Datacenter X64, with M1060 Teslas - it fails to install, citing no hardware found that matches. When manually installing through device manager, I’m informed this operating system is not supported. Also note that the .exe referenced in the release notes (NVIDIADisplayWinServer2008x64(196_28)Int.exe) is NOT part of the driver package. Attempted via local console, remote KVM, and RDP, as the ‘real’ local administrator, and as other domain users, both administratively and not. No dice. Any suggestions? We’ve looked forward to this driver for a while…

4

milson correctly determined that I messed up the INF for M1060. This will be fixed in the next build. For now, if you’re having a problem with M1060, replace the nvwd.inf in the package with this nvwd.inf.

5

I tried installing this driver on Windows Server 2008 R2 x64 with a GeForce GT 220. I understand this is not an officially supported graphics card for this driver, but my server only accepts single-slot low-profile graphics cards and I have to make do with this. I tried adding the GT 220 device ID to the INF file, but when I try to manually assign the driver to the device I get the following error message:

“The driver selected for this device does not support this version of Windows”

The release notes document’s wording seems to imply that non-supported graphics cards should work as CUDA devices with this driver. I’d love to have this driver working in this configuration because my server is supposed to be headless and I have to manage it through Remote Desktop.

Any ideas?

6

Ok, I revisited this problem and found out that I had incorrectly modified the INF file for the TCC driver. I now have the driver loading for my GT220 and CUDA programs running through Remote Desktop, which is fantastic.

In short, these are the modifications I had to do to NVWD.inf from the TCC package:

[NVIDIA_SetA_Devices.NTamd64.6.0]
%NVIDIA_DEV.0A20.01% = Section001, PCI\VEN_10DE&DEV_0A20

[NVIDIA_SetA_Devices.NTamd64.6.1]
%NVIDIA_DEV.0A20.01% = Section002, PCI\VEN_10DE&DEV_0A20

[Strings]
NVIDIA_DEV.0A20.01 = “NVIDIA GeForce GT 220”

7

Hi,

We are also having a problem getting CUDA to work on a new box and from my reading this driver would appear to be the answer to our problems. Unfortunately, following the link provided ends up returning a null page. Has the driver been removed? Does anyone have a copy?

Thanks,

Mark

8

Hi

Few questions:

1)Is this driver compatible with Windows Vista and Windows 7 x64?

2)Is this driver compatible with Windows XP x64?

3)Is the driver compatible with Tesla C1060?

Thanks!

Sergio

9

Answering myself:

  1. No

2)No

  1. Yes

Sergio

10

I’ve done the method with jblazquez and it works smoothly. Thanks.

11

I downloaded the 32 bit version for Windows 7 and have been trying to do the same with that but am not having any luck. Really would be nice to have a driver that just allows us to use the CUDA capabilities of these cards.

I have this card in this machine specifically for processing and in fact am using an on-board video for connection when required. I see a lot of cards with fewer cores listed then the GT set. It would be nice to just make it inclusive so cutting up the inf file wouldn’t be required.

K.

12

So, just to clarify, we need to select the driver for a 1U S1070 to get one that will work for a C2050?

13

So, just to clarify, we need to select the driver for a 1U S1070 to get one that will work for a C2050?

14

As there isn’t any answer on oscarb’s question corresponding to OpenCL, I’d like to repeat the question:
Is there now OpenCL support on the TCC driver (i.e. 259.03) for Windows Server 2008?
Will there be in future? When?

15

As there isn’t any answer on oscarb’s question corresponding to OpenCL, I’d like to repeat the question:
Is there now OpenCL support on the TCC driver (i.e. 259.03) for Windows Server 2008?
Will there be in future? When?

16

Could a mod or anyone with knowledge explain the differences between these drivers and normal GeForce gaming drivers? I see that these support GeForce cards but are they recommended for gaming? What would be the purpose of installing these on a GeForce card and would TCC interfere with functionality at all?

Thanks.

EDIT: Sorry, I just noticed the date of this post, so apologies for reviving a dead thread. I was referring to the new 260.61 Tesla drivers.

17

Could a mod or anyone with knowledge explain the differences between these drivers and normal GeForce gaming drivers? I see that these support GeForce cards but are they recommended for gaming? What would be the purpose of installing these on a GeForce card and would TCC interfere with functionality at all?

Thanks.

EDIT: Sorry, I just noticed the date of this post, so apologies for reviving a dead thread. I was referring to the new 260.61 Tesla drivers.

18

You can’t install TCC drivers on a GeForce card. (or at least, you can’t use the TCC mode on a GeForce card)

The 260 release is a very improved version of TCC–essentially it’s per-device rather than across the entire system.

19

You can’t install TCC drivers on a GeForce card. (or at least, you can’t use the TCC mode on a GeForce card)

The 260 release is a very improved version of TCC–essentially it’s per-device rather than across the entire system.

20

I have GTX 470 - I’m not interested in gaming at all - only CUDA. However I really need CUDA on RDP, is it doable via this driver with modified inf files ?

many thanks

Mirko