CUDA 2.0 Beta 2 GTX support, more Linux distros...
  1 / 2    
We're releasing a new CUDA 2.0 beta today. Support for GT200-based cards with new features (double precision, shared atomics, voting, double registers--compile with --arch compute_13), more documentation, more support for newer Linux distributions. more examples, and more...

It should be on the website soon, but in the interim, download links:

[url="http://developer.download.nvidia.com/compute/cuda/2.0-Beta2/linux/driver/NVIDIA-Linux-x86-177.13-pkg1.run"]Linux 177.13 driver for x86[/url]
[url="http://developer.download.nvidia.com/compute/cuda/2.0-Beta2/linux/driver/NVIDIA-Linux-x86_64-177.13-pkg2.run"]Linux 177.13 driver for x86-64[/url]

UPDATED WINDOWS DRIVERS - now works on all CUDA cards
This is a beta driver intended for CUDA development only.

[url="http://developer.download.nvidia.com/compute/cuda/2.0-Beta2/windows/driver/NVIDIADisplayWin2K(177_35)Int.exe"]WinXP 177.35 driver[/url]
[url="http://developer.download.nvidia.com/compute/cuda/2.0-Beta2/windows/driver/NVIDIADisplayWin2KAMD64(177_35)Int.exe"]WinXP 64-bit 177.35 driver[/url]
[url="http://developer.download.nvidia.com/compute/cuda/2.0-Beta2/windows/driver/NVIDIADisplayWinVista(177_35)Int.exe"]Vista 32-bit 177.35 driver[/url]
[url="http://developer.download.nvidia.com/compute/cuda/2.0-Beta2/windows/driver/NVIDIADisplayWinVista64(177_35)Int.exe"]Vista 64-bit 177.35 driver[/url]

CUDA Toolkit 2.0b2:

[url="http://developer.download.nvidia.com/compute/cuda/2.0-Beta2/linux/toolkit/NVIDIA_CUDA_Toolkit_2.0beta2_Fedora8_x86.run"]Fedora 8 x86[/url]
[url="http://developer.download.nvidia.com/compute/cuda/2.0-Beta2/linux/toolkit/NVIDIA_CUDA_Toolkit_2.0beta2_Fedora8_x86_64.run"]Fedora 8 x86-64[/url]
[url="http://developer.download.nvidia.com/compute/cuda/2.0-Beta2/linux/toolkit/NVIDIA_CUDA_Toolkit_2.0beta2_rhel4.6_x86.run"]RHEL 4.6 x86[/url]
[url="http://developer.download.nvidia.com/compute/cuda/2.0-Beta2/linux/toolkit/NVIDIA_CUDA_Toolkit_2.0beta2_rhel4.6_x86_64.run"]RHEL 4.6 x86-64[/url]
[url="http://developer.download.nvidia.com/compute/cuda/2.0-Beta2/linux/toolkit/NVIDIA_CUDA_Toolkit_2.0beta2_rhel5.1_x86.run"]RHEL 5.1 x86[/url]
[url="http://developer.download.nvidia.com/compute/cuda/2.0-Beta2/linux/toolkit/NVIDIA_CUDA_Toolkit_2.0beta2_rhel5.1_x86_64.run"]RHEL 5.1 x86-64[/url]
[url="http://developer.download.nvidia.com/compute/cuda/2.0-Beta2/linux/toolkit/NVIDIA_CUDA_Toolkit_2.0beta2_Suse10.2_x86.run"]SuSE 10.2 x86[/url]
[url="http://developer.download.nvidia.com/compute/cuda/2.0-Beta2/linux/toolkit/NVIDIA_CUDA_Toolkit_2.0beta2_Suse10.2_x86_64.run"]SuSE 10.2 x86-64[/url]
[url="http://developer.download.nvidia.com/compute/cuda/2.0-Beta2/linux/toolkit/NVIDIA_CUDA_Toolkit_2.0beta2_Suse10.3_x86.run"]SuSE 10.3 x86[/url]
[url="http://developer.download.nvidia.com/compute/cuda/2.0-Beta2/linux/toolkit/NVIDIA_CUDA_Toolkit_2.0beta2_Suse10.3_x86_64.run"]SuSE 10.3 x86-64[/url]
[url="http://developer.download.nvidia.com/compute/cuda/2.0-Beta2/linux/toolkit/NVIDIA_CUDA_Toolkit_2.0beta2_sled10sp1_x86.run"]SLED 10 SP1 x86[/url]
[url="http://developer.download.nvidia.com/compute/cuda/2.0-Beta2/linux/toolkit/NVIDIA_CUDA_Toolkit_2.0beta2_sled10sp1_x86_64.run"]SLED 10 SP1 x86-64[/url]
[url="http://developer.download.nvidia.com/compute/cuda/2.0-Beta2/linux/toolkit/NVIDIA_CUDA_Toolkit_2.0beta2_Ubuntu7.04_x86.run"]Ubuntu 7.04 x86[/url]
[url="http://developer.download.nvidia.com/compute/cuda/2.0-Beta2/linux/toolkit/NVIDIA_CUDA_Toolkit_2.0beta2_Ubuntu7.04_x86_64.run"]Ubuntu 7.04 x86-64[/url]
[url="http://developer.download.nvidia.com/compute/cuda/2.0-Beta2/linux/toolkit/NVIDIA_CUDA_Toolkit_2.0beta2_Ubuntu7.10_x86.run"]Ubuntu 7.10 x86[/url]
[url="http://developer.download.nvidia.com/compute/cuda/2.0-Beta2/linux/toolkit/NVIDIA_CUDA_Toolkit_2.0beta2_Ubuntu7.10_x86_64.run"]Ubuntu 7.10 x86-64[/url]

[url="http://developer.download.nvidia.com/compute/cuda/2.0-Beta2/windows/toolkit/NVIDIA_CUDA_toolkit_2.0beta2_Win32.exe"]32-bit WinXP or Vista[/url]
[url="http://developer.download.nvidia.com/compute/cuda/2.0-Beta2/windows/toolkit/NVIDIA_CUDA_toolkit_2.0beta2_Win64.exe"]64-bit WinXP or Vista[/url]

CUDA SDK 2.0b2:
UPDATED WINDOWS RELEASE
Fixes glew64.dll errors.

[url="http://developer.download.nvidia.com/compute/cuda/2.0-Beta2/linux/sdk/NVIDIA_CUDA_sdk_2.0beta2_linux.run"]Linux[/url]
[url="http://developer.download.nvidia.com/compute/cuda/2.0-Beta2/windows/sdk/NVIDIA_CUDA_sdk_2.0beta2b_Win32.exe"]32-bit WinXP or Vista[/url]
[url="http://developer.download.nvidia.com/compute/cuda/2.0-Beta2/windows/sdk/NVIDIA_CUDA_sdk_2.0beta2b_Win64.exe"]64-bit WinXP or Vista[/url]

CUDA Visual Profiler:

[url="http://developer.download.nvidia.com/compute/cuda/2.0-Beta2/profiler/CudaVisualProfiler_windows_1.0_13June08.zip"]Windows[/url]
[url="http://developer.download.nvidia.com/compute/cuda/2.0-Beta2/profiler/CudaVisualProfiler_linux_1.0_13June08.tar.gz"]Linux[/url]

Docs (included, but in case you're curious before downloading):

[url="http://developer.download.nvidia.com/compute/cuda/2.0-Beta2/docs/Programming_Guide_2.0beta2.pdf"]CUDA 2.0b2 Programmers Guide[/url]
[url="http://developer.download.nvidia.com/compute/cuda/2.0-Beta2/docs/PG_Revision_Hist_2.0beta2.txt"]CUDA 2.0b2 Programmers Guide Revision History[/url]
[url="http://developer.download.nvidia.com/compute/cuda/2.0-Beta2/docs/CudaRefMan_2.0beta2.pdf"]CUDA 2.0b2 Reference Manual[/url]
[url="http://developer.download.nvidia.com/compute/cuda/2.0-Beta2/docs/SDK_Rel_Notes_Win_2.0beta2.pdf"]CUDA 2.0b2 SDK Release Notes - Windows[/url]
[url="http://developer.download.nvidia.com/compute/cuda/2.0-Beta2/docs/SDK_Rel_Notes_Linux_2.0beta2.txt"]CUDA 2.0b2 SDK Release Notes - Linux[/url]
[url="http://developer.download.nvidia.com/compute/cuda/2.0-Beta2/docs/CudaVisualProfiler_README_1.0_13June08.txt"]CUDA 2.0b2 Visual Profiler Readme[/url]
[url="http://developer.download.nvidia.com/compute/cuda/2.0-Beta2/docs/CudaVisualProfiler_windows_release_notes_1.0_13June08.txt"]CUDA 2.0b2 Visual Profiler Release Notes - Windows[/url]
[url="http://developer.download.nvidia.com/compute/cuda/2.0-Beta2/docs/CudaVisualProfiler_linux_release_notes_1.0_13June08.txt"]CUDA 2.0b2 Visual Profiler Release Notes - Linux[/url]
We're releasing a new CUDA 2.0 beta today. Support for GT200-based cards with new features (double precision, shared atomics, voting, double registers--compile with --arch compute_13), more documentation, more support for newer Linux distributions. more examples, and more...



It should be on the website soon, but in the interim, download links:



Linux 177.13 driver for x86

Linux 177.13 driver for x86-64



UPDATED WINDOWS DRIVERS - now works on all CUDA cards

This is a beta driver intended for CUDA development only.



WinXP 177.35 driver

WinXP 64-bit 177.35 driver

Vista 32-bit 177.35 driver

Vista 64-bit 177.35 driver



CUDA Toolkit 2.0b2:



Fedora 8 x86

Fedora 8 x86-64

RHEL 4.6 x86

RHEL 4.6 x86-64

RHEL 5.1 x86

RHEL 5.1 x86-64

SuSE 10.2 x86

SuSE 10.2 x86-64

SuSE 10.3 x86

SuSE 10.3 x86-64

SLED 10 SP1 x86

SLED 10 SP1 x86-64

Ubuntu 7.04 x86

Ubuntu 7.04 x86-64

Ubuntu 7.10 x86

Ubuntu 7.10 x86-64



32-bit WinXP or Vista

64-bit WinXP or Vista



CUDA SDK 2.0b2:

UPDATED WINDOWS RELEASE

Fixes glew64.dll errors.



Linux

32-bit WinXP or Vista

64-bit WinXP or Vista



CUDA Visual Profiler:



Windows

Linux



Docs (included, but in case you're curious before downloading):



CUDA 2.0b2 Programmers Guide

CUDA 2.0b2 Programmers Guide Revision History

CUDA 2.0b2 Reference Manual

CUDA 2.0b2 SDK Release Notes - Windows

CUDA 2.0b2 SDK Release Notes - Linux

CUDA 2.0b2 Visual Profiler Readme

CUDA 2.0b2 Visual Profiler Release Notes - Windows

CUDA 2.0b2 Visual Profiler Release Notes - Linux

#1
Posted 06/18/2008 10:59 PM   
so what is about the osx users which are still stuck with 1.1 beta...

do we get some love too?
so what is about the osx users which are still stuck with 1.1 beta...



do we get some love too?

#2
Posted 06/18/2008 11:30 PM   
The new beta installed fine in Suse 10.3 and seems to work fine - SDK compiled, tests passed, etc. Thx.

The new beta does not seem to want to let me overclock my 8800GT like the previous beta, however. Attempts to do so have locked up the system. Before I was getting 600->700 MHz core, 900->1000 MHz memory, now not so much. The autoscan for overclocking in nvidia-settings suggested 700/1100, which promptly froze the system.

(AMD Phenom 9600 on a Gigabyte MA790X motherboard; PNY 8800GT.)

(Still searching for code for matrix division: solution to A*X=B; glimmers of happiness over the horizon, but no joy yet...)
The new beta installed fine in Suse 10.3 and seems to work fine - SDK compiled, tests passed, etc. Thx.



The new beta does not seem to want to let me overclock my 8800GT like the previous beta, however. Attempts to do so have locked up the system. Before I was getting 600->700 MHz core, 900->1000 MHz memory, now not so much. The autoscan for overclocking in nvidia-settings suggested 700/1100, which promptly froze the system.



(AMD Phenom 9600 on a Gigabyte MA790X motherboard; PNY 8800GT.)



(Still searching for code for matrix division: solution to A*X=B; glimmers of happiness over the horizon, but no joy yet...)

#3
Posted 06/19/2008 12:26 AM   
It seems that the 177.35 driver for windows only support GT200 gpu.
Which version is the best for g80/g92 gpu?
It seems that the 177.35 driver for windows only support GT200 gpu.

Which version is the best for g80/g92 gpu?

#4
Posted 06/19/2008 12:45 AM   
1) The 177.35 don't work with my GF8500GT/GF8600GTS under Vista x64(SP1).

2) On the other hand, to separate the x86 and x64 installers is a very bad idea... because if I want to cross-compile from Vista x64 to winXP 32 I cannot... the toolkit says it's already installed and I need to uninstall it... why you cannot just make one installer and deploy the DLLS and .LIBs into two different folders(x86 and x64)? /no.gif' class='bbc_emoticon' alt=':no:' />

3) On the other hand you continue deploying only the /MT(Multithreaded ) cudart.lib... For IJW/CLI/CLR/.NET/.NET STL we [b]NEED[/b] also the /MD ( Multithreaded DLL) libraries /crying.gif' class='bbc_emoticon' alt=':'(' /> If not, it's impossible to link the CUDA libraries without getting a zillion of linking errors, /NODEFAULTLIB problems, etc...

4) And, finally, the libraries generates memory leaks under VS2008... because the deployment you did its for VS2005... and the VS2008 crt is different...

For example... see the FBX SDK... Autodesk deploys the .LIB as:

fbx_vs2003_mt.lib -> x86, Multithread for Visual Studio 2003
fbx_vs2003_md.lib -> x86, Multithread DLL for Visual Studio 2003
fbx_vs2003_mtd.lib -> x86, Multithread Debug for Visual Studio 2003
fbx_vs2003_mdd.lib -> x86, Multithread DLL Debug for Visual Studio 2003
fbx_vs2003_mt_x64.lib -> x64, Multithread for Visual Studio 2003
fbx_vs2003_md_x64.lib -> x64, Multithread DLL for Visual Studio 2003
fbx_vs2003_mtd_x64.lib -> x64, Multithread Debug for Visual Studio 2003
fbx_vs2003_mdd_x64.lib -> x64, Multithread DLL Debug for Visual Studio 2003

fbx_vs2005_mt.lib -> x86, Multithread for Visual Studio 2005
fbx_vs2005_md.lib -> x86, Multithread DLL for Visual Studio 2005
fbx_vs2005_mtd.lib -> x86, Multithread Debug for Visual Studio 2005
fbx_vs2005_mdd.lib -> x86, Multithread DLL Debug for Visual Studio 2005
fbx_vs2005_mt_x64.lib -> x64, Multithread for Visual Studio 2005
fbx_vs2005_md_x64.lib -> x64, Multithread DLL for Visual Studio 2005
fbx_vs2005_mtd_x64.lib -> x64, Multithread Debug for Visual Studio 2005
fbx_vs2005_mdd_x64.lib -> x64, Multithread DLL Debug for Visual Studio 2005

fbx_vs2008_mt.lib -> x86, Multithread for Visual Studio 2008
fbx_vs2008_md.lib -> x86, Multithread DLL for Visual Studio 2008
fbx_vs2008_mtd.lib -> x86, Multithread Debug for Visual Studio 2008
fbx_vs2008_mdd.lib -> x86, Multithread DLL Debug for Visual Studio 2008
fbx_vs2008_mt_x64.lib -> x64, Multithread for Visual Studio 2008
fbx_vs2008_md_x64.lib -> x64, Multithread DLL for Visual Studio 2008
fbx_vs2008_mtd_x64.lib -> x64, Multithread Debug for Visual Studio 2008
fbx_vs2008_mdd_x64.lib -> x64, Multithread DLL Debug for Visual Studio 2008

That's a correct deployment... the current CUDA 2 toolkit/sdk is incomplete... because without that you will get lots of problems with .NET and memory leaks due to C CRT incompatibilities... If you could add a very simple Managed C++/CLR/IJW/WinForms/.NET STL example to the SDK you could see what i'm referrering to...

5) some examples complain because can't find glew64.dll
1) The 177.35 don't work with my GF8500GT/GF8600GTS under Vista x64(SP1).



2) On the other hand, to separate the x86 and x64 installers is a very bad idea... because if I want to cross-compile from Vista x64 to winXP 32 I cannot... the toolkit says it's already installed and I need to uninstall it... why you cannot just make one installer and deploy the DLLS and .LIBs into two different folders(x86 and x64)? /no.gif' class='bbc_emoticon' alt=':no:' />



3) On the other hand you continue deploying only the /MT(Multithreaded ) cudart.lib... For IJW/CLI/CLR/.NET/.NET STL we NEED also the /MD ( Multithreaded DLL) libraries /crying.gif' class='bbc_emoticon' alt=':'(' /> If not, it's impossible to link the CUDA libraries without getting a zillion of linking errors, /NODEFAULTLIB problems, etc...



4) And, finally, the libraries generates memory leaks under VS2008... because the deployment you did its for VS2005... and the VS2008 crt is different...



For example... see the FBX SDK... Autodesk deploys the .LIB as:



fbx_vs2003_mt.lib -> x86, Multithread for Visual Studio 2003

fbx_vs2003_md.lib -> x86, Multithread DLL for Visual Studio 2003

fbx_vs2003_mtd.lib -> x86, Multithread Debug for Visual Studio 2003

fbx_vs2003_mdd.lib -> x86, Multithread DLL Debug for Visual Studio 2003

fbx_vs2003_mt_x64.lib -> x64, Multithread for Visual Studio 2003

fbx_vs2003_md_x64.lib -> x64, Multithread DLL for Visual Studio 2003

fbx_vs2003_mtd_x64.lib -> x64, Multithread Debug for Visual Studio 2003

fbx_vs2003_mdd_x64.lib -> x64, Multithread DLL Debug for Visual Studio 2003



fbx_vs2005_mt.lib -> x86, Multithread for Visual Studio 2005

fbx_vs2005_md.lib -> x86, Multithread DLL for Visual Studio 2005

fbx_vs2005_mtd.lib -> x86, Multithread Debug for Visual Studio 2005

fbx_vs2005_mdd.lib -> x86, Multithread DLL Debug for Visual Studio 2005

fbx_vs2005_mt_x64.lib -> x64, Multithread for Visual Studio 2005

fbx_vs2005_md_x64.lib -> x64, Multithread DLL for Visual Studio 2005

fbx_vs2005_mtd_x64.lib -> x64, Multithread Debug for Visual Studio 2005

fbx_vs2005_mdd_x64.lib -> x64, Multithread DLL Debug for Visual Studio 2005



fbx_vs2008_mt.lib -> x86, Multithread for Visual Studio 2008

fbx_vs2008_md.lib -> x86, Multithread DLL for Visual Studio 2008

fbx_vs2008_mtd.lib -> x86, Multithread Debug for Visual Studio 2008

fbx_vs2008_mdd.lib -> x86, Multithread DLL Debug for Visual Studio 2008

fbx_vs2008_mt_x64.lib -> x64, Multithread for Visual Studio 2008

fbx_vs2008_md_x64.lib -> x64, Multithread DLL for Visual Studio 2008

fbx_vs2008_mtd_x64.lib -> x64, Multithread Debug for Visual Studio 2008

fbx_vs2008_mdd_x64.lib -> x64, Multithread DLL Debug for Visual Studio 2008



That's a correct deployment... the current CUDA 2 toolkit/sdk is incomplete... because without that you will get lots of problems with .NET and memory leaks due to C CRT incompatibilities... If you could add a very simple Managed C++/CLR/IJW/WinForms/.NET STL example to the SDK you could see what i'm referrering to...



5) some examples complain because can't find glew64.dll

#5
Posted 06/19/2008 03:13 AM   
Installed and on XP64 and it seems to work fine except for the demos requiring glew64.dll.
/thumbsdown.gif' class='bbc_emoticon' alt=':thumbsdown:' />

Any idea why it is not part of the distrib?

I agree with the above issues regarding the /MD and cross compiling.

But I am already quite happy to have a GTX280 and a CUDA version for it so soon :)

Good work guys. /thumbup.gif' class='bbc_emoticon' alt=':thumbup:' />
Installed and on XP64 and it seems to work fine except for the demos requiring glew64.dll.

/thumbsdown.gif' class='bbc_emoticon' alt=':thumbsdown:' />



Any idea why it is not part of the distrib?



I agree with the above issues regarding the /MD and cross compiling.



But I am already quite happy to have a GTX280 and a CUDA version for it so soon :)



Good work guys. /thumbup.gif' class='bbc_emoticon' alt=':thumbup:' />

#6
Posted 06/19/2008 07:06 AM   
The driver does not work with FX570m on Vista, even with a modded .inf. Very sad :(
However, on XP it works (with modded .inf), and slightly faster than the previous.
The driver does not work with FX570m on Vista, even with a modded .inf. Very sad :(

However, on XP it works (with modded .inf), and slightly faster than the previous.

#7
Posted 06/19/2008 07:27 AM   
Hello,

has anyone successfully installed the new driver 177.35 (NVIDIADisplayWin2K_177.35,...)? What kind of hardware is required?

I tried to install two drivers on two different setups:

Windows XP (x86), GeForce 8600 GT, NVIDIADisplayWin2K_177.35
Windows Vista (x64), GeForce 9600 GT, NVIDIADisplayWinVista64_177.35

In both cases setup is terminated:
"NVIDIA Setup Error
The NVIDIA Setup program could not locate any drivers that are compatible with your current hardware. ..."

Any ideas?

Thanks & best regards
Hello,



has anyone successfully installed the new driver 177.35 (NVIDIADisplayWin2K_177.35,...)? What kind of hardware is required?



I tried to install two drivers on two different setups:



Windows XP (x86), GeForce 8600 GT, NVIDIADisplayWin2K_177.35

Windows Vista (x64), GeForce 9600 GT, NVIDIADisplayWinVista64_177.35



In both cases setup is terminated:

"NVIDIA Setup Error

The NVIDIA Setup program could not locate any drivers that are compatible with your current hardware. ..."



Any ideas?



Thanks & best regards

#8
Posted 06/19/2008 08:48 AM   
GTX280 , Driver 177.35 and CUDA 2.0 Beta 2 works fine on XP64 for me.

I have not tried yet under Vista64.
GTX280 , Driver 177.35 and CUDA 2.0 Beta 2 works fine on XP64 for me.



I have not tried yet under Vista64.

#9
Posted 06/19/2008 02:02 PM   
Would it be possible to have the reference manual either in HTML form or to include links within the PDF? It's nice to be able to scroll through the list of function names and then jump straight to ones of interest.
Would it be possible to have the reference manual either in HTML form or to include links within the PDF? It's nice to be able to scroll through the list of function names and then jump straight to ones of interest.

#10
Posted 06/19/2008 03:59 PM   
[quote name='StickGuy' date='Jun 19 2008, 11:59 AM']Would it be possible to have the reference manual either in HTML form or to include links within the PDF?  It's nice to be able to scroll through the list of function names and then jump straight to ones of interest.
[right][snapback]396285[/snapback][/right]
[/quote]

Additionally, could you please kill the "see also" sections? Searching for a function name always finds these, but not the function you're looking for.
[quote name='StickGuy' date='Jun 19 2008, 11:59 AM']Would it be possible to have the reference manual either in HTML form or to include links within the PDF?  It's nice to be able to scroll through the list of function names and then jump straight to ones of interest.

[snapback]396285[/snapback]






Additionally, could you please kill the "see also" sections? Searching for a function name always finds these, but not the function you're looking for.

#11
Posted 06/19/2008 06:50 PM   
Same problem here on Vista 64 + 8800GTX:

"The NVIDIA Setup program could not locate any drivers that are compatible with your current hardware."
Same problem here on Vista 64 + 8800GTX:



"The NVIDIA Setup program could not locate any drivers that are compatible with your current hardware."

#12
Posted 06/19/2008 09:23 PM   
Hey, have you noticed the NVIDIA's web still selects the 177.35 driver when you select the "download CUDA"? That driver does not work with GF8/9 series... only with the GTX ... any idea how to fix that, pls? Touching the .ini file or...?
Hey, have you noticed the NVIDIA's web still selects the 177.35 driver when you select the "download CUDA"? That driver does not work with GF8/9 series... only with the GTX ... any idea how to fix that, pls? Touching the .ini file or...?

#13
Posted 06/21/2008 01:15 AM   
That is not the first time nvidia makes real mess with official driver. Somebody should be fired for such mistake. I expect after spending over 2000$ for Tesla, 9800GX2 and 790i Ultra mobo, drivers for those expensive pieces should work without mods and tweaks. Unfortunately history repeats and it looks like nvidia doesn't feel responsible (again). Hey NGuys wtf you are doing?

This is the last time I do their job.
This is workaround for windows

1. Find display driver 175.70 (www.guru3d.com) and download it
2. Uninstal current Nvidia Display driver
3. start 175.70 selfextracting exe but DO NOT install that driver
4. find folder where exe extracts files and copy nv4_disp.inf file from that folder to some temp location
5. start installation of 177.35 driver
6. when setup fails (it reports error and terminate installation but all files are extracted usually in C:\NVIDIA)Winblablabla\177.35\english)
7. take nv4_disp.inf from temp location and paste (overwrite) existing nv4_disp.inf file in 177.35\english folder
8. start setup again
9. now it should recognize your hardware
10. reboot your comp

Followed steps are necessary only if you use nTune and have problems with GPU's fan(s) when windows boot (it is necessary only once)
11. when windows boots go to folder
C:\Documents and Settings\Administrator\Local Settings\Application Data\NVIDIA Corporation\nTune\Profiles
12. using wordpad open file osbootpf.nsu
13. find lines:
GPUFAN3D0=0
GPUFAN3D1=0 // this line exists only if you have more than one GPU
GPUFAN3D2=0 // this line exists only if you have more than two GPU

and replace zeros with percentage initial values. For example if you want GPU0 fan works on 60% when windows boot then line should looks like
GPUFAN3D0=60

14. Save changes
15. reboot


thats it

Note: if you have new Nvidia g2xx card you shouldn't do this because original nv4_disp.inf already has support for it
That is not the first time nvidia makes real mess with official driver. Somebody should be fired for such mistake. I expect after spending over 2000$ for Tesla, 9800GX2 and 790i Ultra mobo, drivers for those expensive pieces should work without mods and tweaks. Unfortunately history repeats and it looks like nvidia doesn't feel responsible (again). Hey NGuys wtf you are doing?



This is the last time I do their job.

This is workaround for windows



1. Find display driver 175.70 (www.guru3d.com) and download it

2. Uninstal current Nvidia Display driver

3. start 175.70 selfextracting exe but DO NOT install that driver

4. find folder where exe extracts files and copy nv4_disp.inf file from that folder to some temp location

5. start installation of 177.35 driver

6. when setup fails (it reports error and terminate installation but all files are extracted usually in C:\NVIDIA)Winblablabla\177.35\english)

7. take nv4_disp.inf from temp location and paste (overwrite) existing nv4_disp.inf file in 177.35\english folder

8. start setup again

9. now it should recognize your hardware

10. reboot your comp



Followed steps are necessary only if you use nTune and have problems with GPU's fan(s) when windows boot (it is necessary only once)

11. when windows boots go to folder

C:\Documents and Settings\Administrator\Local Settings\Application Data\NVIDIA Corporation\nTune\Profiles

12. using wordpad open file osbootpf.nsu

13. find lines:

GPUFAN3D0=0

GPUFAN3D1=0 // this line exists only if you have more than one GPU

GPUFAN3D2=0 // this line exists only if you have more than two GPU



and replace zeros with percentage initial values. For example if you want GPU0 fan works on 60% when windows boot then line should looks like

GPUFAN3D0=60



14. Save changes

15. reboot





thats it



Note: if you have new Nvidia g2xx card you shouldn't do this because original nv4_disp.inf already has support for it

#14
Posted 06/21/2008 04:27 AM   
Mandrak is completely right. This is a huge mess. This driver only works for a VERY small subset of CUDA compatible cards. (just do a diff between beta1-174.55 and beta2-177.35 inf files to see) I thought about doing the workaround, but I would much rather have a "supported" driver. NV guys, any idea when you will get this fixed?? Very few people can use beta2 in its current form on windows. And how did this ever get past QA?!

Thanks.
Mandrak is completely right. This is a huge mess. This driver only works for a VERY small subset of CUDA compatible cards. (just do a diff between beta1-174.55 and beta2-177.35 inf files to see) I thought about doing the workaround, but I would much rather have a "supported" driver. NV guys, any idea when you will get this fixed?? Very few people can use beta2 in its current form on windows. And how did this ever get past QA?!



Thanks.

#15
Posted 06/21/2008 04:09 PM   
  1 / 2    
Scroll To Top