GTX 560 Ti: Driver 301.24 Fixed PSOD, but at a Price Tons of System Interrupts
  1 / 2    
This driver fixed my PSOD issue, but now I have 1 core on my i7 pegged consistently at 50% (at least). I imagine any fix for this will have to be included in a driver update from Nvidia, so is there a workaround that doesn't involve rebooting?
This driver fixed my PSOD issue, but now I have 1 core on my i7 pegged consistently at 50% (at least). I imagine any fix for this will have to be included in a driver update from Nvidia, so is there a workaround that doesn't involve rebooting?

#1
Posted 05/04/2012 03:30 PM   
try disabling the C3/C6 states and see if the interrupts stop
try disabling the C3/C6 states and see if the interrupts stop



In Memory of Chris "ChrisRay" Arthington, 1982-2010

CPU:Intel i7 920 @ 3.8(D0), Mainboard:Asus Rampage II Gene, Memory:12GB Corsair Vengeance 1600
Video:EVGA Geforce GTX 680+ 4GB, Sound:Creative XFI Titanium Fatal1ty Pro, Monitor:BenQ G2400WD
HDD:500GB Spinpoint F3, 1TB WD Black, 2TB WD Red, 1TB WD Black
Case:NZXT Guardian 921RB, PSU:Corsair 620HX, OS:Windows 7 SP1

#2
Posted 05/04/2012 06:51 PM   
[quote name='Sora' date='04 May 2012 - 12:51 PM' timestamp='1336157506' post='1404342']
try disabling the C3/C6 states and see if the interrupts stop
[/quote]

Hi, thanks for the reply. Are you implying this could be a processor issue? If so, it's not a processor issue, and I'd rather keep my low power states available. This is 100% a product of the new driver. I didn't have System Interrupt errors before updating to 300 series, I do now. I rolled back the driver and I no longer had the issue. PSOD necessitated updating the driver to 300 series once more, System Interrupt errors are back.
[quote name='Sora' date='04 May 2012 - 12:51 PM' timestamp='1336157506' post='1404342']

try disabling the C3/C6 states and see if the interrupts stop





Hi, thanks for the reply. Are you implying this could be a processor issue? If so, it's not a processor issue, and I'd rather keep my low power states available. This is 100% a product of the new driver. I didn't have System Interrupt errors before updating to 300 series, I do now. I rolled back the driver and I no longer had the issue. PSOD necessitated updating the driver to 300 series once more, System Interrupt errors are back.

#3
Posted 05/04/2012 06:56 PM   
No, it probably is a driver issue, but removing C states from the equation can go towards debugging where the issue is and get a quick fix in the next set.

There are apparently a couple of issues that some users got in .24 that were not present in .25 (and remain to be not present in .34) so you could try that set as well by replacing the contents of nv_disp.inf in the .34 driver set attached with the pastebin inf dump attached

http://www.nvidia.com/object/win7-winvista-64bit-301.34-driver.html
http://pastebin.com/gH6YW5uM

also, could you post a screenshot of Latencymon's first tab, it will display the driver with the highest interrupt and dpc times
No, it probably is a driver issue, but removing C states from the equation can go towards debugging where the issue is and get a quick fix in the next set.



There are apparently a couple of issues that some users got in .24 that were not present in .25 (and remain to be not present in .34) so you could try that set as well by replacing the contents of nv_disp.inf in the .34 driver set attached with the pastebin inf dump attached



http://www.nvidia.com/object/win7-winvista-64bit-301.34-driver.html

http://pastebin.com/gH6YW5uM

also, could you post a screenshot of Latencymon's first tab, it will display the driver with the highest interrupt and dpc times



In Memory of Chris "ChrisRay" Arthington, 1982-2010

CPU:Intel i7 920 @ 3.8(D0), Mainboard:Asus Rampage II Gene, Memory:12GB Corsair Vengeance 1600
Video:EVGA Geforce GTX 680+ 4GB, Sound:Creative XFI Titanium Fatal1ty Pro, Monitor:BenQ G2400WD
HDD:500GB Spinpoint F3, 1TB WD Black, 2TB WD Red, 1TB WD Black
Case:NZXT Guardian 921RB, PSU:Corsair 620HX, OS:Windows 7 SP1

#4
Posted 05/04/2012 07:04 PM   
[quote name='Sora' date='04 May 2012 - 01:01 PM' timestamp='1336158116' post='1404349']
No, it probably is a driver issue, but removing C states from the equation can go towards debugging where the issue is and get a quick fix in the next set.

There are apparently a couple of issues that some users got in .24 that were not present in .25 (and remain to be not present in .34) so you could try that set as well by replacing the contents of nv_disp.inf in the .34 driver set attached with the pastebin inf dump attached

http://www.nvidia.com/object/win7-winvista-64bit-301.34-driver.html
http://pastebin.com/gH6YW5uM
[/quote]

Thanks!

I'm on .24 now. I was unaware that .34 (and above) was available for 500 series cards. The link you posted is for a 600 series card. Does that make a difference?
[quote name='Sora' date='04 May 2012 - 01:01 PM' timestamp='1336158116' post='1404349']

No, it probably is a driver issue, but removing C states from the equation can go towards debugging where the issue is and get a quick fix in the next set.



There are apparently a couple of issues that some users got in .24 that were not present in .25 (and remain to be not present in .34) so you could try that set as well by replacing the contents of nv_disp.inf in the .34 driver set attached with the pastebin inf dump attached



http://www.nvidia.com/object/win7-winvista-64bit-301.34-driver.html

http://pastebin.com/gH6YW5uM





Thanks!



I'm on .24 now. I was unaware that .34 (and above) was available for 500 series cards. The link you posted is for a 600 series card. Does that make a difference?

#5
Posted 05/04/2012 07:05 PM   
Tried running Latencymon. I got a cascading "invalid FLOP error" after running it for a minute or so.
Tried running Latencymon. I got a cascading "invalid FLOP error" after running it for a minute or so.

#6
Posted 05/04/2012 07:09 PM   
Technically, the driver is for the 690 and 670 as these are the only 2 cards the inf supports, However thats what the pastebin dump is for.
nVidia has maintained a UDA (Unified Driver Architecture) since the early Detonator days so any fixes checked in between .24 and .34 that may affect older cards will still be present

Driver 301.24 is r301_07-12 DriverVer = 04/03/2012, 8.17.13.0124
Driver 301.25 is r301_07-14 DriverVer = 04/04/2012, 8.17.13.0125
Driver 301.34 is r301_07-38 DriverVer = 04/27/2012, 8.17.13.0134

As it stands, i too was getting random interrupt spikes in the nvklddm driver under 301.25, but that is no longer the case with .34.

[quote name='JazzMac251' date='05 May 2012 - 06:09 AM' timestamp='1336158545' post='1404353']
Tried running Latencymon. I got a cascading "invalid FLOP error" after running it for a minute or so.
[/quote]

yeah, thats actually a really strange bug that only occurs when the processor is allowed to enter C3/C6 states.
I got it myself until i set the C state limit to C1(HALT) in the bios. Either the resplendence driver tries to do stuff with a sleeping core or its caused by the higher dpc and interrupts caused by waking a core to execute an instruction.
Technically, the driver is for the 690 and 670 as these are the only 2 cards the inf supports, However thats what the pastebin dump is for.

nVidia has maintained a UDA (Unified Driver Architecture) since the early Detonator days so any fixes checked in between .24 and .34 that may affect older cards will still be present



Driver 301.24 is r301_07-12 DriverVer = 04/03/2012, 8.17.13.0124

Driver 301.25 is r301_07-14 DriverVer = 04/04/2012, 8.17.13.0125

Driver 301.34 is r301_07-38 DriverVer = 04/27/2012, 8.17.13.0134



As it stands, i too was getting random interrupt spikes in the nvklddm driver under 301.25, but that is no longer the case with .34.



[quote name='JazzMac251' date='05 May 2012 - 06:09 AM' timestamp='1336158545' post='1404353']

Tried running Latencymon. I got a cascading "invalid FLOP error" after running it for a minute or so.





yeah, thats actually a really strange bug that only occurs when the processor is allowed to enter C3/C6 states.

I got it myself until i set the C state limit to C1(HALT) in the bios. Either the resplendence driver tries to do stuff with a sleeping core or its caused by the higher dpc and interrupts caused by waking a core to execute an instruction.



In Memory of Chris "ChrisRay" Arthington, 1982-2010

CPU:Intel i7 920 @ 3.8(D0), Mainboard:Asus Rampage II Gene, Memory:12GB Corsair Vengeance 1600
Video:EVGA Geforce GTX 680+ 4GB, Sound:Creative XFI Titanium Fatal1ty Pro, Monitor:BenQ G2400WD
HDD:500GB Spinpoint F3, 1TB WD Black, 2TB WD Red, 1TB WD Black
Case:NZXT Guardian 921RB, PSU:Corsair 620HX, OS:Windows 7 SP1

#7
Posted 05/04/2012 07:10 PM   
-del-
-del-



In Memory of Chris "ChrisRay" Arthington, 1982-2010

CPU:Intel i7 920 @ 3.8(D0), Mainboard:Asus Rampage II Gene, Memory:12GB Corsair Vengeance 1600
Video:EVGA Geforce GTX 680+ 4GB, Sound:Creative XFI Titanium Fatal1ty Pro, Monitor:BenQ G2400WD
HDD:500GB Spinpoint F3, 1TB WD Black, 2TB WD Red, 1TB WD Black
Case:NZXT Guardian 921RB, PSU:Corsair 620HX, OS:Windows 7 SP1

#8
Posted 05/04/2012 07:13 PM   
I know it's a little off topic, but will disabling those C states have any noticeable impact on my daily usage? I do try to keep my power consumption to a minimum when possible.
I know it's a little off topic, but will disabling those C states have any noticeable impact on my daily usage? I do try to keep my power consumption to a minimum when possible.

#9
Posted 05/04/2012 07:15 PM   
Also, how do I go about this pastepin procedure? Do I just download the driver, extract the files, do the paste job? Do I have to rezip? I guess not. I'm thinking back to my experience doing similar things on the Android platform...but I think the only reason rezipping was necessary was to make the .zip flashable.

So i guess it's just extract, paste, and run the installer?
Also, how do I go about this pastepin procedure? Do I just download the driver, extract the files, do the paste job? Do I have to rezip? I guess not. I'm thinking back to my experience doing similar things on the Android platform...but I think the only reason rezipping was necessary was to make the .zip flashable.



So i guess it's just extract, paste, and run the installer?

#10
Posted 05/04/2012 07:19 PM   
As far as i've see,n the impact of those C states is nothing special. C1 halt state disconnects the processor from the bus without completely powering it down which removes a fair mount of dpc and interrupt latency since the core does not need waking to resume operation. The OS can still park the core but it won't get the benefit of being completely disconnected from power. depending on your current processor it might actually be best to disable it anyway since there are hardware erratums in the first gen Core i processors supporting C3/C6.

I have never seen anyone go into the power usage benefits of the C states other then skimming over their existence in most reviews. For a home consumer i'd say the result would be similar to the <5c/y decrease that Green hd's provide.

[quote name='JazzMac251' date='05 May 2012 - 06:19 AM' timestamp='1336159165' post='1404362']
Also, how do I go about this pastepin procedure? Do I just download the driver, extract the files, do the paste job? Do I have to rezip? I guess not. I'm thinking back to my experience doing similar things on the Android platform...but I think the only reason rezipping was necessary was to make the .zip flashable.

So i guess it's just extract, paste, and run the installer?
[/quote]

yep, open up the inf file which usually defaults to c:\NVIDIA\DisplayDriver\301.34\WinVista_Win7_64\English\Display.Driver\nv_disp.inf (for the US/English release of 301.34 anyway) and then paste the replace the contents of nv_disp.inf with the contents of pastebin.

Then all you need is to run setup.exe @ c:\NVIDIA\DisplayDriver\301.34\WinVista_Win7_64\English\
As far as i've see,n the impact of those C states is nothing special. C1 halt state disconnects the processor from the bus without completely powering it down which removes a fair mount of dpc and interrupt latency since the core does not need waking to resume operation. The OS can still park the core but it won't get the benefit of being completely disconnected from power. depending on your current processor it might actually be best to disable it anyway since there are hardware erratums in the first gen Core i processors supporting C3/C6.



I have never seen anyone go into the power usage benefits of the C states other then skimming over their existence in most reviews. For a home consumer i'd say the result would be similar to the <5c/y decrease that Green hd's provide.



[quote name='JazzMac251' date='05 May 2012 - 06:19 AM' timestamp='1336159165' post='1404362']

Also, how do I go about this pastepin procedure? Do I just download the driver, extract the files, do the paste job? Do I have to rezip? I guess not. I'm thinking back to my experience doing similar things on the Android platform...but I think the only reason rezipping was necessary was to make the .zip flashable.



So i guess it's just extract, paste, and run the installer?





yep, open up the inf file which usually defaults to c:\NVIDIA\DisplayDriver\301.34\WinVista_Win7_64\English\Display.Driver\nv_disp.inf (for the US/English release of 301.34 anyway) and then paste the replace the contents of nv_disp.inf with the contents of pastebin.



Then all you need is to run setup.exe @ c:\NVIDIA\DisplayDriver\301.34\WinVista_Win7_64\English\



In Memory of Chris "ChrisRay" Arthington, 1982-2010

CPU:Intel i7 920 @ 3.8(D0), Mainboard:Asus Rampage II Gene, Memory:12GB Corsair Vengeance 1600
Video:EVGA Geforce GTX 680+ 4GB, Sound:Creative XFI Titanium Fatal1ty Pro, Monitor:BenQ G2400WD
HDD:500GB Spinpoint F3, 1TB WD Black, 2TB WD Red, 1TB WD Black
Case:NZXT Guardian 921RB, PSU:Corsair 620HX, OS:Windows 7 SP1

#11
Posted 05/04/2012 07:27 PM   
I noticed your driver (pastebin) is 64 bit. Mine is 32. Would the copy/paste cause any potential issues?
I noticed your driver (pastebin) is 64 bit. Mine is 32. Would the copy/paste cause any potential issues?

#12
Posted 05/05/2012 05:17 AM   
The installer seems to have noticed that one of its files was tampered with. After editing NV_Disp.inf, the installer fails.

"NVIDIA Installer Cannot Continue. Required files are missing."
The installer seems to have noticed that one of its files was tampered with. After editing NV_Disp.inf, the installer fails.



"NVIDIA Installer Cannot Continue. Required files are missing."

#13
Posted 05/05/2012 05:26 AM   
ah, totally my bad (well, i just assumed you had x64) yeah some of the file names are different so it will spit out an error like that

if you still have the .24 driver set unpacked all you need to do is copy the header of the .34 inf and replace the same section of the .24 inf, then save it over the top of .34 inf file and rerun setup.

it looks similar (but obviously says 32bit) to this
[code]; NVIDIA Windows Vista / Windows 7 (64 bit) Display INF file
; Copyright (c) NVIDIA Corporation. All rights reserved.

[Version]
Signature = "$Windows NT{:content:}quot;
Provider = %NVIDIA%
ClassGUID = {4D36E968-E325-11CE-BFC1-08002BE10318}
Class = Display
DriverVer = 04/27/2012, 8.17.13.0134
CatalogFile = NV_DISP.CAT[/code]

So just copy that part of the .34 inf into the .24 inf then save it over the top of the .34 inf and then it will install. (you can reunpack the driver to get the original file)
ah, totally my bad (well, i just assumed you had x64) yeah some of the file names are different so it will spit out an error like that



if you still have the .24 driver set unpacked all you need to do is copy the header of the .34 inf and replace the same section of the .24 inf, then save it over the top of .34 inf file and rerun setup.



it looks similar (but obviously says 32bit) to this

; NVIDIA Windows Vista / Windows 7 (64 bit) Display INF file

; Copyright (c) NVIDIA Corporation. All rights reserved.



[Version]

Signature = "$Windows NT{:content:}quot;

Provider = %NVIDIA%

ClassGUID = {4D36E968-E325-11CE-BFC1-08002BE10318}

Class = Display

DriverVer = 04/27/2012, 8.17.13.0134

CatalogFile = NV_DISP.CAT




So just copy that part of the .34 inf into the .24 inf then save it over the top of the .34 inf and then it will install. (you can reunpack the driver to get the original file)



In Memory of Chris "ChrisRay" Arthington, 1982-2010

CPU:Intel i7 920 @ 3.8(D0), Mainboard:Asus Rampage II Gene, Memory:12GB Corsair Vengeance 1600
Video:EVGA Geforce GTX 680+ 4GB, Sound:Creative XFI Titanium Fatal1ty Pro, Monitor:BenQ G2400WD
HDD:500GB Spinpoint F3, 1TB WD Black, 2TB WD Red, 1TB WD Black
Case:NZXT Guardian 921RB, PSU:Corsair 620HX, OS:Windows 7 SP1

#14
Posted 05/05/2012 12:44 PM   
[quote name='Sora' date='05 May 2012 - 06:44 AM' timestamp='1336221850' post='1404587']
ah, totally my bad (well, i just assumed you had x64) yeah some of the file names are different so it will spit out an error like that

if you still have the .24 driver set unpacked all you need to do is copy the header of the .34 inf and replace the same section of the .24 inf, then save it over the top of .34 inf file and rerun setup.

it looks similar (but obviously says 32bit) to this
[code]; NVIDIA Windows Vista / Windows 7 (64 bit) Display INF file
; Copyright (c) NVIDIA Corporation. All rights reserved.

[Version]
Signature = "$Windows NT{:content:}quot;
Provider = %NVIDIA%
ClassGUID = {4D36E968-E325-11CE-BFC1-08002BE10318}
Class = Display
DriverVer = 04/27/2012, 8.17.13.0134
CatalogFile = NV_DISP.CAT[/code]

So just copy that part of the .34 inf into the .24 inf then save it over the top of the .34 inf and then it will install. (you can reunpack the driver to get the original file)
[/quote]

Hey Sora.

So I think I went through the process correctly. I copied the 34 header over the top of the 24 header, then I replaced the 34 inf with the edited 24 contents via copy/paste. The driver installed, but core 3 is still going nuts with system interrupts. :(

not sure if the solution didn't work or I just biffed the process somehow. Just like before, upon boot into windows, system usage is normal for a while, but after a minute or two I start getting the spike in core 3 and it stays there for the remainder of my system's up-time.

I also checked into disabling c3/c6. My BIOS only offers the options of disabling "Intel C-State Tech" altogether, which I'm not sure is desirable, and disabling enhanced C1. Doesn't look like I can pick and choose, unfortunately. (Asus P6T SE)
[quote name='Sora' date='05 May 2012 - 06:44 AM' timestamp='1336221850' post='1404587']

ah, totally my bad (well, i just assumed you had x64) yeah some of the file names are different so it will spit out an error like that



if you still have the .24 driver set unpacked all you need to do is copy the header of the .34 inf and replace the same section of the .24 inf, then save it over the top of .34 inf file and rerun setup.



it looks similar (but obviously says 32bit) to this

; NVIDIA Windows Vista / Windows 7 (64 bit) Display INF file

; Copyright (c) NVIDIA Corporation. All rights reserved.



[Version]

Signature = "$Windows NT{:content:}quot;

Provider = %NVIDIA%

ClassGUID = {4D36E968-E325-11CE-BFC1-08002BE10318}

Class = Display

DriverVer = 04/27/2012, 8.17.13.0134

CatalogFile = NV_DISP.CAT




So just copy that part of the .34 inf into the .24 inf then save it over the top of the .34 inf and then it will install. (you can reunpack the driver to get the original file)





Hey Sora.



So I think I went through the process correctly. I copied the 34 header over the top of the 24 header, then I replaced the 34 inf with the edited 24 contents via copy/paste. The driver installed, but core 3 is still going nuts with system interrupts. :(



not sure if the solution didn't work or I just biffed the process somehow. Just like before, upon boot into windows, system usage is normal for a while, but after a minute or two I start getting the spike in core 3 and it stays there for the remainder of my system's up-time.



I also checked into disabling c3/c6. My BIOS only offers the options of disabling "Intel C-State Tech" altogether, which I'm not sure is desirable, and disabling enhanced C1. Doesn't look like I can pick and choose, unfortunately. (Asus P6T SE)

#15
Posted 05/05/2012 08:26 PM   
  1 / 2    
Scroll To Top