[JenBell]

Hello People,

While installing Vista32 I came across the following error:

nvlddmkm.sys has stopped responding and has recovered

I looked around the net and found that it was a common problem. I managed to fix the same problem on both my home PCs and my clients by doing the same thing. It might not work for you but totally cured mine...actually I can replicate the error / crash on demand. Its a software issue and not a driver issue - fact (as u will see).

Method

Note: I am just explaining it as simply as I can. U can modify the solution so it is more precise, if you wish. I just wanted to fix the problem with the minimum of trouble.

1. UAC was turned on. ZoneAlarm was running (FW/AV)
2. First thing, download DriverCleaner. More recent the revision the better. Install this app.
3. Remove the Nvidia display drivers. My mobo is an Asus Extreme Striker so I only removed the display drivers.
4. Restart the PC into SM (safe mode) - (F8 at boot)
5. Login in while into SM
6. Run DriverCleaner. Choose the Nvidia option only. Only option I used but you can try all of them if you feel frisky
7. Empty RecycleBin (I set my RB to delete perm all the time so I don’t usually do the empty thing…)
8. Reboot into normal mode Vista
9. Using Explorer or open My Computer and browse C:\Windows\
10. Right-Click on System32 folder and click ‘search’
11. Enter NV*.sys
12. This will show you all the Nvidia system files in the folder. The ones we are interested in is the ‘nvlddmkm.sys’ and how many copies we have.
13. If you tried multiple install/uninstall you will a few copies of it. I had 19 copies. If you look at the location, they were all in the ‘FileRepository’ folder under different subfolder. Something like nv_as***, nv_io***, etc
14. Note down the name of the sub folders where the file is located, nv_as***
15. So we now know that we have X copies of the nvlddmkm.sys file and what folders they are located in under the FileRep folder. There might be one under the main system32 folder as well. Note it down.
16. Vista has a totally awesome function. ACLs for network geeks…Access-Control Lists or Access-Lists or whatever you want to call them
17. Browse to Accessories from the Start menu (win icon>all programs)
18. Right-Click on command and click ‘Run as Admin’
19. There are other ways 2 do the step above…I just got it pinned to my start menu or u can just press the win key on ur keyboard and type command
20. OK so in our Admin CMD window we type the following commands in sequence
21. takeown /f C:\Windows\System32\DriverStore\FileRepository\*.*
22. The command above will give you ownership of all the files under the FileRep folder
23. cacls C:\Windows\System32\DriverStore\FileRepository\*.* /G Username:F
24. Replace username with your windows account name or login name
25. Like magic we are done. Oh and if ur wondering…u cant do the above via GUI. Windows will give you permission but you would still not be able to delete anything 
26. Earlier we did the search for the nvlddmkm.sys under the system32 folder. Do that again. In the search results, select all the instances of the nvlddmkm.sys file and delete them. OK, next is a totally optional step but I did it to make 100% sure it was all gone.
27. We noted down all the nv_as**, etc folder which had the nvlddmkm.sys file in it, delete the folders and empty ur RecBin
28. Fire up your ForceWare drivers for the install. Disabled ur AV before hand, it does make a difference (git)
29. Restart, login, set ur resolution and enjoy the whole error-free nvlddmkm.sys effect.
30. Few notes. Make sure u can see the file extensions and system files are not hidden (blasphemous behaviour for techGeeks). Google search to show hidden window files and extension if u have no idea what to do

The whole procedure might sound like a really long way to get the problem sorted out buts its not. When u actually do it u will see that in totally in takes around 10 mins tops to get the whole thing sorted out.

The techguys had this problem while rolling our vista to the execs. One of the execs asked me to look into it. I declined as I was too busy until they added a contract addendum (read bonus…£££s (^-^)…) and I took a look at the whole problem. Spent an hour trying various things. I did the most logical thing…make Vista like XP…as much as I can…the takeown/cacls commands do this for us.

Piece of advice, I don’t recommend u apply the ACL commands to your whole storage systems. I usually enter it for the FileRep folder and ALL my HDDs that used for only data storage, like avi’s, etc.

I am sure someone on the net will create a pic version of this guide. As of right now all the first 1000 hits from Google on the 11/07/2007 11pm…no-one has this method written but should see it pop up on more sites.

Well. First of all. Anyone that knocks Vista32 needs to be shot. MS did a good job with the OS. Nvidia should not be blamed for this problem or MS. For me it’s a security issue that is in a grey area.

One final thing. I am not a Vista32 expert by any means.

Feel free to try it out let me know how u get on.

ScreamMyName?

PS: Just finished a piece of work. To run the normal install of Vista32 entirely ‘virtualised’ with shadow copies. This means that when ANY service or app crashes you can recover from it from a few simple key presses and carry on working like nothing happened. This is not VMware or anything like that. There are runtimes in V32 that allow u to run effectively 2+ copies of the whole OS and flip between them. For example if the RPC service fails…BSOD follows or a screen-freeze crash. With this ‘tweak’ the RPC shadow takes over because fundamentally…the applications hand-off code is running in separate memory spaces (-_-). Might write this thing up but will wait as my client saw the tweaked version and offered a lot of £££s to install it on all the there systems. The thing about the whole tweak…resource usage is only 3% higher than not running it…

EDIT: Found out that if you do all the WindowsUpdates do not get installed / failed then the NVL error will appear. The method above will not fix the problem. You will need to use the expand method. Basically in the System32\drivers folder you need to remove the NVL file and manually copy across the new NVL, that are part of your drivers. This seems to sort the isssue out.
Some people are reinstalling there whole PCs, etc and then post that there problems are resolved regarding the issue. Its because the updates were installed completely with no failures, etc.
As not everyone can be asked to keep re-installing, I am trying to narrow down the exact order of updates required as a minimum to avoid the problem. Later on in the week I plan to uninstall all the AutoUpdates and do them one by one until the NVL appears.

People....post your details of install/update methods used. I am sure between us we can sort this out and find a work-around until NV/MS get there heads together and sort it out properly.

Update 03/11/2007: PDF of how I upgraded from .44 to the latest beta's with pics...enjoy!

http://www.megaupload.com/?d=3K898TO3
http://rapidshare.com/files/67066628...ummon.pdf.html

Update: 13/Dec/2007: I think the whole single dimm stick fix is totally bogus or not exactly correct. I am running 2x 2GB RAM sticks on the lastest beta's released yesterday and its all working fine. I tried the following sticks: Corsair 2x2GB PC6400, Corsair 2GB PC8888 and Crucial 2x2GB and all checked out fine on vista32.

[Neoclasiccl]

This fix didn't work for me. I removed all the nvlddmkm.sys and installed the nvidia driver and the problem persists.

System Quad core intel, twin 8800gtx's 4 gigs ram. Xfi elite pro 680i EVGA motherboard.

Hopefully this problem gets resolved with a new driver.

[Surfrider77]

I have had this problem since JANUARY with both SLI 7900GTX and also with a single 8800 Ultra.

75 pages long thead on the nVidia forums and this is still not resolved for me.

[JenBell]

Neo...carry on hoping...its a problem that is more hardware related than anything when it comes to NVs side and as for MS, there drivers comm path is very strict but here is the thing. For example, when the NVL file wants to do something it goes through the usual procedures...which is fine. When there are 2 readers (2 RAM sticks) then u need the whole thing becomes a issue. We even tested this out. For the BFG 680i mobo's that some the artists are using we hired a proper pro. I hired him personally as I has worked with this guy before and know exactly how good he is. He showed so many factors that could cause the problem and also mentioned so many different ways to minimize the effects. He will be delivering his final report on Tuesday so hopefully I will update this thread.

Sur...OK...get this through ur head...this problem is not related to ur 2 GPUs. This is a vista specific / 680 platform problem. Drivers, software, vista driver structure and sub-level timings, etc...simplest way to get rid of the problem is to run one single RAM stick...I kid u not. I ram stick and u will see the issue vanish almost instantly!

Few things to note. Google search for the 'expand' method. This involves extracting the NVL file again and copying across manually in safe mode to the right folders. Also there is another method which is basically a dirty install. Install different drivers version over each other. Some people have had success with these methods.

Personally on my EVGA 680i A1 which had 2x1GB Corsair PC8888 ram is now running 1 dimm of Corsair 2Gb PC6400 @ stock 800. I just wanted the whole PC stable for now.

Having searched all over the place and spent a number of hours with a hardware/software master-level-yoda-spewing-geek-of-geeks tech superman I learnt a few things. One...he is willing to pay me £15k for my virtual method which he says is something that is not a tweak but a software in itself -_-
two...he made the companies BFG systems with 2x8800GTXs, QX6700, 8GB Corsair dimms all totally rock stable. I ran the usual benchies and all passed the 12hr mark fine and since then had to repeats on any of the PCs. Now here is the thing...talking to this uber-tech...he says that technically we are breaking our support contract with NVL...who at this present moment in time...still cannot offer us a damn solution....nvidiots. Also....MS wont be happy if they found out what we have done. The MGMT in the company have pretty much left the whole thing up to me. As the audit dept/sec/IT, etc all come under my logical control...I am already having sleepless nights thinking of how I am going to get past all these barriers and still keep everyone happy.

Speaking as a normal personal/home user...I am so not happy. I initially thought that NVL and MS had to come together to play but hearing a person who actually showed us a complete solution that actually works just gets to me. I like so many other people worked really hard and paid alot of money for some of the 'best' hardware available. I really feel let down but the whole thing. Ah....now here is the thing...a uber-techied fixed it...great...but I am sure MS and NV have the very people that created the damn thing...so why cant they go one better and sort it out once and for all?

JenBel

PS: My current client is so not happy with me because I barred them from releasing our solution to any of the partners, suppliers...basically anyone that was not internal to the company. Its simple...I personally hired the techie through my own company and such he is contracted to me. As per the contract, if the company discloses and information owned / created or 'custodian-ed' to my company then they are in breach of the contract. If a CISSP reads this...come one u must be smiling right now right? See...I practice what I am learning :-) oh and people...I kid u not when I say never trust senior mgmt...3 of them tried to email our initial report to there friends in other companies...thank god I designed the email system and have 'control' over it (^-^)...

[Surfrider77]

Originally Posted by JenBell

Sur...OK...get this through ur head...this problem is not related to ur 2 GPUs. This is a vista specific / 680 platform problem. Drivers, software, vista driver structure and sub-level timings, etc...simplest way to get rid of the problem is to run one single RAM stick...I kid u not. I ram stick and u will see the issue vanish almost instantly!

Read my post. I have the issue when using 7900GTXs in SLI AND when using a single 8800 Ultra. The thread on nVidia forums makes it perfectly clear it tracks across 6xxx, 7xxx, and 8xxx series cards. No **** its a software issue and yes people have posted some limited success by severely underclocking RAM, removing all but one stick, etc. But this is a B.S. fix. These systems run memtest flawlessly and the same hardware works flawless in WinXP.

Trust me, you dont need to lecture me. I've been at this for 7 months now and have read every claimed "fix", in which nothing seems to work.

I do appreciate the effort, honestly. I will take anything to get Vista and my system running at this point. The vast majority of your suggestion was posted months back in the middle of the 75 page long thread. Works for some, but not for most. Some that said it fixed it ended up back a week later with the exact same problem.

Ooops... thread is up to 81 pages now. If this isnt **** poor engineering / support, I honestly dont know what else is.
http://forums.nvidia.com/index.php?showtopic=25381

[JustChill]

I can't even begin to explain what kind of frustration I have went through with this god forsaken problem. I have truly spent more time with my system, "fixing" this problem than anything else. Obv. to no avail.

Jen Im not sure I get all this, does this fix it or do i still need to go as far as to rmove ram sticks from my system to be truly stable....ha thats a funny statement "truly stable" anyway can I get a quick rundown to what i known of this problem and what it effects, Ex. 680i? 8800? drivers? vista? and do any of you know if you were to build a system is there anyway to avoid this problem. like, is there any configuration of hardware that are not affected by this problem?

[JenBell]

Just...quick answer is that even the P35 mobo's or for that matter any hardware combo with vista can through up this error. As for the stable fix...tried on the following mobo's and it cured the problem...P35 DS6, P5K, Striker Extreme, EVGA, XFX...those are all the ones tested and fixed at the moment. Got tons of paper work to get through before Mondays meeting to hammer this all out with my client. Yes...the one dimm solution is the only current fix that cures the problem instantly...give a go if u dont believe me. I know its not the best solution.

U talk about frustration...I can understand that totally. Hopefully we can all work together to create our own solution for ourselves.

Edit: Just a reminder...2GB of RAM is the Vista sweet spot. I have a single 2GB dimm so I stuck that in and the error went away and its been ROCK stable since...I was encoding 2 instances of H264 media while surfing the net and watching some Kdramas and the PC was still kicking along fine. Lost Planet DX10 also played fine and Halo2 was fine as well. I have read the NV thread like 5 times. If I am selectively ignorant then I am thinking only about the fix I get for my client and as a bonus my own PC. I dont intend to OC the PC like crazy as I value stability that little bit more until the contract is over. For everyone OCing..my advice to avoid all the headaches is to buy a single OCing 2GB RAM dimm and stick that in. After that u can pretty much OC to your hearts content with upto the 2x 8800Ultra's in SLI without the dreaded NVL error popping up. If you need/use vista64 then I cant really offer u proper solution that works instantly. If there are any 'coders' out there...I recommend you look into the way vista sends/rec's data from RAM ;-). Spoke to NV person today who is still feeding me crap like...we are looking into it, etc...MS are just gits at the moment...here is something u might find interesting...with our SSstorage HDDs testbed...threw one these HDDs inside the a PC that was getting the NVL and installed Vista32 and strangely the problem is not coming up...even when we try to induce it...have effectively hardware some settings to make it fall over! There is a potential fix I am help work with which is much more simple and I could release this to everyone...but just dont have the time to hammer it all out so will update this thread as I go.

[Delsana]

Actually that's not true.

This is not a 680i platform issue.

Though most people with this error get the problem more often then myself it is indeed not specifically only for 680i.

When playing games after about 6 - 7 hours of play my monitor turns off without losing power and the screen is gone, and I have to restart just to get it back.

Well today I played WiC beta and it took only about 45 minutes - 1 hour for the monitor to stop showing anything but then it reverted after I alt tabbed for once and gave me this error everyone has.

I have a nForce590 not a 680i board.

Geforce 8800.

This is a nvidia issue plain and simple...

I tried following the advice, unfortunately DriverCleaner is not free anymore.

Any other ways?

[JenBell]

Delsana...I presume u are using vista right? Its strange u seem to hell bend on thinking this is a NV-only issue. Maybe you should look into the way vista actually works and the infrastructure in place. Maybe u missed my earlier posts but last time I checked...P35 remained an Intel chipset. I am concentrating more on 680i chipset as thats the mobo I am using and my client are using. As for drivercleaner...enter the right search in google and u can get 'a copy'.

Think of like this, if it was purely a driver issue then I am sure teams like extreme-g would be able to stream the drivers and get rid of the problem.

One thing to bear in mind about driver cleaner...it is effectively a script that deletes files. Next time I get time to play with the drivers (this weekend) I will update the first post with a manual method of removing the files the DC does.

Another thing to note...if it is a driver issue then can you tell me if NV also install memory management for the mobo/RAM? I do not think so...MS are keeping vista really close to them until SP1 is out and SP2 beckons...like XP.

Speaking to few people via IRC, it seems some people have tried combinations of things and not fixed the issue.

To everyone else: I am trying to help us fix the problem. Please help me fix the issue and not go off on who to blame, etc...that never helps...we can do that after we have a workable solution...right?

People...help me please. Tell me things u have tried other than the normal methods listed here and on the net.

[Delsana]

Well remember that my case was a bit different, most people had this error quickly and then it would just say resolved. Mine wouldn't say resolved it would shut down the display.

I managed to do everything you said but... I have no idea if the display error is gone now, I will see what happens. Also I had a BSOD on attempt to restart the computer after I followed advice, it would not go away and it would not stay soon enough for me to read it... so I had to repair the computer using the Vista Disk.

Now I have a bluetooth HID Device: Access is Denied. Which I also had for the graphics card until I manually installed the drivers. Now I can't use my keyboard and no solutions to fix it have helped.

I'm using an old one that you couldn't use for gaming if you even wanted to.

[SuperDave1685]

Hi there JenBell...

1)

This is a vista specific / 680 platform problem.

This is NOT true. I had this issue on my nForce 4 based DFI board. People have also reported having this issue on P965 chipsets as well. Yes, its a Vista specific issue, but it not tied to the 680i platform exclusively.

2)

.enter the right search in google and u can get 'a copy'.

Getting software via "alternative" means is prohibited by forum rules.

The way in which the nvidia driver interacts with Vista's kernel is *very* different than that of XP. MS has made numerous changes in the way Vista handles I/O requests on the driver level. There is a good explanation over at Creative's website about this new driver model. In XP, writing drivers was much easier because of the simpler API's used. Now with Vista, it has thrown hardware manufacturers for kind of a loop because of the new driver requirements. This is NOT a MS issue. I refuse to believe that. Look at all these other hardware companies who have successfully written drivers for their hardware for Vista. Then look at nVdia. The finger squarely points at them. I am lead to believe this issue stems from how the driver is trying to access the card's memory buffers for a read/write command, i.e. the way the driver communicates I/O requests to and from the card. I have read MANY of those treads you've mentioned and through my own experience with vista, I strongly believe it has to do with memory I/O's. The foundation for nvidia's drivers is flawed somewhere. I'm sure they are working on it in their labs, but have not been able to isolate the issue exactly. I also partially blame MS for their kernel design.

-Dave

[JenBell]

Supes...I take back the 680i only thing.

Hence the reason why one dimm makes the problem vanish.

Anyways, thank u for the input...saves me typing it!

[Surfrider77]

And again, everyone is back at square one... waiting on nVidia... hence the 80+ pages of gripes on the nVidia driver forum. Another day goes by making nearly 7 months in total that my "certified for Vista" vid card is useless in Vista.

Am a ticked off? You bet I am!

[JustChill]

k well bitching does not resolve anything. SO lets try to keep the thread more informative and less, well ..... opinionated!!! ....MMM-Kayyy . . . Thanx Peter.

[Section8]

I have a fix.....run Win XP I have a copy of Vista sitting here that I bought and ran for several months but got tired of banging my head against the wall so I loaded Win XP Pro back on my machine and it has been a sweet dream ever since. I run every game, application and software you can think of with no issues. When Service Pack 1 comes out for Vista I may load it on one of my other machines and give it a go again but until then Viva Win XP

[JenBell]

Can we please keep the 'XP' responses to a minimum. Yes XP is the current main OS but I had problems back then with XP and before switching to vista I still did. Vista is more stable. Apart from the NVL error...I would go vista every single time. Yes there are things (way too many things) that totally bug me but I am willing to use workarounds to enjoy a 'stable-in-my-eyes' OS.

Some interesting things...if someone has time...try googling for registry edits related to GPU timings. I am closing the contract with my current client next week as I simply hate having to be a total control freak. Told them that I was buying the rest of the contract out and will concentrate on working with companies who dont try to steal your tech and staff.

Rant: Can you believe it. These gits tried to poach my tech-guru who flat our refused them because he's happy where he is (does what he wants as long as he delivers results and looks presentable when in front of clients - oh and gets paid well) and then they tried my network admin who just stuck up 2 fingers (he totally hates them) before finally trying to poach my baby sis who just laughed at them.

Update: In my mini lab test I used a P35 DQS6, ExtremeStriker and an eVGA 680i mobo. My method was simple...cycle through my driver collection for the 8800ultra and see what happens. Results were kinda of strange. The DQS6 mobo was happy with the 158.42 for a little while but totally NVL-ErrorFree with 158.24. StrikerExtreme only liked the 162.15 and 158.24, for the .24 I had to use the expand method in safe mode to make it all play nice...after this it was pretty kewl. eVGA was ONLY happy with 158.24. It just did not like 162.22...tested out the 'removed' 163.11 which strangely did not give me the NVL error at all. Must remember though that those drivers have a HUGE bug inside so there is a good reason they must of removed it. I tried several flavors
of Extreme-G drivers as well...none of them were close to stable on any of the mobo's.

I did all my test using my rig below...changing the mobo but keeping everything else the same...inc 1 dimm.

I am currently using the 158.24 drivers which are rock solid for me. If you are still being killed by the NVL then I think u should do what I did and simply download all the drivers released by nvidiots so far. Install the drivers cleanly and give each one a go. Eventually u might hit one that works fine.

For me, my drivers means I cant play LostPlanet but then again I can play Halo2 fine. CSS just screams along nicely:-). Currently my rig is stable but when the latest and greatest games come out I will have to upgrade the drivers and I know for sure that if the NVL issue is still there then I will be left...so not happy. I mean imagine not being able to play crysis because of ur gpu drivers. Cant play LP but I dont mind that...If it was Crysis I would of killed myself.

I've decided...I am gonna wait for AMD to get Barcelona out and then see if the problem is there as well...I just think that it will not be there and my whole BarcaCPU life will be NVL-free.

I am so sick of the drivers/vista/everyone-NVL-error problem. If anyone stumbles upon a different method of fixing things, feel free to let us know. I for one have started looking forward to the BarcaCPU from AMD. When I was AMD I had problems but most of them were easily fixed and driver wise pretty kool. Let's put it this way...if Barca comes out and does not have the NVL error what would u do if u had the money? Carrying on using a 8800ultra and NOT be able to play ur games or get a slower CPU and play everything?

[JustChill]

are we the only ones wit this problem? I thougght that this would be a hot topic.............

[JenBell]

Just...I think most people are just giving up if they have the problem. Some people are willing to actually find a fix. Been contacted by a well-known blogging site...they want me to do a full-write-up on my whole experience but I currently dont have the time. Last week with my current client before I leave this not-so-nice place/people. I am still searching for the grail-of-hole-li-ness...latest drivers without NVLs killing my last remaining braincell :-|

Just wanted to let people know that there are some vista beta mini-service-pack-update-like things flying around. I am will try and get hold of these and try them out...u never know...they might actually fix the problem as they fix memory corruption problems...which is very interesting.

[Surfrider77]

Unfortunately, I think JenBell is correct. This has been such an ongoing problem that I know of people who have gone so far as changing hardware to get rid of the problem. Some have even given in to underclocking RAM in order to achieve stability.

My hat is off to you JB for being so persistent. Unfortunately, I think its going to take a miracle patch on nVidia and/or Vista's behalf before we see a fix.

[JenBell]

Surf...ur most likely right and nvidiots gits and MShte's will sort it out but not before nvidia release another mobo revision that magically fixes the problem...even better vista sp1 and a new nv mobo....ooooo....feel like strangling someone...gonna start with myself....who's after me?

The beta patches dont fix the problem by the way. Under clocking ram...does work for some but I tried that as well and took the ram all way down to 200 and it still gave me the error when running 2 sticks.

Oh and I would wait for the beta patches to come through...uninstalled them as I would rather wait for the real thing from MS....I have enough dealing with NVL issues and not being able to play LP.

People...please help us with ur thoughts...pretty please with a cherry on top...some whipped cream and caramel syrup (^-^)...sry...just talking about my icecream :-|