SLI with different cards

[afraxass]

Thanks for your help ! I'll do everthing you wright but this weekend ! Ssoooooo tired and it's almost midnight here in france xD Thanks again for your help ^^

normally,
i can't copy patched nvlddmkm.sys to C:\windows\system32\DriverStore\FileRepository\~~~~
so i have to boot with safemode

cmd - msconfig - check safemode boot
after enter safemode, then copy patched driver "nvlddmkm.sys" to windows directory
if copy successfully,
cmd - msconfig - uncheck safemode boot

 

[tytynono]

..............................

Meanwhile Win10 users please use the version attached to this post
I've also included an updated version of Install.cmd

I've not "nv_dispi.inf_amd64_7209bde3180ef5f7" folder with Win 10 build 17017, my nvlddmkm.sys is in "nv_dispi.inf_amd64_d0a31ff5946203b6"
The patch work but no nvidia panel after reboot

I think the patched file nvlddmkm.sys is not patched correctly, see below

 

[Swatson]

I've not "nv_dispi.inf_amd64_7209bde3180ef5f7" folder with Win 10 build 17017, my nvlddmkm.sys is in "nv_dispi.inf_amd64_d0a31ff5946203b6"
The patch work but no nvidia panel after reboot



I think the patched file nvlddmkm.sys is not patched correctly, see below

You need to edit the install.cmd with your "nv_dispi.inf_amd64_d0a31ff5946203b6" instead of "nv_dispi.inf_amd64_7209bde3180ef5f7", then run the install.cmd in safe mode to sign it. This is why you see the old signatures.

 

[tytynono]

You need to edit the install.cmd with your "nv_dispi.inf_amd64_d0a31ff5946203b6" instead of "nv_dispi.inf_amd64_7209bde3180ef5f7", then run the install.cmd in safe mode to sign it. This is why you see the old signatures.

I have of course done that but my nvlddmkm.sys file is like that

 

[Swatson]

I have of course done that but my nvlddmkm.sys file is like that

did you run the "bcdedit" commands? the patch probably worked it's the signatures that are missing it seems.

 

[tytynono]

did you run the "bcdedit" commands? the patch probably worked it's the signatures that are missing it seems.

Yes

...

ok patcher done but...when I 'launch instal.cmd :

L’opération a réussi.
Le processus ne peut pas accéder au fichier car ce fichier est utilisé par un autre processus.
0 fichier(s) copié(s).
Reboot now for changes to take effect

sorry it's french : (

UPDATE : I rebooted ans still no SLI on the NVidia panel

Salut CGFlyer, il faut biensûr utiliser le fichier install.cmd mais il vaut mieux copier/coller ton fichier nvlddmkm.sys manuellement et modifier le nom de l'original en .back.
Pour cela il faut te rendre propriétaire du dossier ou se trouve nvlddmkm.sys, regarde la video jointe.

Hi CGFlyer, you must use the file install.cmd but it is better to copy and paste your file nvlddmkm.sys manually and change the name of the original in .back.
To do this you have to make the owner of the folder where nvlddmkm.sys is, look at the video attached.

 

[Arkanggel10]

387.92 working properly?

 

[Swatson]

387.92 working properly?

Works for me but control panel will not open after connecting to the internet. Yes I know that's odd but after re-reinstalling it I could still open control panel until I re-enabled my internet. I think Nvidia is getting smart....

Seriously, when I re-installed to check why it wouldn't open I disabled my internet to make sure windows didn't try to auto update the driver (it has interrupted me a few times), it actually still opened after enabling SLI. Once my internet was re-enabled control panel no longer opens....

Edit: Just gave it a try since you reminded me. Turned off my network adapter and reboot, I can open control panel now.... this is seriously odd. Can open it even after turning network adapter back on, but if I reboot again with network adapter turned on, it will stop opening again...

 

[Arkanggel10]

I lost nvidia icon some driver updates ago also...

I will try 387.92 and your solution...

Thanks

 

[CGFlyer]

Salut CGFlyer, il faut biensûr utiliser le fichier install.cmd mais il vaut mieux copier/coller ton fichier nvlddmkm.sys manuellement et modifier le nom de l'original en .back.
Pour cela il faut te rendre propriétaire du dossier ou se trouve nvlddmkm.sys, regarde la video jointe.

Hi CGFlyer, you must use the file install.cmd but it is better to copy and paste your file nvlddmkm.sys manually and change the name of the original in .back.
To do this you have to make the owner of the folder where nvlddmkm.sys is, look at the video attached.

YYYYEEEEEEEEE AAAAHHHHHHHH !!!!!!!!!!!!!!!!!

Thanks every one for hel"pink" me !!!! especialy @P!nkpanther and @tytynono

 

[P!nkpanther]

Great!
I noticed that you connected each of your monitors to a different card. While this might look more 'symmetrical', it puts a heavy load on the PCIe bus, lowering SLI scaling.
I recommend you to monitor your bus usage using a tool like nVidiaInspector, comparing bus load in the current configuration with that when both cards are connected on the same card.

 

[CGFlyer]

Comparing bus load in the current configuration with that when both cards are connected on the same card.

In fact when SLI is enable only 1 display is active, so I guess it's not that badd on the bus bandwith ?

 

[P!nkpanther]

If the secondary display is completely inactive, then it's no problem.
But even it displays your desktop only, while the primary display renders a game, bus load will be heavy, if different cards are used to connect these monitors.

 

[LightlyEngineered]

I'm going to give this another attempt tonight. Hopefully, I can make some progress.

 

[CGFlyer]

If the secondary display is completely inactive, then it's no problem.
But even it displays your desktop only, while the primary display renders a game, bus load will be heavy, if different cards are used to connect these monitors.

When in SLI mode my second display is completly black, not inactive but black. And in fact each GPU is handeled by a different CPU so I think bus load isnt' realy an issue XD

 

[tytynono]

"SignTool Error: No certificates were found that met all the given criteria." could you tell ma why tere is this error?

someone would like to help me?

 

[P!nkpanther]

When in SLI mode my second display is completly black, not inactive but black. And in fact each GPU is handeled by a different CPU so I think bus load isnt' realy an issue XD

Black or not, as soon as you have active connectors on multiple cards you have that increased bus load. Might just be the common desktop area, hidden behind that blacked screen.
But my point was, since it's so easy to check, why not test it instead of speculating.

And the GPUs are handled by the driver, so unless you wrote your own driver you have no control over which CPU handles which GPU. You're at the mercy of the driver devs how they optimized the handling to fit your respective hardware architecture.

"SignTool Error: No certificates were found that met all the given criteria." could you tell ma why tere is this error?

someone would like to help me?

You're in safe mode and ran install.cmd as admin?

 

[CGFlyer]

Black or not, as soon as you have active connectors on multiple cards you have that increased bus load. Might just be the common desktop area, hidden behind that blacked screen.
But my point was, since it's so easy to check, why not test it instead of speculating.

Wich parameter do I need to look at ? MCU ? VPU ?
Sorry if this sound stupid but not realy ban expert in this domain

 

[P!nkpanther]

If you're using nVidiaInspector, right-click on the sensor area and select 'Bus Usage'.
Other tools, like MSI Afterburner, have a similar measure available.

Remember that the percentage used by Bus Usage is lost for GPU processing, so even 10% of it means 10% less GPU calculating power.
Also, it's very much dependent on the application you use. For stressing bus load, I usually use 'Star Swarm Stress Test' in RTS Scenario (it's on Steam).

 

[CGFlyer]

If you're using nVidiaInspector, right-click on the sensor area and select 'Bus Usage'.
Other tools, like MSI Afterburner, have a similar measure available.

Remember that the percentage used by Bus Usage is lost for GPU processing, so even 10% of it means 10% less GPU calculating power.
Also, it's very much dependent on the application you use. For stressing bus load, I usually use 'Star Swarm Stress Test' in RTS Scenario (it's on Steam).

So...with Gulia4D stress test I've got around 17%Bus usage on 1 card and 27%on the other...is that normal ?
And when I try another benchmark,well...my PC just shut down !!! No blue/black screen, just off... this is wowering me right now...

This is wowering me right now...

Alright...find my problem...my PSU, even if it's an 1250W one, is only capable of delivering 18A per GPU so around 216W and...the 780 can consume up to 260W peak... I guess I need to not do bench mark anymore xD (or run a dual PSU setup...)

 

[LightlyEngineered]

If you're using nVidiaInspector, right-click on the sensor area and select 'Bus Usage'.
Other tools, like MSI Afterburner, have a similar measure available.

Remember that the percentage used by Bus Usage is lost for GPU processing, so even 10% of it means 10% less GPU calculating power.
Also, it's very much dependent on the application you use. For stressing bus load, I usually use 'Star Swarm Stress Test' in RTS Scenario (it's on Steam).

I'm totally stuck. I've re-tried all of the steps multiple times and I'm not sure what I am missing. Only once have I gotten the drivers to work properly but it didn't allow me to run SLI.

The patcher comes back with a successful, the driver installs in safe mode (but no certificate), I can reboot and everything appears to be working properly but there's no SLI option. I'd be willing to paypal someone if they could help me figure this out. I can repost all of my screen shots of all the steps if that would be helpful.

 

[P!nkpanther]

So...with Julia4D stress test I've got around 17%Bus usage on 1 card and 27%on the other...is that normal ?
And when I try another benchmark,well...my PC just shut down !!! No blue/black screen, just off... this is wowering me right now...

Some bus usage certainly is normal, this may vary widely, depending on the application.
What I was warning of, is EXCESS bus usage due to inter-videobuffer communication (figuratively speaking), when monitors are connected to different cards.
So you want to compare bus load under these different circumstances (both on card #1, both on card #2, each one on one card, etc.).

In your case of 17% on one card and 27% on the other, what comes to mind is that your second card seems to be seated in the x8 PCIe slot, while the first one is in a x16 slot. You might want to experiment with different slots for your second card.

According to this document, #2 and #5 are the x16 slots:
http://h20331.www2.hp.com/Hpsub/downloads/Optimizing_PCI-Express_on_Z800_3-18-09.pdf

The patcher comes back with a successful, the driver installs in safe mode (but no certificate), I can reboot and everything appears to be working properly but there's no SLI option.

From this summary I take it that your system still loads an unmodified driver.

Device Manager → Double click your main display adapter → Tab: Driver → Driver Details
Take note where the active nvlddmkm.sys is located.
Go to that location and open the file properties of nvlddmkm.sys→ Tab: Digital Signatures
An unmodified driver would have 3 certificates listed, one from MS the rest from nVidia.
Your modified driver however should list the certificate from DifferentSLIauto.

 

[afraxass]

I'm totally stuck. I've re-tried all of the steps multiple times and I'm not sure what I am missing. Only once have I gotten the drivers to work properly but it didn't allow me to run SLI.

The patcher comes back with a successful, the driver installs in safe mode (but no certificate), I can reboot and everything appears to be working properly but there's no SLI option. I'd be willing to paypal someone if they could help me figure this out. I can repost all of my screen shots of all the steps if that would be helpful.

so, strange

Assume nvidia driver installed succesfully in normal mode(windows 10)

anyway nvlddmkm.sys will be modified 2times

First Step.
if run DifferentSLIAuto.exe successfully, make .bak file.
this time, the size of patched nvlddmkm.sys is exactly same as nvlddmkm.bak (because of modifying some of string to "00" or "90")

Second Step
if run install.cmd, the size of nvlddmkm.sys is decreased ~kB

At the Second Step
Someone copied patched nvlddmkm.sys to driver's directory (windows/system32~~~) succesfully
but, my case it always denied it. so i have to reboot with Safe mode, and copy nvlddmkm.sys to windows dir. manually

please make sure, the size of nvlddmkm.sys and be careful driver DIR.

 

[John Johnson]

Installed 388.00 today and it works perfeoktly.

Can somebody kindly provide me an win7 driver file of 387.92
Discovered some crap in the os detection mechanism
Win7/8/8.1 users should use version 1.7.1 with 387.92 just fine

Meanwhile Win10 users please use the version attached to this post
I've also included an updated version of Install.cmd

Installed 388.00 today on Windows 10 Pro latest update 1709 Build 16299.19 using this edditon of DifferentSLI and it works perfectly.

Good luck.

 

[CGFlyer]

In your case of 17% on one card and 27% on the other, what comes to mind is that your second card seems to be seated in the x8 PCIe slot, while the first one is in a x16 slot. You might want to experiment with different slots for your second card.
According to this document, #2 and #5 are the x16 slots:
http://h20331.www2.hp.com/Hpsub/downloads/Optimizing_PCI-Express_on_Z800_3-18-09.pdf

Well, I checked and I'm on the two X16 slot...so weird...
But I have another problem :

as you can see I can't enable Suround mode. I've checked Nvidia Web'site and my screen are pluged in the right port... don't know what to do right now