Upgrading the DGFF GPUs in the Precision 7530 & 7730

Turning off driver signing is fine for initial testing but I do recommend self-signing for the long-term, it will save you a lot of hassle when it comes to updating the driver and so forth.

Can you share what hardware ID is showing up for the card in your system and the specific INF alterations that you made?
I suspect that you just didn't make quite the correct tweak to the INF file.

I also have a personalized script for

 

So I added these lines, based on the info on page 1 of this thread :-

%NVIDIA_DEV.1F36.0832.1028% = Section051, PCI\VEN_10DE&DEV_1F36&SUBSYS_08321028
%NVIDIA_DEV.1F36.1832.1028% = Section059, PCI\VEN_10DE&DEV_1F36&SUBSYS_18321028

And then added :-

NVIDIA_DEV.1F36.0832.1028 = "NVIDIA Quadro RTX 3000"
NVIDIA_DEV.1F36.1832.1028 = "NVIDIA Quadro RTX 3000"

As for device ID's I used 0832 / 1832 based on gpuz :-



Thanks for your help

 

Let's confirm that the device ID is exactly right...
You can get it from Device Manager. Open up information on the GPU, go to the "Details" tab, and then select "Hardware IDs" from the list.

 

Thanks for your quick reply

 

This is what I am going to recommend.

Rather than adding lines, let's do "find-and-replace" — some versions of the driver have multiple "lists" of sections for different versions of Windows 10 and this will make sure that all necessary occurrences are replaced.

You only need to modify the device ID in the INF file, the bit that starts with "PCI\VEN...".
The bits that start with "NVIDIA_DEV" are basically variable names, and it does not matter if they do not match the hardware ID, it just matters that they match at the top and bottom parts of the file.

Since you have a Precision 7730, the closest match would be a Precision 7740, so we'll take the Precision 7740 RTX 3000 entries and adjust them to match the Precision 7730 hardware ID.

So, just do these find-and-replace's on a clean version of the file:
PCI\VEN_10DE&DEV_1F36&SUBSYS_19271028 → PCI\VEN_10DE&DEV_1F36&SUBSYS_18321028
PCI\VEN_10DE&DEV_1F36&SUBSYS_09271028 → PCI\VEN_10DE&DEV_1F36&SUBSYS_08321028

...And that's it.

(Using this method, we also don't have to do any work to figure out which "Section" number to use. The section numbers in my INF file do not match yours. I am looking at version 461.40 from NVIDIA's site, DCH version.)

 

Argh......

I hit the same problem, nvidia installer fails and update driver via device manager fails like in #319

When I go to Nvidia site I get a slightly newer driver 461.72

461.72-quadro-rtx-desktop-notebook-win10-64bit-international-dch-whql

Update, i have not tried changing the full string to read :-

%NVIDIA_DEV.1F36.0832.1028% = Section136, PCI\VEN_10DE&DEV_1F36&SUBSYS_08321028

%NVIDIA_DEV.1F36.1832.1028% = Section140, PCI\VEN_10DE&DEV_1F36&SUBSYS_18321028

Still the same error

 

Just for kicks, try this file. Extract and overwrite the nvdmwi.inf in your extracted package (461.40 DCH), and then try running setup.exe.
If it doesn't work, honestly stumped, that's the only step that I ever have to take to get it to work.

 

Just realized how close the RTX4000 actually is to the RTX5000, i have almost the same score with the RTX4000 using the 90W vBios :O

 

shrug

The two GPUs as configured in the 7740/7750 are also very close. There just isn't enough power to really let the RTX 5000's 3072 cores put out as much performance as they should. And if there was, this GPU would be in a desktop and rated at 250 W, and not a laptop... At least the 16 GB GDDR6 VRAM means that I can crank up the textures in a game in 1440p and still have no stuttering.

 

I use a 330W brick with my 7540 when plugged in, just to make sure that it never has any issues with not enough power and unstable voltage due to PSU doing overtime.

Yeah, then there is also the thermal problem, even at 90W limit while it can be cooled well when playing games, this is not the case during benchmarks.
The problem is that they share a heatpipe and the heat gets transfered from one to the other, that's all fine and dandy untill you hit them both with full load at the same time, the heatsink cant handle that 200W of heat and the CPU/GPU have to lower boosts closer to stock to prevent overheating.