Not an option, I simply don't have it anymore.
The thing is that before getting rid of the old one, I tested the 4090 to be sure it work, but sadly, I only tried rendering, at the moment it simply didn't occur to me that the dforce could not work, so like a nice simpleton friend I offered it to a good friend of mine thinking everything was ok... and it wasn't...
Not a simpleton, just a nice friend.
Incidentally, it depends on the space in your computer, power source and the manufacturer of the 4090 (and eventual 3080) you will get (in terms of units, there is the NVIDIA, and then the derived ones from other manufacturers), but if you end up deciding to get an older card (but careful, I saw a post about dForce problems with a 3080 Ti as well), you can always try to use the two cards in parallel.
It is not the NVlink kind of performance, but NVIDIA allows that - though I admit I know it can work well with CUDA, not sure about OpenCL (well, OpenCL itself yes, at least with some scientific libraries, but the OpenCL implementation by NVIDIA as used by DAZ, I don't know enough and never tried mysefl)
Suggestions that maybe you tried already, but one never knows...:
1) "Rollback" to a previous version of the drivers, with NVIDIA means uninstall and install the old version;
2) If you have a problem getting the old version of the driver, try to download the CUDA development envirnoment, and use that installer, CUDA installers come also with the driver, but it is in general a bit older, e.g. on the NVIDIA main download you get 522.30, CUDA 11.8 comes with 522.06 (and CUDA 11.8 does support 4090, so, it must work)
For point 2, you do not need even to install the full CUDA development environment, if you chose the "custom" install, you can select and deselect elements, and if I remember well, one of the elements is the drivers.