Probably possible to do.

Then again, if they are the same GPU, why not just edit the .inf file for the newer drivers?

Much easier than modding a BIOS and the card stays stock.

Not like there are games that are going to check the PCI ID or try to read what the card is from the card's BIOS.

Doable, the only difference between cards could be memory timings and speeds. Make sure there is a fallback solution, such as a flash programmer with a backup bios or blind booting (or with a pci video card)

If software expects a GT 710 or better, it probably wouldn't like the Fermi version of the GT 730 anyway. It would probably need the Kepler version.

What’s sense of this ?
Software does anyway check for feature set, not for video bios string or id

So after going crosseyed poring over spec lists, and looking up old reviews, I can't actually find a top tier* graphics card manufacturer model that used the A108 96 core chip for a 730. Or they were short lived and never got a bios upgrade or anything. I also suspect that TechPowerUp's GPU database might have a few errors and omissions in it.

*i.e. any of the big names whose BIOS and claimed specs I'd actually trust.

So, guess I'm just rolling with the 440 GT stock configuration until I bang my head against something that really needs this. I figure matze and debs are probably right, and cyclone3D probably has a viable, less problematic solution there. Mihai makes a good point, I would have been running it as a secondary card for all flash operations, you can do this by using onboard graphics as primary or using it in a dual "16x" slot SLI/Crossfire motherboard (Though some configs will have them go 8+8 or 16+8 or whatevs depending on what number of lanes the chipset supports and what else it does with them.)

You know what's weird "out there" though, you can find benchies where 440 and 730 seem neck and neck, and you think, yah, should be if they're both same core, but they never mention which specific card they're using, or then again you find 730 and 1030 neck and neck, so those gotta be the 384 core version. It seems obvious now, but at the outset I was somehow misunderapprehending that I could get a 440 as fast as a 1030.

There are some issues with that. BIOSes are often board specific.

I've had to deal with ATIs AtomBIOS when developing the first free driver for R500 and upwards back in 2007. It does a lot of board specific things: bring up the ram, describe the display connector layout, even adjust the voltage of display signals to cope with differences in board layout. Plus, ATI engineers, when trying to convince management and other departments of their shortsighted "let's do everything in the BIOS" plan, promised, pinky swear, that they would never put such board specific values in AtomBIOS "functions", and would abstract them away in board specific Atombios "data" tables. But that of course went straight out of the door when the R600 came around.

Afaict, the same is true for Nvidia, but since Nvidia never went open source like AMD, there is little documentation there.

There is a ATOMBios disassembler called atomdis (https://cgit.freedesktop.org/~mhopf/AtomDis/) that Matthias Hopf wrote as part of our open source project.
You can use my version of flashrom (https://github.com/libv/flashrom/tree/ati) for a linux tool that can flash many slightly older ATI pci-express boards. I have no access to vega and newer, and the way prices are evolving, it'll take a few years until i go there.

Oh, and we of course never went to tooling to generate our own BIOSes. SeaBIOS, combined with an atombios interpreter (the MIT licensed xfree86 code that Egbert Eich wrote for instance), and the reverse of AtomDIS would've been an interesting project.