On Tue, 2021-11-23 at 12:58 -0600, Roger Heflin wrote:
> This is the magic decoder ring web page. The most recent cards that
> I think were kicked out are anything with a GK* in the code name
> (kepler). So any kepler cards seem to stop at 470.X. And GF*
> stopped a while ago at 340 I think. If your card just stopped
> working it is probably a GK* variant card.
>
> But some models have 2 generations of chips.
>
> https://en.wikipedia.org/wiki/List_of_Nvidia_graphics_processing_units#GeForce_700_series
I gave up on NVidia video a long time ago, and this is one of the
things that bugs me about having cards that *could* be used, but if
only you could figure out how to do it:
I have no problem using older NVidia hardware and living within
nouveau's limitations, but I will go to macOS or Windows to get
a job done.
*It's* the damn computer, why should *we* have to figure out which
driver to install?
There is no business case for graphics hardware vendors to invest
in helping people use old hardware. Open source,however, needs
time to reverse engineer drivers, so by the time the vendor drops
support, there is often some level of open source support.
At my former work in a group that used remote sensing data, linux
was never viable due to the extra effort needed to maintain colour
fidelity, so we moved from SGI to Apple hardware, using linux only
for work that did not have demanding colour fidelity or 3-D requirements.
Either they could have one large package with everything it needs to
install the one that you do (somewhat akin to CUPS, though I know it
doesn't have absolutely everything within itself). Or, they could have
a detector package which surveys your graphics hardware and picks the
right driver package to download and install.
This approach leads to every increasing complexity so developers
move to a new more capable low-level protocol:
CUPS FUTURE: Remove printer driver and raw queue support. #5271:
Like Xorg, CUPS was becoming unmaintainable. IPP Everywhere is
the CUPS analog of wayland. I would like to see a proper analysis
of the issues that slowed Nvidia's wayland support. My impression
of IPP everywhere is that it provides a standard interface for
applications by collecting extensive details of the printer's
parameters and generic page description formats. Newer printers
support IPP directly, and legacy printers can use translation layers
(like Xwayland?).
I had to go through the same crap with Windows 98SE, eons ago: You'd
buy a card and it came with a driver CD-ROM full of a gazillion drivers
for your card, and a plethora of other things (often with dozens of
things unrelated to your kind of hardware). Then, you had to figure
out which one to install by divine inspiration. It didn't probe and do
it (or if it did, it got it wrong). Your best chance was to look at
the chipsets on the card, and try to find a driver with similar details
in the filename.
A long time ago, I encountered cards that used ATI chips but had
eliminated one of the ancillary chips used in real ATI cards. The
card came with drivers that replaced the missing functions with
software, but only for one Windows version. Real ATI drivers
told us the cards were broken.