Please start a new thread about your problem.
Since this is kind of political, I think a reply to you is more appropriate, so I'll do that and then start a new thread.
I have four machines running in this household. One AOpen 1547 laptop running an i810 chipset, one P2-266 mongrel running a Banshee, and two nForce-2-based workstations.
Despite the i810 being a fairly sucky chipset, the laptop's graphics work flawlessly, and at full speed. The video driver is 100% FOSS. It has never crashed.
Despite the P2-266 being as old as the hills, its graphics also work flawlessly (although not so fast, due to the CPU and the Banshee). It also has never crashed; my daughter leaves it on full time. The driver is 100% FOSS.
My own desktop, an MSI-nForce2-based "NVCrush11 [GeForce2 MX Integrated Graphics] (rev b1)" white box, runs the FOSS nv driver. As long as I do that, it never crashes, but it is very slow, even in 2D, apparently because NVidia won't publish enough detail for the authors to make it go fast.
Running the closed nvidia driver results in 3-5 lockups per day (the X driver freezes, spinlocked on something, and crashes the card if you kill -9 it). This has persisted through four different versions of Linux, and despite every imaginable combination of AGP and other options.
My wife's desktop, another nForce-2-based white box "NV18 [GeForce4 MX - nForce GPU] (rev a3)" locks up approximately daily if it's run on the closed driver, never on the open one. When I say "locks up" I mean it's doorbell time, and sometimes even that doesn't work, you have to pull the cable out. Likewise, all reasonable combinations of acceleration and AGP options produce no appreciable change.
Now... if the NVidia drivers were FOSS, I would stand a chance of debugging them enough to have 100% reliable, fully accelerated machines.
There is now a FOSS driver for the nForce2 LAN chipset, done by two teams, one to analyse and spec, one to write a driver.
Why? I've not seen any reliability issues with the LAN interface. The answer is that they can freely distribute it without raising any complicated licence issues. With it, a 100% FOSS distribution can be installed and have enough networking capabiltiies to fetch anything else it needs "out of the box". Not so nvnet, which has a binary component which recently caused me trouble with different compiler versions.
If NVidia GPLed as much of their code as was not completely nailed down by technology licences, this would not be necessary.
Whole countries (the latest Vietnam) are switching to Open technologies; it is to NVidia's direct benefit to play along with this. The more freedom that the groups using FOSS have with your drivers, the more of them will specify NVidia-based hardware.
I'm currently negotiating with the XGI people about opening their Linux drivers. If they open their drivers but NVidia keep them closed, the cheap-workstation world will be flooded with grandsons of S3 instead of potent NVidia chipsets.
Just to head off the "competitive advantage" phantom: all of NVidia's real magic is in the hardware. FOSS people don't have the resources to reverse-engineer your software effectively, ATI and XGI do. NVidia have nothing to lose by Opening their drivers.
How about it?