@dleone Thank you for taking the time to address our concerns. I’m afraid, however, that the driver that you’re referring to, 535.43.11, has been the main topic in the vastly popular thread about Starfield for the past few days. Many people have tested it and shared already their reports.
The good news is that it does seem to fix most – though not all – problems with Starfield on the Turing architecture. The bad news is that on Ampere, compared to 535.43.10, it reintroduces the Xid 109 errors, known from the stable 535 series. This seems like a clear regression compared to 535.43.10. The latter did have its own problems, mainly performance-related, but at least the Xid 109 seemed to be fixed also on Ampere.
In fact, even the developer of vkd3d-proton noticed the reintroduced crashes and had reported them to you. Other people in the Starfield thread seem to be noticing Xid 109s in other games too, which may be even more concerning.
I’m sure everyone here appreciates highly the hard work of Nvidia’s Linux team and I believe that each one of you is accomplishing well beyond what can be expected from the undoubtedly limited resources that you have. I personally have always liked your drivers: sometimes slow to catch up, but dependable and performing.
But the situation with Starfield – and Cyberpunk’s Xid 31 errors, for that matter – isn’t that reassuring, I’m afraid. While I have no doubt that your team is very dedicated and giving your very best, in my humble opinion your management does need to seriously consider allocating more resources to the Linux development.
Linux gaming might not be that big, but it’s certainly showing a growing trend. Most importantly perhaps, many of us are not only gamers. Nvidia’s reputation in fields like AI is also relevant to many of us. And the experience we have with Nvidia’s Linux drivers can influence our future purchasing decisions, even in these broader domains. It may not be the primary factor, but it will be an important factor nonetheless. It’s a matter of trust.