3 ms·

The thing to understand about Nvidia is that they use the same model as Apple: Produce really nice software, and use it to drive hardware sales. On the driver

by wtracy 8y ago

The thing to understand about Nvidia is that they use the same model as Apple: Produce really nice software, and use it to drive hardware sales. On the driver side, a lot of the features and performance advantages of Nvidia cards are rumored to be largely implemented in software. (Think of things like having a really good optimizing compiler for shaders.) Open sourcing that code would allow other brands to copy a lot of their techniques. On the application side, Nvidia produces a lot of developer tools (like CUDA) that are either tied to or heavily optimised for Nvidia hardware. All this means that Open Source is largely antithetical to the way they do business. OTOH, ATI Linux drivers went through a very rough patch when ATI decided that Open Source meant, "Hey, we can get a bunch of suckers to write drivers for us!" Everyone involved seems to have actually come out the other side stronger for the experience, and I'm very optimistic about their future. That said, my understanding is that you are committed to Linux, you still get a more robust OpenGL implementation (ATI used to be buggy as sin) and better shader performance with Nvidia hardware/software. The situation is changing rapidly, and I would love to know if the information at the start of this paragraph is out of date. Now that I've provided an infomercial for Nvidia: I'm willing to defend Nvidia's products, but I'm not really ready to defend Nvidia the company. Even if Nvidia's hardware beats the competition on all the benchmarks relevant to your use case, it's entirely reasonable to prefer choosing a company that is more willing to play nice with the greater community.

nybble41 8y ago
> That said, my understanding is that you are committed to Linux, you still get a more robust OpenGL implementation (ATI used to be buggy as sin) and better shader performance with Nvidia hardware/software. That's probably true, if your use case requires the absolute best GPU performance. For me, though, the Intel and AMD open source drivers have gotten good enough that the marginal improvements possible with proprietary drivers simply aren't worth the hassle. At least with AMD you are free to choose between Mesa and Catalyst, and can be assured of a decent experience either way. If you opt for nVidia hardware you're effectively locked in to their proprietary driver if you want to use even half of your GPU's capabilities. It helps that Intel and AMD and all the non-PC Mesa drivers share common infrastructure, so improvements to one driver tend to enhance the entire ecosystem—except nVidia. As you noted, the situation is changing rapidly; I think it won't be much longer before nVidia's proprietary drivers forfeit whatever performance advantage they have left to the combined efforts of the open source community.
- wtracy 8y ago
  You know, my advice is colored by the use cases I have: I'm usually in one of two situations: First, situations where performance doesn't matter, and it needs to be cheap and Just Work. Intel beats everyone else here. (I'm really eager to see what Intel does with their upcoming discrete video card.) Second, situations where performance is a Big Deal, and I'm willing to both pay money and jump through hoops to get it. Nvidia was winning here the last time I looked closely. ATI products exist somewhere in the middle, and I can't remember the last time that I needed anything like that.

by wtracy 8y ago

nybble41 8y ago

> That said, my understanding is that you are committed to Linux, you still get a more robust OpenGL implementation (ATI used to be buggy as sin) and better shader performance with Nvidia hardware/software. That's probably true, if your use case requires the absolute best GPU performance. For me, though, the Intel and AMD open source drivers have gotten good enough that the marginal improvements possible with proprietary drivers simply aren't worth the hassle. At least with AMD you are free to choose between Mesa and Catalyst, and can be assured of a decent experience either way. If you opt for nVidia hardware you're effectively locked in to their proprietary driver if you want to use even half of your GPU's capabilities. It helps that Intel and AMD and all the non-PC Mesa drivers share common infrastructure, so improvements to one driver tend to enhance the entire ecosystem—except nVidia. As you noted, the situation is changing rapidly; I think it won't be much longer before nVidia's proprietary drivers forfeit whatever performance advantage they have left to the combined efforts of the open source community.

wtracy 8y ago
You know, my advice is colored by the use cases I have: I'm usually in one of two situations: First, situations where performance doesn't matter, and it needs to be cheap and Just Work. Intel beats everyone else here. (I'm really eager to see what Intel does with their upcoming discrete video card.) Second, situations where performance is a Big Deal, and I'm willing to both pay money and jump through hoops to get it. Nvidia was winning here the last time I looked closely. ATI products exist somewhere in the middle, and I can't remember the last time that I needed anything like that.