I’m investigating suitability of graphics cards for a future product. I’m a hardware/systems designer, not a software developer.
Are there restrictions on software that which prevent it running on gaming cards but allows running on workstation cards? The GTX1080 hardware appears to be only slightly less powerful than the Quadro P5000 but is much less expensive.
Are some features or libraries blocked or not supported on the gaming cards?
Can software developed on a Quadro P5000 then be run on a GTX 1080? If yes, what needs to be done to ensure compatibility?
Can you describe, at least in generic terms, what kind of product this is? If the focus is on GPU computing, I would suggest looking at the Tesla line for comparison with consumer GPUs, rather than Quadro. The Quadro line is geared towards professional graphics and image processing applications.
This is outside my area of expertise, but my understanding is that the value of the Quadro line is mostly in the certified drivers that are optimized for a large number of relevant products used in professional graphics processing. See this link for details: [url]http://www.nvidia.com/object/quadro-certified-drivers.html[/url].
I have vague recollections of Quadro drivers providing significantly higher line-drawing performance, compared to consumer graphics cards, which is (or at least, was) useful in CAD applications. I seem to recall that there was some comparison data for of consumer cards and Quadro in the SPECviewperf benchmark, but checking spec.org right now, it seems SPEC very recently introduced a new version of the benchmark, and none of the new results are for consumer cards: [url]http://spec.org/gwpg/gpc.data/vp12.1/summary.html[/url]
The application is video processing for medical applications. We would like to run proprietary algorithms on the video stream as well as use computer vision and ind image processing libraries.
Sorry, I don’t have insights into video stream processing. In order for others to give relevant advice, it might be good to know whether you expect your custom pipeline to be dominated by NVENC usage, or by custom processing via CUDA kernels.