- Moderator
- #21
Jeff Kilgroe
Well-known member
GTX TITAN Signature Superclocked edition has just landed here.
Testing will commence this evening.
Readback performance is going to continue being better on Quadro cards for now -- or at least in this manner as Scratch and many other OpenGL based software handles it. Comes back to legacy OpenGL 2.5 and prior implementations using glReadPixels as the primary readback method. Newer OpenGL specs have brought us superior ways to handle readback with direct framebuffer access and large frame memory transfers, many of which are handled through extensions, so the shift over is not always that simple. One of the common work-arounds I'm seeing is that developers are resorting to using a function built with CUDA or OpenCL to replicate the same functionality in a way that's faster on the GeForce and Radeon cards, whereas the older OpenGL functions are faster on Quadro and FireGL cards. ...Yes, AMD has this same issue.
Based on the Scratch profiler, if I were a Scratch user, I wouldn't be yanking that Quadro 6000 out of my workstation anytime soon.
Anyway, I'll be starting tests with the Quadro 6000, Titan and GTX 690 and various combinations of each. I'm expecting the Titan + 690 to be the winning combo for Resolve. Titan should be the winner in Premiere, assuming drivers are up to snuff. Quadro is still going to rule for pro OpenGL stuff, as we're seeing with the Scratch profiler -- just the way I'm expecting it to unfold.
I'm curious to see just how far I can push the clock speed on the Titan...
Readback performance is going to continue being better on Quadro cards for now -- or at least in this manner as Scratch and many other OpenGL based software handles it. Comes back to legacy OpenGL 2.5 and prior implementations using glReadPixels as the primary readback method. Newer OpenGL specs have brought us superior ways to handle readback with direct framebuffer access and large frame memory transfers, many of which are handled through extensions, so the shift over is not always that simple. One of the common work-arounds I'm seeing is that developers are resorting to using a function built with CUDA or OpenCL to replicate the same functionality in a way that's faster on the GeForce and Radeon cards, whereas the older OpenGL functions are faster on Quadro and FireGL cards. ...Yes, AMD has this same issue.
Based on the Scratch profiler, if I were a Scratch user, I wouldn't be yanking that Quadro 6000 out of my workstation anytime soon.
Anyway, I'll be starting tests with the Quadro 6000, Titan and GTX 690 and various combinations of each. I'm expecting the Titan + 690 to be the winning combo for Resolve. Titan should be the winner in Premiere, assuming drivers are up to snuff. Quadro is still going to rule for pro OpenGL stuff, as we're seeing with the Scratch profiler -- just the way I'm expecting it to unfold.
I'm curious to see just how far I can push the clock speed on the Titan...