"P" is for Pascal : The NVIDIA Quadro professional line-up will be blessed with the latest GPU architecture... and the highest performance ever.

NVIDIA is accelerating their GPU release schedules. The Titan series launched just shortly after the initial release of the new Pascal architecture. Now the professional line up will see performance boost in just a couple of months. With professional users converging on Siggraph 2016, NVIDIA is giving them a mid-summer Christmas present: the promise of a fast new set of high-performance GPUs. 

Quadro P6000 : the fastest NVIDIA graphics board ... ever

Here are the highlights: 

• NVIDIA's fastest architecture comes to Quadro. The Quadro P6000 and P5000 are based on the GP102 graphics processor
• The Quadro P6000 is the most powerful graphics board in the NVIDIA line-up ... ever. It has nearly 300 more CUDA cores than the Titan X announced a week earlier.
• Simultaneous, multi-projection allows for left & right eye projections to be created in a single pass, boosting VR / stereo 3D applications
• The boards are loaded with the fastest GDRR5X memory which delivers twice the bandwidth. This is critical for GPU computing.
• Unified virtual memory will accelerate GPU-computing applications 
• Dynamic load-balancing of graphics and computing applications delivers better GPU-computing & graphics mixed-mode operation
• These new Quadro products are made to accelerate iRay and Mental Ray GPU-based ray-tracing, global illumination, and other realistic rendering techniques
• DisplayPort 1.4 support allows the new GPUs to deliver 8K resolution and both products can drive four, 5K resolution displays simultaneously

The basic characteristics of the two new Quadros are:

It is worth repeating: The Quadro P6000 is absolutely the fastest, most powerful graphics board in the NVIDIA family. Not only does it have 24 GB of GDDR5X memory, twice the memory of the Titan X, it also has 3840 CUDA cores compared to the Titan X with 3584 CUDA cores.   There is no faster GPU in the NVIDIA line up. 

Another key point for professionals, the Quadro P6000 and P5000 unified memory architecture is perfect for compute applications, especially realistic rendering using advanced ray-tracing techniques. The unified memory architecture allows scenes of unlimited size to be calculated and rendered. It's important to understand that the unified memory architecture is possible on Linux, but it does not exist under Windows. 

Virtual Reality, VR, is all the rage in the consumer space. Professionals, on the other hand, have been struggling with VR, stereoscopic 3D displays, and the relevant technical problems for over two decades.  Thanks to the Pascal GPU architecture, the new Quadro products use a technique called single-pass multiprojection. The technique allows the Quadro P5000 and P6000 to process the 3D scene one time yet generate two perspective views: one for the right eye and one for the left eye. This doubles the performance for stereo projections which, in turn, doubles the complexity and fidelity of the VR, 3D stereo image. 

The ability to dynamically balance graphics and computing work on the GPU enhances the Quadro's ability to work interactively with realistic, GPU-computed raytracing in the application viewport. This example of a computing and graphics combined workflow fits perfectly for An example: imagine iRay realistic rendering in a viewport of Autodesk Maya or 3DS MAX. The resulting workloads combining compute and graphics are handled easily and naturally by the GPU resulting in a smoother, more efficient working environment. 

The professional world is moving beyond 4K. The new Quadro P5000 & P6000 will be perfect fits to this need. The new GPUs support 8K resolutions. They also can drive four simultaneous 5K resolution displays. That is a boon to professionals in video and special effects and it will be a benefit to designers and engineers when the high-resolution displays fall into the sub-$1000 category as 4K monitors have done today.

NVIDIA loves to have customers buying product higher up the performance curve. With all the great tools, SDKs, unique technologies, performance and productivity that the NVIDIA professional products bring to the workstation market, the company gives the marketing team just a bit too much leash. NVIDIA regularly over-specs the match-up of Quadro products to software applications. Take this slide from NVIDIA. As a workstation user, you can shift every single category one unit to the left. 

What this means is that the category "Office, Sketchup" falls off the left side of the chart and you just run that on Intel integrated graphics. AutoCAD is the same. This leaves Revit, Inventor, Photoshop, Illustrator, etc running on entry-level GPUs. All of your basic 3D modeling and CAD applications can run on a Quadro K1200. If you followed the original "recommendations" of the NVIDIA table, you might think that a Quadro M2000 could bring you some benefit over a Quadro K1200 (both Maxwell architectures, by the way) for Autodesk Inventor. It won't. Forget it and save some money.   At PW, we believe in getting the right tool for the job and when you need a really fast GPU or a very high end workstation, then the extra 1000s of dollars are worth every penny. But to claim that you might want an M2000 for Inventor is just foolish.  Likewise, a Quadro K1200 does a fine job on Adobe Premier Pro. Yes, you might want a faster GPU in some cases, but the K1200 is a great GPU for a lot of Premier Pro work.

A final point on the NVIDIA chart, if you are using Ansys, Abaqus, and Simulia products for simulation, the chances are good that a Tesla GPU with fast double-precision floating point calculations is the class of GPU that you require.  These GP102 GPUs do not have the double-precision performance that a GP100 GPU offers. 

The PW Perspective

NVIDIA is upgrading the Quadro family with the Pascal Architecture faster than any architecture change that we can remember. The Quadro P6000 is not only an equal in raw performance to its consumer counterpart, it is easily superior to it with nearly 300 more CUDA cores and double the graphics memory. That is refreshing. Finally, the new architecture, combined with NVIDIA's design software tools like Iray and Mental Ray, as well as the features for accelerating VR, make this an excellent professional GPU architecture.