Rendering and Compute Performance

Luxmark 3.1 

OpenCL rendering performance is important for workstation-level graphics card hardware. Luxmark, one of the most widely used OpenCL performance tests, provides a good look at how different GPUs perform in typical OpenCL rendering workloads. For this test, we are using the "Hotel" Scene in Luxmarkthe most compute-intensive scene, consisting of almost 5000 triangles.

Going from the NVIDIA Titan Xp in Luxmark to the Titan V, we see a massive 78% increase in performance. Similarly, the Titan V manages to double the score of the AMD Radeon Vega 64.

Cinebench R15

The performance depends on various factors, such as the GPU processor on your hardware, on the drivers used. The graphics card has to display a huge amount of geometry (nearly 1 million polygons) and textures, as well as a variety of effects, such as environments, bump maps, transparency, lighting and more to evaluate the performance across different disciplines and give a good average overview of the capabilities of your graphics hardware. The result is measured in frames per second (fps). The higher the number, the faster your graphics card is.

While the CPU portion of Cinebench is a tool we often use to evaluate the performance of new processors, it also offers an OpenGL test for assessing GPU performance.

Although both the NVIDIA offerings are about 50% faster than the AMD Vega 64, there's virtually no difference between the Titan V and the previous generation Titan Xp in this test.

V-Ray Benchmark

V-Ray is popular third-party renderer that plugs into the most powerful CAD and 3D modeling applications. With plugins for 3ds Max, Maya, Revit, Rhino, and more, V-Ray is widely used for high-quality renderings in commercial applications such as architecture and product design.

V-Ray Benchmark is a free standalone application which allows users to evaluate hardware without having to install a full suite of software or provide a software license. For AMD GPUs, V-Ray Benchmark uses an OpenCL renderer, while for NVIDIA GPUs a CUDA-enabled renderer is used.

In V-Ray rendering, the Titan V is over 3.3x the speed of Vega 64, and over 25% faster than the previous generation Titan Xp.

SiSoft Sandra 2017 GPGPU Compute

SiSoft Sandra is a suite of benchmarks covering a wide array of system hardware and functionality, including an extensive range of GPGPU tests, which we are looking at today. 

The first GPGPU test in SiSoft Sandra evaluates the shader performance of a given GPU at different precision levels. 

While both the Titan Xp and AMD Vega 64 are within 25% of the performance of the Titan V in single precision workloads, the GV100 silicon in the Titan V obliterates every other GPU when it comes to double precision workloads. The Titan V is 14.7x faster than the Titan Xp and 6.6x faster than the Vega 64!

SiSoft Sandra 2017 GPGPU Financial Analysis

With the GPGPU Financial Analysis test in Sandra 2017, we see a more real-world application of the double precision floating point performance with 8x the performance of the Vega 64, and over 13x the compute performance of the Titan Xp.

SiSoft Sandra 2017 GPGPU Scientific Analysis

In the Scientific analysis test, we can see the increased double precision performance of the Titan V in N-Body simulation and GMM workloads, but the playing field is more leveled in single precision FFT workload.

SiSoft Sandra 2017 GPGPU Image Processing

The image processing test shows a different outcome than the rest of our previous benchmark. AMD's Vega 64 GPU takes the lead here with an almost 25% lead over the Titan V. This is due to benchmark being able to take advantage of half-precision (FP16) shaders.

The Vega architecture has the ability to run half-precision workloads at exactly double the rate of FP32 operations, which is a huge advantage in tasks that don't need high precision as we see here. This is a good sign for the potential of developer optimaztion for Vega changing the competitive landscape.

« PreviousNext »