Titan RTX Review

Compute Benchmarks

The Turing architecture is very compute oriented. With Turing being a revision on the Volta architecture, Nvidia moved forward the cache and scheduler changes that made Volta great at what it was for. The benefits show here. Most of the data presented in this section is going to be based on SpecViewPerf 13, a bench suite designed to test cards on professional workloads. This test covers 3d Studio Max, Solidworks, Maya, Catia, Creo, and others. Our compute bench selection will grow going forward, but due to limited time those tests wont be present here. Blender is another test that we want to test, but due to 2.78 not working without a recompile, and 2.8 being in beta we are taking time to evaluate which route would be best for our viewers. On to the graphs!

SpecViewPerf 13 Weighed Results

The graph above is a weighed result between all sub tests in each category. As you can see here, for the most part the Titan RTX doesn’t offer much over the much cheaper (never thought I’d use that adjective for turing) 2080ti, with a few exceptions. If you are working in Catia, Energy (not familiar with this specific software) or Siemens NX, the Titan is solid winner. Below you will find a break down for each test.

Rendering

For this section, we are currently using LuxMark 3.1 and V-Ray.

LuxMark – Based on the Luxcore renderer, Luxmark was developed as a showcase benchmark for rendering with OpenCL. Its been a staple in the benchmark community for some time and we see no reason to stop using it now.

The results really speak for themselves here. The Titan RTX is 14.6% faster on average with the biggest lead on the scene with the least complex geometry.

V-Ray is the next bench up. V-Ray is a benchmark based on the popluar V-Ray plugin that works with many rendering programs. V-Ray is an add on rendering engine used to accelerate rendering in 3d applications while supporting more up to date rendering methods.

Here you can see that the 2080ti is 8% slower than the Titan RTX (forgive the chart, still figuring out how to set min/max values). This lines up with the physical differences in rendering power between the cards.