How We Configured Our Test Systems: We tested the graphics cards in this article on a Gigabyte Aorus X299 Gaming 7 Pro motherboard powered by an Intel Core i7-7820X octal-core processor and 32GB of G.SKILL DDR4 RAM clocked at 2666MHz. The first thing we did when configuring the test system was enter the UEFI and set all values to their "high performance" default settings and disable any integrated peripherals that wouldn't be put to use. The memory's clock was manually dialed in, to ensure optimal memory performance at the processor's maximum supported speed of 2666MHz (without overclocking) and the solid state drive was then formatted and Windows 10 Professional x64 was installed and fully updated. When the Windows installation was complete, we installed all of the drivers, games, and benchmark tools necessary to complete our tests.
NVIDIA TITAN V Installed In Our Test Bed
We should note that the
AMD Radeon RX Vega 64 was tested with its default "Balanced" power mode. Power Saver (slower) and Turbo (faster) power modes are also available with Vega-based cards, which would affect performance, noise output, and peak power consumption. In addition, it behooves us to say that the
NVIDIA TITAN V IS NOT TARGETED AT GAMERS. The TITAN V is destined for AI and Machine Learning applications, and it is priced as such. The card is not specifically tuned for gaming and despite its killer performance (as you'll see); the value proposition for gamers is simply not practical. However, game performance of the TITAN V may give us a glimpse of what future consumer-targeted GPUs based on NVIDIA's
Volta architecture may offer, and we're sure many of you reading this are curious as to what the card can do. Of course, we have an array of compute-focused tests on tap as well that are more targeted on the TITAN V's intended audience.
|
|
| HotHardware's Test System |
| Intel Skylake-EX Powered |
|
|
|
| Unigine Superposition |
| Pseudo-DirectX / OpenGL Gaming |
|
Superposition is a new benchmark from Unigine, powered by the UNIGINE 2 Engine. It offers an array of benchmark modes, targeting gaming workloads as well as VR, with both DirectX and OpenGL code paths. There is an extreme hardware stability test built-in as well. Unigine Superposition uses the developer’s unique SSRTGI (Screen-Space Ray-Traced Global Illumination) dynamic lighting technology, along with high quality textures and models, to produce some stunning visuals. We ran Superposition in two modes using the DirectX code path – 1080p Extreme and VR Maximum -- to compare the performance of all graphics cards featured here.
Unigine Superposition
The NVIDIA TITAN V simply crushed every other card in this benchmark. This test is very compute-intensive, and as such, the TITAN V's beastly GV100 GPU is able to churn through the benchmark significantly faster than anything else on the market currently. Note that performance is more than double the GeForce GTX 1080.
NVIDIA's flagships GPUs are able to max-out the overall score in this benchmark (it's time to move onto VR Future mode), hence the grouping of 10K scores. If you look at the individual results, however, you'll see that the TITAN V offered measurably higher framerates and lower latencies than every other card we tested.
|
| Futuremark VRMark |
|
Testing Rift And Vive Readiness |
|
Futuremark’s VRMark is designed to test a PC’s readiness for the
HTC Vive and
Oculus Rift virtual reality headsets. The benchmark does not, however, require that one of the headsets is attached to the PC and it uses an in-house graphics engine and content to ensure comparable results between different platforms. We ran the "Orange Room" VRMark test at defaults settings here.
VRMark's Orange Room benchmark is mostly CPU-bound with a high-end GPU, hence the tight grouping here. Regardless, the TITAN V jumped out to yet another lead over every other card in the mix.
|
| 3DMark Time Spy |
|
Direct X 12 Performance |
|
3DMark Time Spy is a relatively new DirectX synthetic benchmark test from Futuremark. It features a DirectX 12 engine, built from the ground up, to support bleeding-edge features like asynchronous compute, explicit multi-adapter, and multi-threading. Time Spy is designed to test the DirectX 12 performance of the latest graphics cards using a variety of techniques and varied visual sequences. This benchmark was developed with input from AMD, Intel, Microsoft, NVIDIA, and the other members of the Futuremark Benchmark Development Program, to showcase the performance and visuals potential of close-to-the-metal, low-overhead APIs.
3DMark Time Spy
The
TITAN V surged to yet another huge lead in the DX12 3DMark Time Spy benchmark. It didn't quite double the performance of the GTX 1080 or Vega 56, but its lead was significantly nonetheless.
|
|
| Futuremark 3DMark Fire Strike |
| Synthetic DirectX Gaming |
|
3DMark Fire Strike has multiple benchmark modes: Normal mode runs at 1920x1080, Extreme mode targets 2560x1440, and Ultra mode runs at a 4K resolution. GPU target frame buffer utilization for normal mode is 1GB and the benchmark uses tessellation, ambient occlusion, volume illumination, and a medium-quality depth of field filter. The more taxing Extreme mode targets 1.5GB of frame buffer memory and increases detail levels across the board. Ultra mode is explicitly designed for high-end and CrossFire / SLI systems and cranks up the quality even further. GT 1 focuses on geometry and illumination, with over 100 shadow casting spot lights, 140 non-shadow casting point lights, and 3.9 million vertices calculated for tessellation per frame. And 80 million pixels are processed per frame. GT2 emphasizes particles and
GPU simulations.
3DMark Fire Strike
The TITAN V held onto small leads in the Combined and Game Test 2 portions of this benchmark, but it was markedly faster than every other card in Game Test 1. With how the scores are weighted though, the TITAN V ends up finishing about 10% higher than a TITAN Xp.
