Professional Visualization and Rendering

With AMD strongly pushing the Radeon VII as a prosumer content creation card, it behooves us to look at rendering, CAD, and professional visualization performance. However, accurate and applicable benchmarks for this field are not so easy to find, especially since performance is highly dependent on workflow and proprietary licensed ISV software. Given AnandTech’s audience, which often includes engineers using these applications in critical production environments, our goal is to provide the most relevant metrics. However, as Ian has discussed previously, the route to the most accurate workstation benchmarking for professional applications is in the hands of ISVs, who are at best blasé and more typically negative about providing access, even at the prospect of lending limited software licenses in return for ongoing discussion and third-party benchmark data of their software.

Those caveats in mind, the next best thing for evaluating overall GPU workstation performance is the venerable SPECviewperf, recently updated to version 13. Separated into ‘viewsets,’ which are a group of application-specific workloads derived from real-world datasets, SPECviewperf has been a longstanding suite for generalized workstation/CAD GPU performance. For SPECviewperf 13, the viewsets are based on:

  • Autodesk 3ds Max 2016 (Nitrous DX11 driver)
  • Dassault Systèmes CATIA V6 R2012
  • PTC Creo 3 & Creo 4
  • Geosurvey software, with workloads based on rendering techniques utilized by the open-source OpendTect seismic visualization application
  • Autodesk Maya 2017
  • Radiological (i.e. CT, MRI scans) rendering, with workloads using the Tuvok rendering core of the ImageVis3D volume visualization application
  • Autodesk Showcase 2013
  • Siemens NX 8.0
  • Dassault Systèmes Solidworks 2013 SP1

While we didn’t have time for complete benchmarking of video editing/production software such as Adobe Premiere Pro CC, we will be looking to include that in the future.

Compute/ProViz: SPECviewperf 13 - 3dsmax-06

Compute/ProViz: SPECviewperf 13 - catia-05

Compute/ProViz: SPECviewperf 13 - creo-02

Compute/ProViz: SPECviewperf 13 - energy-02

Compute/ProViz: SPECviewperf 13 - maya-05

Compute/ProViz: SPECviewperf 13 - medical-02

Compute/ProViz: SPECviewperf 13 - showcase-02

Compute/ProViz: SPECviewperf 13 - snx-03 (Siemens NX)

Compute/ProViz: SPECviewperf 13 - sw-04 (Solidworks)

Looking over the results, it's clear that certain viewsets tend to perform better on one vendor's hardware than the other's. In those cases, the Radeon VII doesn't buck the trend, though in Siemens NX the lower performance is more likely than not related to driver maturity. In the reverse scenarios like in creo-02 or maya-05, the Radeon VII is in a similar spot, naturally ahead of the RX Vega 64 but behind the competing RTX and GTX cards. If anything, the results highlight the importance of software maturity for newer hardware, but there are definite signs of Vega 20 being a powerful workstation card. The caveat is that it doesn't seem to change the overall landscape for worksets that traditionally perform well on NVIDIA hardware.

Our next set of benchmarks look at rendering performance. To be clear, given the nature of ‘render wars’ as well as the adoption of CUDA, the featured render engines are not necessarily indicative of the overall GPU renderer landscape. Because we are looking at the Radeon VII, it’s not applicable to include some of the more popular renderers, such as Redshift and Octane, which are CUDA-only, and similarly the presence of Indigo Renderer helps as another datapoint even though it is less popular.

Compute/ProViz: LuxMark 3.1 - LuxBall and Hotel

Compute/ProViz: Cycles - Blender Benchmark 1.0b2

Compute/ProViz: V-Ray Benchmark 1.0.8

Compute/ProViz: Indigo Renderer 4 - IndigoBench 4.0.64

To note, official Blender releases have yet to incorporate CUDA 10, and so RTX 20 series cards are not officially supported.

V-RAY here is the only test that utilizes CUDA for NVIDIA cards, while the rest all use OpenCL. The results seem broadly similar to SPECviewperf, where the Radeon VII continues to excel at workloads where AMD hardware generally fare well.

Synthetics Radeon VII and RX Vega 64 Clock-for-Clock
POST A COMMENT

293 Comments

View All Comments

  • i4mt3hwin - Thursday, February 07, 2019 - link

    So FP64 is 1:4 and not 1:8 or 1:2 as previously known? Reply
  • tipoo - Thursday, February 07, 2019 - link

    Yep, looks like they changed the cap in vBIOS based on feedback.

    Which also means they could have uncapped it, but it's still cool that they did that.
    Reply
  • Ganimoth - Thursday, February 07, 2019 - link

    Does that mean it could be potentially unlocked by some bios mod? Reply
  • tipoo - Friday, February 08, 2019 - link

    I hope so! Reply
  • Hul8 - Thursday, February 07, 2019 - link

    I don't think it was ever reported or assumed to be 1/2 - that best possible ratio is only for the pro MI50 part. Early reports said 1/16. Reply
  • Ryan Smith - Thursday, February 07, 2019 - link

    For what it's worth, when we first asked AMD about it back at CES, FP64 performance wasn't among the features they were even throttling/holding back on. So for a time, 1/2 was on the table. Reply
  • GreenReaper - Thursday, February 07, 2019 - link

    So it was *your* fault! ;-p Reply
  • BigMamaInHouse - Thursday, February 07, 2019 - link

    Asrock just posted vBios: is this with the FP 1:4 or newer?
    https://www.asrock.com/Graphics-Card/AMD/Phantom%2...
    Reply
  • Ryan Smith - Thursday, February 07, 2019 - link

    We're not currently aware of any Radeon VII cards shipping with anything other than 1/4 rate FP64. Reply
  • BigMamaInHouse - Friday, February 08, 2019 - link

    So maybe it's new bios with some fixes?
    Did you tried it since all cards are the same reference design?
    Reply

Log in

Don't have an account? Sign up now