Introduction

TechPowerUp is one of the most highly cited graphics card review sources on the web, and we strive to keep our testing methods, game selection, and, most importantly, test bench up to date. Back in May 2018, we updated to a 6-core "Coffee Lake" system, and today, we are pleased to introduce the latest 2019.2 version of our review setup.



With a more capable processor, an easier-to-use motherboard, and the latest updates to all our testing software and games, the new bench will equip us for a stream of graphics card launches throughout the second half of 2019. In this mini article, which is a TechPowerUp first, we describe our new VGA test machine, list the changes we've made, and, as a quick bonus, put out summarized performance numbers for our current selection of graphics cards on the new bench.

A VGA rebench is a monumental task for TechPowerUp. It involves testing 18 graphics cards in 21 games at 3 resolutions, or 63 game tests per card, which works out to 1,134 benchmark runs in all.

Just like for every other retest, every single graphics card has been retested on this system—we don't recycle results. Anything else would be wildly inaccurate.

Hardware

For the past twelve months, we've used an Intel Core i7-8700K, overclocked to 4.8 GHz and running on an ASUS Maximum X Hero (Z370). This setup was replaced by an eight-core, sixteen thread Intel Core i9-9900K we overclocked to 5 GHz all-core. We chose a fairly conservative overclock to ensure the system stays stable at all times—even during hot summer days. The increase in CPU power will help avoid a CPU bottleneck with the RTX 2080 Ti at lower resolutions, like 1920x1080.

The 16 threads of the 9900K will come in handy when testing CPU scaling as a wider range of CPU count configurations are now available.


As the motherboard, we chose an EVGA Z390 DARK, which I have to admit is the best motherboard I've ever used. It contains so many little quality of life features in its BIOS that make setup easier, and the angled connectors for SATA and 24-pin ATX help with cable routing. Performance is excellent, too, and the diagnostic LEDs will definitely come in handy for troubleshooting. Our CPU overclock is not that big, so overclocking headroom wasn't a factor for our motherboard choice.

We continue using all other components. 32 GB of memory is still far from being required for gaming, and nothing has changed in requirements regarding everything else. I guess further down the road, I could see me wanting larger SSDs as 1 TB is simply getting too small with our large game selection and how much space each individual game takes up.

Zen 2/Ryzen 3000 is around the corner, and if it manages to deliver gaming performance on-par with Intel—even at lower resolutions—I might be open to switching to AMD for the graphics card test system. But for that to happen, performance and market penetration has to be there, or the graphics card review comments will be full of posts "..but on Intel...".

Power Consumption Testing

For almost ten years (!), I've tested media playback power consumption in reviews using Blu-ray playback with an actual optical drive, with the Batman Dark Knight disc (to ensure AACS still has to be decrypted). Times have changed, and the vast majority of movie playback is now through streaming services.

I specifically don't want to test media streaming over the Internet as that would add a dependency on network speed, and I have no control over the data rate, format, and compression providers use or might change at any time.

That's why I've settled for power consumption testing with VLC Media Player, while watching a 4K 30 FPS video that's encoded with H.264 AVC at 64 Mbps bitrate. In a year or so, we'll update to an HEVC encoded video.

All other power consumption testing data is still valid (idle/gaming) because we measure graphics card power only, not that of the whole system.

Software and Games

• Windows 10 was updated with all the latest patches, including MDS mitigations
• The AMD graphics driver used for all testing is now Adrenalin 19.5.1 Beta
• All NVIDIA cards use 430.64 WHQL

• All existing games have been updated to their latest available version

The following titles were removed:

• Deus Ex Mankind Divided: Uses DX12, but is simply getting too old
• Dragon Quest XI: Not very popular. Uses Unreal Engine 4, but Darksiders 3 and Ace Combat 7 remain, which use the same engine
• Ghost Recon Wildlands: Too old
• Grand Theft Auto V: Too old
• Hellblade: Senua's Sacrifice: Not very popular. Uses Unreal Engine 4, but Darksiders 3 and Ace Combat 7 remain, which use the same engine
• Just Cause 4: Not very popular. Replaced by RAGE 2, which uses Apex Engine as well and supports Vulkan.

The following titles were added:

• Ace Combat 7: One of few recent flight sim releases. Uses Unreal Engine 4.
• Anno 1800: City builder/economy simulation. "Ultra" uses 8xAA for no reason, reduced to 4xAA.
• Devil May Cry 5: Very popular recent release; uses RE Engine (same as Resident Evil VII)
• Metro Exodus: Hugely popular recent release using 4A Engine with DX12. Support for RTX (disabled during normal benchmarks)
• RAGE 2: New release; uses Apex Engine (not id Software's id Tech) and supports Vulkan
• Sekiro Shadows Die Twice: One of the best-selling titles in Q1 2019. Fairly light on hardware requirements. We removed the 60 FPS limit to get proper benchmark scores on all cards

The full list of games is now: Ace Combat 7, Anno 1800, Assassin's Creed Odyssey, Battlefield V, Civilization VI, Darksiders 3, Devil May Cry 5, Divinity Original Sin 2, F1 2018, Far Cry 5, Hitman 2, Metro Exodus, Monster Hunter World, Rage 2, Rainbow Six Siege, Sekiro, Shadow of the Tomb Raider, Shadow of War, Strange Brigade, Witcher 3, and Wolfenstein 2.

Performance results are on the next page. If you have more questions, please do let us know in the comments section of this article.