Investigating NVIDIA GeForce Gamescom Driver Performance Improvements

NVIDIA recently released a new driver at Gamescom, in which they claim performance uplifts as high as 23% in select titles. Today, we’ll be taking a look at a few of the key titles NVIDIA is claiming performance improvements in, as these are titles we typically benchmark for our own graphics card reviews.

While the company showed numbers for their entire RTX SUPER lineup and the 2080 Ti at three different resolutions; we’ll be focusing only on the RTX 2060 SUPER at 1440p today. That said, if you’d like us to follow-up with another configuration, feel free to sound out in the comments!

Graphics Card Test Setup and Methodology

Summer 2019 Update:

We’ve undergone a fairly significant change in both our testing methods, how we display data and our test setup. Firstly, we’ve upgraded our graphics card test bench to a 3rd gen Ryzen-based system. Second, we’ll no longer be measuring ‘99.9th percentiles’ or as we’ve labeled them in the past ‘0.1% lows’ instead we’ll focus purely on both average FPS and ’99th percentiles’ as we believe they most accurately represent actual GPU performance without over complicating things for the viewer.

All game benchmarks are taken from a 3-run average either in-game or using the title’s built-in benchmark utility. Performance is recorded using OCAT: the open-source frontend for PresentMon, which itself is an open-source capture and analytics tool which supports modern APIs such as DirectX 12 and Vulkan.

For our graphics card reviews, we use a 3rd generation Ryzen-based system with components supplied by AMD, GIGABYTE, Western Digital, DEEPCOOL and Nixeus. Thanks to all those who support us by supplying vital components for our test systems, without them reports like this would not be possible.

Graphics Card Test Bench
CPUAMD Ryzen 9 3900X – PBO Enabled
MotherboardGIGABYTE X570 AORUS PRO WiFi
RAMG. Skill Trident Z Neo – DDR4-3600 CL18 16GB Memory Kit – Timings at CL-16-19-19-38
GPUGIGABYTE GeForce RTX 2060 SUPER Gaming OC
StorageWD Black SN750 1TB NVMe SSD (OS), 480GB Patriot Burst SSD, 960GB ADATA SU655 SSD
CoolerDEEPCOOL Castle 240 RGB all-in-one liquid cooler.
CasePhanteks P350X
PSUDEEPCOOL Quanta DQ1250
MonitorNixeus EDG27 v2
Software Configuration
OSWindows 10 1903
DriversNVIDIA GeForce 431.60 – WHQL / NVIDIA GeForce 436.02 WHQL

GeForce Game Ready 436.02 Performance

Forza Horizon 4

We test this title using the game’s built-in benchmark utility.

Taking a look at the above results, we can see that the new drivers offer a massive 16% performance improvement in both average FPS and the 99th percentile. This is very much in line with the 17.6% improvement NVIDIA claimed in their own tests. Not bad!

Battlefield V

We test this game during a 60-second gameplay segment of the ‘War Stories’ single-player game mode on the first level of the ‘Tirailleur’ campaign.

Here the gains aren’t quite as sexy, but definitely still significant. We have an average FPS increase of 3%, which is actually higher than the 1.6% NVIDIA claimed (although that was at DX12, which we don’t typically test at for this title). More importantly, the 99th percentile performance increases by 7%.

World War Z

We test this title using the game’s built-in benchmark utility.

Here, we have a very similar story to that of Battlefield V. Average performance improves by 3%, and 99th percentile performance by 7%. NVIDIA claims an up to 7.5% performance increase in this title, and it’s hard to argue that they’ve not delivered.

Conclusion

So, after an admittedly fairly quick analysis, we were able to easily verify NVIDIA’s performance claims. In fact, in some cases performance was even better than they promised, which is always great to see. We did also test a number of other titles in our benchmarking suite, but found that there was no actual difference in performance, or any difference observed was within the margin of error. So, we did not include those results in this report.

In our recent review of the GIGABYTE RTX 2060 SUPER Gaming OC, we found it to be on average, 7.5% slower than the RX 5700 XT. After updating our summary with these new results, we find the RTX 2060 SUPER to be just 5.7% slower on average at 1440p. That’s a fairly significant change with just a single driver update, and it really is a testament to NVIDIA’s focus on performance optimization and their ability to eke out as much performance as possible from Turing.

That said, while I’m pleased to see the strides NVIDIA are making here, this doesn’t really change my view of the RTX 2060 SUPER vs the RX 5700 XT. I would still recommend the Radeon card over the GeForce one if you’re looking for the most performance for your ~$400. However, this does make it easier for folks who really want ray tracing to justify going with the slightly slower graphics card in order to get those features.

It is worth mentioning that we’ve got a couple of custom RX 5700 XT models coming in for review next, so, this will definitely not be the last you hear from us on this matter. Stay tuned!

UPDATE: We’ve published a more detailed report, including 15 different titles tested. Check it out here.

More Stories
Seagate FireCuda 510 NMVe 1TB SSD Review
Do NOT follow this link or you will be banned from the site!