People have been talking about 4K gaming for several years now, so we thought it was about time that we took a close look at whether 4K gaming is something gamers can enjoy on today's hardware. We equipped our overclocked six-core gaming system with GeForce GTX 980 Ti 6GB video cards in both single and SLI configurations, hooked it to our 4K TV, and ran it through our suite of benchmarks. But we didn't stop there. We also ran our two configurations through our entire test suite at 2560 x 1440 in order to determine if this resolution provided an overall better gaming experience using our new Acer XB270HU G-Sync monitor. To complete the picture, we'll also be analyzing SLI scaling, PCIe scaling, and VRAM usage in this article, and once all the dust and data have settled, we bet you'll be surprised at our findings!
We took up the 4K challenge because we've seen a lot of gamers making purchasing decisions based on a desire to run at 4K, even though the number of gaming-specific 4K screens is limited. We thought it was important to provide gamers the information they need to determine what it really takes to build a 4K gaming system today, and whether its worthwhile to make the investment.
Here's is the test system we used to run the benchmarks for this article:
- CPU: Intel Core i7-5820K @ 4.4GHz
- Motherboard: Asus X99-Pro/USB3.1
- Video Card #1: EVGA GeForce GTX 980 Ti ACX 6GB
- Video Card #2: Asus GeForce GTX 980 Ti 6GB
- RAM: G.Skill 4x4GB Ripjaws4 DDR4-3000
- SSD #1: Samsung SM951 M.2 256GB
- SSD #2: 2x Samsung 850 Evo 500GB in RAID0
- Case: Corsair Carbide 500R
- Power Supply: EVGA Supernova 1000 PS
- CPU Cooler: Corsair Hydro H100i
- Operating System: Windows 10
- 1440p Test Screen: Acer XB270HU 27" 144Hz G-Sync
- 4K Test Screen: Samsung UN55JS8500 55" HDTV
The system was loaded up with Windows 10, along with GeForce Driver version 353.62, desinged for use with the new OS. To eliminate bottlenecks as much as possible, we ran our six-core i7-5820K at 4.4GHz throughout the test. For our 4K tests, our benchmarking system was hooked to our HDTV via HDMI 2.0, while for our 1440p tests, the system was connected to our Acer XB270HU G-Sync monitor via DisplayPort. We used eight games for benchmarking, four using a built-in benchmark, four using in-game runs. We also ran 3DMark Fire Strike in each configuration to set a baseline.
Shown here are the video cards we used, a pair of reference-clocked GTX 980 Ti 6GB cards, the top card being an open-air EVGA ACX model, and the bottom being a blower-style model from Asus. Interesting, despite both being spec'd at 1,000MHz base clock, they did not run at the same speed in game. The ACX model was 25MHz faster. In order to collect meaningful benchmarks, we set them to run at the same speed, using MSI Afterburner to sync them, and also increased the power and temperature limits to the maximum (110% and 91 °C, respectively). Once sync'd, they ran at a maximum of 1202MHz in-game, but dropped to 1189MHz almost immediately due to passing the 70 °C threshold. Many GeForce owners are aware of the higher 83 °C temperature limit, which can be adjusted using Afterburner as we did, but the lower limit actually cannot be adjusted. By the way, we run a mixed set of cards because we've found that it provides the best overall thermal control. Two open-air cards of this power level really begin to interfere with each other in regard to heat production. Of course, even with the right cards in place, you also need to manage your case cooling carefully. We used a liquid CPU cooler and multiple 140mm case fans to keep our system's temperature under control, but there's no getting around the fact that at full blast, it was pushing out a lot of heat, enough to make the testing room uncomfortable at times.
A final word on power consumption before we jump into benchmarks. Throughout most of our testing, we found our single-card system using around 375W, while our SLI-equipped system used just over 550W on average. In our 3DMark Fire Strike Combined stress test, power use approached 700W in SLI, due to fully loading the CPU and both GPUs. None of these numbers are too shocking, and even a high-quality 850W power supply could handle the power draw without issue (our 1000W Platinum-rated unit didn't even spin up its fan a single time during our testing!). But do keep in mind that SLI requires not just an investment in video cards, but the proper power supply, and to a certain extent, the right motherboard as well. We'll say more about that later on.
As always, we start with baseline data from 3DMark, which is useful to confirm that SLI is working, and also provides some goalposts for what to expect in terms of SLI scaling.
We used 3DMark Extreme, which runs at 2560 x 1440, as well as 3DMark Ultra, which runs at 4K. At 1440p, we achieved 88% scaling of the Graphics Score, which is the most important number to look at when considering GPU power. At 4K, the scaling improves to 90%. Both of these are quite good, but don't quite reach the ideal 100% mark. Perhaps there is some platform bottlenecking occurring, or alternatively we may be seeing a limitation of the SLI drivers. Whatever the case is, this test establishes that we might see around 90% scaling in some of our game benchmarks... if we're lucky!