Asus STRIX Radeon R9 390X Review: Hawaii Gets 8GB
Our Test System And 3DMark Fire Strike Benchmarks
Our test system this time around is based on Intel’s X99 platform, built up with an Intel Core i7-5960x and 8GB of G.Skill DDR4 Ripjaws RAM running at 2400MHz, on top of an ASUS X99-Deluxe motherboard. All of our benchmark tools and games are launched from a Samsung 850 EVO SSD, and we’re running a fresh and fully updated version of Windows 8.1.
To ensure maximum performance, we loaded up clean installations of AMD’s Catalyst 15.7 and NVIDIA’s GeForce 353.30 drivers to retest the GTX 980, AMD’s reference Radeon R9 290X, and the card that’s in today’s spotlight, the Asus STRIX Radeon R9 390X.
Intel Core i7-5960X
Sapphire Tri-X R9 Fury
Radeon R9 390x
GeForce GTX 980
GeForce GTX 980 Ti
8GB G.Skill Ripjaws DDR4-2400
|Relevant Software: |
Windows 8.1 Pro x64
AMD Catalyst v15.7
NVIDIA GeForce Drivers v353.30
Unigine Heaven v4
3DMark "Fire Strike"
Middle Earth: Shadow of Mordor
Metro: Last Light Redux
Grand Theft Auto V
3DMark Fire Strike is a punishing synthetic benchmark that utilizes a number of graphics-related technologies, such as tessellation, ambient occlusion, physics, volume illumination, and depth of field. We’re using Fire Strike’s two most demanding modes: Extreme which targets 2560x1440, and Ultra which takes aim at 4K-capable systems (it can simulate the resolution even if you don’t have an Ultra HD monitor).
While no synthetic test can give us as accurate a performance picture as real gaming, Fire Strike remains a reliable gauge of a video card’s potential and a strong reflection of its place in the competitive landscape.
Running Fire Strike Extreme, we see the 390X with about an 8.5% lead over the 290X, and a 9.3% performance gap when ramping up to Fire Strike Ultra’s 4K benchmark. That slight edge in Ultra HD resolutions seems to be the norm for AMD’s Fury and 300 Series cards.
When we throw the competitive landscape into the picture, we’re seeing NVIDIA’s GTX 980 squarely beating the Radeon 390X. But here’s what’s interesting: This is one such case where synthetic testing isn’t going to be indicative of real-world performance. Read on...