AMD R9 Fury X & CrossFire vs. GTX 980 Ti Gaming Performance Benchmark

The Fury X had curious performance, and for this reason, we urge you to pay close attention to the disparity between the 980 Ti and Fury X as resolution gets lower. This disparity is rooted in a limitation on the Fury X that we will explain as it is revealed.

The below benchmarks show AMD's Radeon R9 Fury X vs. the GTX 980 Ti. Further testing pits the Fury X in 2-way CrossFire against the 980 Ti in SLI. We performed overclocking tests further down.

Note that we presently have two reference GTX 980 Ti cards and one aftermarket GTX 980 Ti Hybrid, which has a bolstered core clock that is more representative of what most board partner cards offer. In our review of the GTX 980 Ti Hybrid, we noted that the card often outperforms the Titan X due to its higher raw frequency (both 'stock'), since 12GB of VRAM is never utilized in current games. For this reason, we've shown both the reference 980 Ti and EVGA's version on the charts.

3DMark FireStrike – GTX 980 Ti Hybrid, 980 Ti, Fury X, & More

R9 Fury X Benchmark: The Witcher 3's Widening Disparity

The Witcher 3 best illustrates the Fury X's initially confusing results. At 4K resolutions, the Fury X closely competes with the GTX 980 Ti with a ~12% gap. To be fair, neither card makes Witcher 3 playable at 4K / Ultra, but we're just stress-testing here. As resolution is lowered, the gap widens considerably, even doubling to a 24% disparity at 1080p. This is shown even better with SLI and CrossFire, where we see an 8% disparity at 4K widen to a 26% disparity.

After some analysis, we believe that the Fury X loses ground not because of ROPs – which would be difference-making at higher resolutions – but because of geometry. AMD's architecture loses favor with geometric complexity and tessellated game elements, and as the games reduce raster operation load (as is done at lower resolutions – fewer pixels shoved into ROPs), the Fury X loses ground rapidly.

Note: We experienced dramatic black flickering in The Witcher 3 when running CrossFire. It is deemed “unplayable” for these purposes, as we would resort to a single-card solution rather than suffer the visual detriment.

R9 Fury X Benchmark: Far Cry 4

Far Cry 4 exhibits similar behavior. The 980 Ti (reference) performs behind the Fury X marginally at 4K, with the SLI and CrossFire configurations tying in average. The 980 Ti SLI wins-out barely in 1% low performance, but is otherwise nearly tied with Fury X cards in CrossFire.

After dropping to 1440p, the GTX 980 Ti reference card jumps ahead by 15%, further illustrating the limitations of the Fury X.

R9 Fury X Benchmark: Metro Last Light

Metro: Last Light is the first game in the list to exhibit the Fury X's odd framerate capping. This is something we encounter with GRID: Autosport in the next charts.

Notice that, up until 1080p is tested, the CrossFire Fury X cards lead the SLI GTX 980 Ti reference cards. At 1080p, we see the two-way GTX 980 Ti cards pull ahead measurably. To prove a point that the Fury X was somehow performance-capped where the 980 Ti was not, we tested Metro: Last Light at 1680x1050 and found a similar 115FPS average. Our next thoughts were that the Fury X may be throttling the clockrate to reduce power consumption, but logging validated that the clock pushed its full 1050MHz with both 1680x1050 and 4K resolutions.

The single Fury X runs behind the GTX 980 Ti reference at all levels, but only marginally so at 4K and 1440p.

R9 Fury X Benchmark: GRID Autosport

GRID: Autosport exhibits a CPU or other hardware limitation when operating at 1080p, evidenced by equalized performance for all of nVidia's top-end cards. The Fury X and Fury X in CrossFire both sit capped at 109FPS average for 1080p. At 4K, the CrossFire Fury X pulls ahead immensely, but is immediately trounced at 1440 for the same cap as experienced at 1080p.

The single Fury X card (without CrossFire) is defeated somewhat easily by the GTX 980 Ti reference, which holds a 15% advantage. The GTX 980 Ti Hybrid – effectively an overclocked 980 Ti, just with an Asetek CLC – holds a 23.5% advantage. This stock OC is easily achieved on every GTX 980 Ti we've worked with and is representative of non-reference performance.

We're unsure of why there'd be such a massive performance disparity for CrossFire vs. SLI compared to single card combat. This is also almost exclusively present with GRID, which has proven to favor some configurations more heavily than other games in the past (accentuated by relatively high performance of the game at 4K).

R9 Fury X Benchmark: Shadow of Mordor

The GTX 980 Ti holds higher performance metrics across the board for Shadow of Mordor, sans raw FPS against the CrossFire config. Unfortunately for AMD, though, the 0.1% lows are abysmal at 4K and suffer slightly at 1440p. This produces more choppiness and noticeable 'gaps' in fluid gameplay on the Fury X, something we don't see much of on the GTX 980 Ti. The Fury struggles particularly hard at 4K, dropping to 18FPS 0.1% lows.

Despite its 1FPS performance gain in averages, the Fury X is the worse card of the two primary contenders for Mordor. Its lows are spiky enough to hinder fluidity of gameplay in a fashion that is noticeable to gamers.

R9 Fury X Benchmark: GTA V

GTA V at Ultra with all “Advanced Graphics” settings enabled (4K) maximally demands 8.3GB of VRAM. This is where we'd potentially see throttling on the Fury X if the HBM isn't fast enough to compensate for lower capacity.

With or without Advanced Graphics enabled, none of these GPUs are particularly capable of playing GTA V at 4K. Stress testing, though, we see that the Fury X has more difficulty keeping up as memory demand is increased. With 4K Ultra and no Advanced Graphics, the Fury X remains more competitive with the GTX 980 Ti.

Overclocking the R9 Fury X vs. 980 Ti & Hybrid – Results Table

Overclocking the Fury X is limited right now. AMD's official tools do not presently allow unlocked memory overclocking, but MSI Afterburner can cheat the OC. Using Afterburner's advanced settings, we can exceed manufacturer limitations to force a memory overclock that is reflected in GPU-Z and Afterburner logging.

Overclocks were stepped incrementally to determine the maximum operating frequency of the core clock and VRAM. We configure the power percent target to its maximum value before adjusting voltage to its own max setting. The 980 Ti allows 110% of base power to be supplied to the GPU for overclocking and seems to hover around 1.224V after overvolting; the R9 Fury X allows 150% of base power, but presently offers no overvolting or official memory OC. We then slowly increment clockrate, observing for visual artifacting or catastrophic failures throughout the process. Each increment is left only for a few minutes before moving to the next step. We're eventually confronted with a driver failure, at which point the clockrate is backed-down and then endurance tested for 25-minutes using 3DMark Firestrike Extreme on loop.

The above shows our AMD overclock stepping. We got stuck at 1110MHz (+60MHz OC) core and 560MHz memory. We tried trading between the two a few times, but were never able to reliably exceed these numbers. The max core overclock was a 5.5% gain.

Below shows our overclocking results for the GTX 980 Ti and GTX 980 Ti Hybrid. 30% gains are seen in reference overclocking (1075 vs. 1450MHz boost); the Hybrid sees a jump to 1514MHz.

GTX 980 Ti Hybrid Overclocking Table

GTX 980 Ti Reference Overclocking Table

R9 Fury X Overclock vs. 980 Ti, 980 Benchmarks

And now for the benchmarks:

As seen, there's generally no worthwhile gain from overclocking the Fury X. The 980 Ti shows large, game-impacting gains from its overclocks (52 to 62FPS average, for instance, in Metro with bolstered 99 and 99.9 percentile performance).

As we did with the 300 series, we've decided that it simply isn't worthwhile to overclock the Fury X. AMD has pushed it as far as it can tolerate as is.