Test Methodology

We tested using our updated 2015 GPU test bench, detailed in the table below. Our thanks to supporting hardware vendors for supplying some of the test components. Thanks to Jon Peddie Research for GTX 970 & R9 280X support.

The latest GeForce 353.06 drivers were used during testing. AMD Catalyst 15.15 was used for 300 series cards. Game settings were manually controlled for the DUT.

VRAM utilization was measured using in-game tools and then validated with MSI's Afterburner and AIDA64, a custom version of the Riva Tuner software. Parity checking was performed with GPU-Z. FPS measurements were taken using FRAPS and then analyzed in a spreadsheet.

Each game was tested for 30 seconds in an identical scenario on the two cards, then repeated for parity.

Average FPS, 1% low, and 0.1% low times are measured. We do not measure maximum or minimum FPS results as we consider these numbers to be pure outliers. Instead, we take an average of the lowest 1% of results (1% low) to show real-world, noticeable dips; we then take an average of the lowest 0.1% of results for severe spikes.

We conducted a large suite of real-world tests, logging VRAM consumption in most of them for comparative analysis. The games and software tested include:

• Far Cry 4 (Ultra 1080, Very High 1080).
• GRID: Autosport (Ultra 1440, Ultra 4K).
• Metro: Last Light (Very High + Very High tessellation at 1080; High / High at 1440).
• GTA V (Very High / Ultra at 1080p).
• Shadow of Mordor (Very High, 1080p).
• 3DMark Firestrike Benchmark
• GTA V
• The Witcher 3

We already know GTA V and Far Cry 4 consume massive amounts of video memory, often in excess of the 2GB limits of some cards. GRID: Autosport and Metro: Last Light provide highly-optimized benchmarking titles to ensure stability on the bench. Shadow of Mordor, GTA V, & The Witcher 3 are new enough that they heavily eat VRAM. 3DMark offers a synthetic benchmark that is predictable in its results, something of great importance in benchmarking.

Thermals were reported using Delta T over ambient throughout a 25-minute burn-in period using 3DMark FireStrike - Extreme on infinite loop, which renders graphics at 1440p resolution. This test loads the VRAM heavily, something Kombustor skips, and keeps the GPU under high load that is comparable to gaming demands. Temperatures were logged using AIDA64.

Critical Note: A Word on Driver Difficulties

No matter the benchmark results in the below charts, the R9 390 is not 4K-ready. Throughout our test period with the 300 series devices, the drivers have exhibited a number of blocking failures that diminish the quality of gameplay or obstruct it entirely. The 300 series launched with version 15.15 of AMD's Catalyst drivers, doing little to stabilize the previous two beta updates. The last stable driver update launched in December with Omega.

During testing, the R9 390 regularly produced black flickering on the display at all game resolutions, but they occurred most frequently at 4K. This flickering issue also presented itself at the desktop level and would occasionally lose the display signal entirely. Requiring a cable pull for detection. We've got more than twenty video cards on the bench, and this is the only one that presents such an issue. The black 'flickers' appear a minimum of twice per minute with an average 'hold' time of 5 seconds. The issue is more pronounced at 1440p and 4K. We have deemed this defect severely prohibitive to gameplay and would consider games “unplayable” while the problem persists.

The R9 380 and R9 390 both have issues gaming at 4K, even when framerates are somewhat reasonable (see - GRID: Autosport), regularly failing to draw textures as more than a black fill on the assigned geometry.

Installation was another issue. Our install of the 300 series drivers did not appear compatible with the 200 series, something we hope is fixed as users receive their cards. The 300 series drivers do not install atop the old drivers (and will present conflicting reports in the 'about' information as to the driver version) and will sometimes struggle with clean installs if Windows Update is enabled to fetch drivers. This is the nature of drivers that aren't WHQL verified, but the 300 series presented more installation difficulties than most other drivers we've recently worked with. Nuking the drivers with Driver Remover can create some issues with boards that utilize multiplexed PCI-e downstream switches, something we provided a fix guide for here.

AMD R9 390 & 380 Temperatures & Power Consumption

Our samples deploy Sapphire-designed Tri-X and Dual-X coolers, so performance will excel beyond what the old reference 200 series coolers were capable of. AMD is known for running hot, but Sapphire cools the chips impressively well – granted, the R9 390 Tri-X card is massive at 12.5” long.

Power draw is about what we'd expect. The *90 series draws in excess of 300W (total system load), *80 in the 350-360W range (total system load), and – for comparison – the GTX 960 draws around 200W, but is significantly lower powered in its raw framerate output.

AMD R9 390 & 380 vs. 290X, 280 – The Witcher 3 Benchmark

Results produced in the Witcher 3 were effectively unplayable at 4K, but that's not saying much – it's a hard game to run, as evidenced by the fact that a Titan X still doesn't enter the “playable” range of performance without fine-tuning the graphics settings. Moving to more common resolutions, 1440p gaming is still just out of reach of the R9 390 and R9 380, but 1080p is a different story. With Ultra settings, the R9 390 produces a 48FPS average output – approaching playable range and easily increased by lowering foliage viewing detail – but sustains a somewhat dismal 13FPS 0.1% low. 99- and 99.9 percentile performance suffers on the new card, shown in the way of somewhat choppy frametimes in the temporal charts.

The R9 380 would need to be dropped to medium settings for reasonable playability, where we saw an unlisted average of 56FPS, 1% low of 37FPS, and 0.1% low of 15FPS.

For the price, the R9 390 appears competitive with nearby alternatives – the GTX 970, mainly – but isn't something we can recommend until the black flickering sorts itself out. The R9 290X is available at roughly the same price that the R9 390 is intended to ship at and exhibited better driver stability. At $325, the 290X is still a reasonable buy for when nVidia is priced out of the build.

AMD R9 390 & R9 380 GTA V Benchmark

AMD R9 390 & 380 vs. 290X, 280 – GRID: Autosport Benchmark

AMD R9 390 & 380 vs. 290X, 280 – Metro: Last Light Benchmark

AMD R9 390 & 380 vs. 290X, 280, 960 – Far Cry 4 Benchmark

As with the above, no one's going to be playing Far Cry 4 at 4K with anything short of a >$500 graphics configuration. Dropping to 1080p, the R9 380 pushes a desirable ~60FPS with agreeable lows, flanked only by the 280X and 285. The GTX 960 and R9 270X fall slightly behind.

AMD R9 390 & 380 vs. 290X, 280, 960 – Shadow of Mordor Benchmark

Playing Shadow of Mordor at 1080p with ultra settings sees playability on the R9 380 and R9 390, sitting at 58FPS and 78FPS averages, respectively. The GTX 970 lands between the two, but would be a significantly better buy than the R9 390 while it has driver issues.

Conclusion: A Disappointing Re-Refresh

AMD's put themselves into a weird cycle of partial refreshes. Fury will be a breath of (hopefully) fresh air, but for now, there's little to be excited about for those who've followed the 7000 and R9 200 series launches. The R9 380 produces marginally better performance than our R9 280 – depending on the game, we're in the ~4% range favoring the 380 – and otherwise operates equivalently. We don't have an R9 290, but the core-advantaged 290X trades blows with the R9 390 and is priced almost identically after factoring rebates and other discounts found on Newegg.

From the specs and our own benchmarks, it would appear that owners of the R9 200 series have no incentive to migrate to the new R9 300 refresh. The R9 285 and R9 380 are priced and perform similarly, but if either one dips below the other in price, it's a reasonable mid-range buy for system builders trying to spend somewhere between 750 Ti and 970 “land” in terms of price. The GTX 960 ($200) is still a good card for its low TDP package, driver support, and software feature-set, but doesn't push the highest raw framerate for some of the harder-hitting modern games.

The R9 390 isn't a device we can recommend right now. Once the black flickering is resolved, it'd be a consideration, but I'd feel a lot more confident recommending that readers spend an extra few bucks toward a more driver-stable GTX 970. The R9 290 ($235 after rebate) is priced cheaper than the R9 390, though, and will perform within 3-6% of the newer card; we'd just buy that instead, really.

AMD's poor showing of driver support has made it difficult to recommend the company's cards even when they're able to push a reasonable framerate. The new devices are OK for the price, but not exciting. NVidia holds a clear and obvious day-1 driver advantage, showing a track record of game-ready, verified drivers on launch day where AMD has failed to compete. For builders who are comfortable waiting an extra day or two on game-ready drivers, the price decrease for AMD can be worthwhile – especially in the mid-range and entry-level markets. At the high-end, nVidia still reigns king with the 980 Ti and 980, though Fury could change things.

NVidia still consumes fewer Watts, no news there, but we were markedly impressed with Sapphire's cooling ability on the 300-series cards. A 45C Delta T over Ambient value is a substantial thermal improvement over our 290X reference card (67C) and talks AMD away from the thermal ledge a little bit. The Tri-X is a massive 12.5” long, but cools well and is worthy of note for its ability to dissipate 275W of heat.

AMD and its board partners are all driving the same angle: For gamers who don't care about additional features -- like CUDA-accelerated application support or a heavier focus on power savings and thermal reduction -- the R9 300 cards are AMD's go-to device. AMD wants to sell to users who care first and foremost about FPS, with little regard for TDP and temperature so long as it's cheap and games. The company needs to get its drivers in order before we can confidently recommend any 300-series products, but it is possible for them to achieve the angle they're targeting -- it's just going to require more joint software-hardware effort.

- Steve “Lelldorianx” Burke.