With all of the focus on AMD’s Fury and its HBM, there isn't much attention being paid to the more affordable Radeon R9 390. Is it possible that we missed a gem?
Back in June, we tested most of AMD’s 300-series cards. We were sent samples of the R9 390X, R9 380 and R7 370, but we didn’t have an R9 390. Sapphire helped fill that gap for us by sending its Nitro R9 390 8G D5, an overclocked Radeon R9 390.
Much like the rest of AMD’s Radeon 300 series, the R9 390 isn’t based on a new GPU. It is a refresh of the R9 290 with some purported manufacturing refinements and configuration changes. The processor formerly known as Hawaii is built on the previous generation's 28nm node, featuring 2560 shaders running at 1010MHz (in this overclocked trim), along with 64 ROP and 160 texture units.
AMD sticks with an aggregate 512-bit memory bus, though the capacity of this card is double most other R9 290 variants. Sporting 8GB of GDDR5, the R9 390 is the most affordable option to offer that much memory. Of course, it's only worth chasing a card with 8GB if the board is fast enough to drive the high resolutions and detail settings capable of pushing beyond the bounds of 4GB.
The Radeon R9 390 is based on AMD’s GCN architecture, which is shared by all of the company’s current graphics products. It supports the same features as the other Hawaii-based cards, such as DirectX 12 in Windows 10 and TrueAudio to accelerate optimized games. The R9 390 supports AMD FreeSync and resolutions up to Ultra HD. These cards also support AMD’s LiquidVR, which enables features like asynchronous shaders and direct-to-display VR HMD support.
At first glance, Sapphire’s Nitro R9 390 8G D5 looks similar to the Fury Tri-X I reviewed recently. It features a similar shape and size. In fact, it also employs a Tri-X cooler without the added colors. The Nitro edition card features a much subtler black color scheme with grey and silver accents, which will be easier to match with other components.
The card's heat sink extends nearly an inch past the back of the PCB, and features five copper heat pipes passing through vertical aluminum fins. Two of the heat pipes are 6mm, two are 8mm and the one going through the center out the back is 10mm. Each pipe starts at the copper contact plate and comes out of the fins under the central fan. The two 6mm pipes loop back through the fins over the GPU, while the remaining three carry on to the rear section of fins. Sapphire claims this setup creates up to 300W of cooling capacity.
To cool off the fins and the heat pipes, Sapphire includes three 90mm dual ball-bearing fans. The company says its dual bearing design is made to keep dust out and ensure a long lifetime. Sapphire also implements Intelligent Fan Control II, which allows individual fans to stop spinning, cutting down on noise when extra airflow isn't needed.
Unlike the Fury that Sapphire sent us, the R9 390's PCB is rather long. Not counting its bracket, the circuit board measures 10.5 inches (267mm) long. When you count the shroud's full length, though, it's 12 inches (395mm) long without the bracket. Or, add 15mm to count the bracket. At the thickest point (from shroud to rear screw), the R9 390 Nitro measures 45mm. It would be a tight squeeze to fit two of these side by side.
The board itself comes equipped with 16K-hour capacitors that use high-polymer aluminum. According to Sapphire, these caps offer superior lifespan over standard aluminum caps. Sapphire also uses what it calls Black Diamond Chokes, which the company claims are 10% cooler and 25% more power efficient than normal chokes.
Along the top edge of the card, you’ll find a pair of eight-pin power connectors. AMD did away with its CrossFire connectors, instead sending multi-GPU data across the PCIe bus. Instead, you'll find a BIOS selection switch along the top edge.
By default, the card comes with its legacy BIOS enabled, but pressing the button switches it to the digitally-signed UEFI. In that mode, boot and resume times should be quicker.
You get several video output options. Similar to the way Nvidia's newest cards are arranged, Sapphire’s Radeon R9 390 Nitro Overclocked exposes three DisplayPort connections, a single DVI-I port and HDMI. The card can output through as many as four of its connectors simultaneously.
Sapphire’s packaging keeps the card safe while still showing it off through a small plastic window in the box. Because of this, the card isn't in an anti-static bag. It is, however, contained in soft foam on all sides, protecting it from impact. Inside the box, Sapphire includes an HDMI cable along with a driver disc and case badge. You’ll also find a quick-start guide and a pamphlet about Sapphire products.
How We Test
In order for us to compare results between cards reviewed by any of our reviewers, Tom’s Hardware standardized the test bench that we all use for 2015. By doing this, each of our locations can have different sets of cards and still compare from the pool of results that any of us have obtained. Starting with an MSI X99S XPower motherboard, each test bench has an Intel Core i7-5930K overclocked to 4.2GHz, 16GB of Crucial Balistix Sport DDR4, two Crucial 500GB SSDs and an 850W power supply from be quiet!.
Software and Drivers
In the recent Sapphire Fury Tri-X article, I received a lot critical feedback about the 15.15 driver. When the card was supplied to me for testing, I only had a limited time with it, and sadly that sample had to be returned after a few days (before AMD released its latest driver). For this evaluation, all tests run on the R9 390 Nitro Overclocked are performed using AMD's 15.7 Catalyst release. An R9 380 is thrown in for comparison at 1080p, if only for a bit of performance perspective at different price points. That board was tested using the 15.7 Catalyst driver as well.
I also have a comparison to the Fury. Its results come from our prior review, so the scores were generated using 15.15. Igor did manage some spot tests that indicated 15.7 had no meaningful impact on frame rates, so our numbers should reflect Fiji's performance accurately. The results from the 290X are also pulled from the Fury Tri-X review, which were actually carried over from the Fury X review from May.
Results from the GTX 970 SC were run on the latest WHQL Nvidia driver, 353.30, as we did not have previous results for this card on this test bench. The GTX 980 results were pulled from our initial GTX 980 Ti review, and as a result use driver version 347.25 (except for GTA V, which was tested using 352.90 beta).
|Direct X||DirectX 11|
|Graphics Drivers||GeForce GTX 980 Ti: Nvidia 352.90 Beta Driver|
All GeForce Cards in Grand Theft Auto V and The Witcher 3: Wild Hunt: Nvidia 352.90 Beta Driver
GeForce GTX Titan X, 980, and 780 Ti in all other games: Nvidia 347.25 Beta Driver
Radeon R9 290 X: AMD Catalyst 15.5 Beta
Radeon R9 Fury X and Fury: AMD Catalyst 15.15
|Middle-earth: Shadow of Mordor||Built-in benchmark, 40-sec Fraps, Ultra preset|
|Battlefield 4||Custom THG Benchmark, 100-sec Fraps, Ultra preset|
|Metro Last Light||Built-in benchmark, 145-sec Fraps, Very High preset, 16x AF, Normal motion blur|
|Tomb Raider||Version 1.01.748.0, Custom THG Benchmark, 40-sec Fraps, Ultimate preset|
|Far Cry 4||Version 1.9.0, Custom THG benchmark, 60-sec Fraps, Ultra preset|
|Grand Theft Auto V||Build 350, Online 1.26, In-game benchmark sequence #5, 110-sec Fraps, FXAA: On, MSAA: 2x, Texture Quality: Very High, Shader Quality: Very High, Shadow Quality: High, Reflection Quality: Very High, Water Quality: High, Particles Quality: Very High, Grass Quality: High, Soft Shadows: Softer, Post FX: Very High, Anisotropic Filtering: 16x|
To measure sound levels, I took readings with a handheld dB meter positioned two inches from the card's I/O bracket. The meter I have at my disposal is only sensitive to 35 dB, so anything lower than that registers as zero. If the card makes an audible noise that the meter does not detect, I mark it as 34 dB. If no audible noise is observed, then it receives a 0 dB result on the graph.
To test power consumption using our reference platform, a bit of creative math is needed. Since Haswell-E processors don’t have integrated GPU cores, we can’t boot the system without a discrete board installed to get a baseline. We are able to estimate consumption based on the approximate power draw of the test bench, though. In our observations, we’ve found that the approximate power draw from everything other than the GPU is 120W. By deducting that from the recorded wattage reported on our in-line power meter, we can calculate the approximate draw of the GPU.
Sapphire says this card is a great value for 1080p gaming, and it is. But at this resolution, the R9 390 trails our 970 until it is overclocked. At higher resolutions, the tables turn.
Using Ultra settings at QHD, it outpaces the R9 290X by a small but consistent margin. The same can be said for the GTX 780 Ti. Remarkably, it manages to maintain an average frame rate only 7 FPS behind the more expensive Fury Tri-X. With a bit of overclocking, that gap naturally shrinks.
At 1440p, our GeForce GTX 970 holds a small lead, while at 4K the R9 390 starts to pull away, keeping its nose above 22 FPS. That's the same minimum observed on Nvidia’s GeForce GTX 980.
The average frame rate at this resolution is too low to play at these settings. This holds true for every card except the Fury Tri-X, and even that experience isn’t what gamers are going to want. Overclocking doesn't help much either.
Far Cry 4
In Far Cry 4, this card really shines. The R9 390 performs similarly to the pricier R9 Fury at 1080p.
At 1440p using Ultra settings, the R9 390 averages over 62 FPS and never dips below 46. The minimum frame rate mirrors the R9 290X, as well as both Maxwell-based GPUs. But the 390 manages a significantly higher average than all of the above. It appears the massive 8GB of GDDR5 may come into play in this test.
The same can be said at 4K. Incredibly, the R9 390 sustains an average frame rate above 34, and only ever dips to 28. While these are not the numbers most gamers are looking for, the comparison between this card and the Fury is notable. The 390 trails by an average of 5 FPS, but its minimum is only 3 FPS lower. After overclocking, the gap is even smaller.
Grand Theft Auto V
Grand Theft Auto V is one of the most demanding games out right now. It’s easy to gobble up more than 4GB of memory in this game, but that would make comparisons with other cards less meaningful.
Even still, as you can see from the graphs, the settings we're using are probably ideal for gaming at QHD with today’s hardware. We average shy of 60 FPS and never dip below 44. Overclocking manages a small 3 FPS increase on the average frame rate, which is not enough to notice.
At 1080p, the R9 390 does well, averaging slightly above the R9 Fury, though its minimum frame rate drops much lower than the Fury and GeForce GTX 970.
Cranking up the resolution to 4K has the expected effect. Frame rates are cut in half across the board. The game dips as low as 22 FPS, and never reaches higher than 38. Overclocking boosts the frame rate ever so slightly, but not enough to be meaningful.
The R9 390 again performs relatively close to the Fury Tri-X in this game.
Metro: Last Light
At 1440p, Metro: Last Light behaves interestingly. If you look at the graph, almost all of the cards hit the same minimum frame rates. The averages vary from card to card, but not by a large margin. From the 290X, which achieves the lowest average, to the R9 390, which only trails the Fury, there is less than a 7 FPS difference.
Our demanding test settings make 1080p a better option in Metro. Minimum frame rates on our unit under test never drop below 69 FPS, equal to the GeForce GTX 970, though its average is more than 6 FPS higher.
Strangely, the tables turn when we test at 4K. The oldest card on the list, the GTX 780 Ti, somehow manages to average 3 FPS higher than the R9 390 before overclocking. Even after bumping the clocks up it trails by a small margin. Each card performs similarly though, with a spread of less than 5 FPS across the board (except for the Fury, which pulls ahead by nearly 20%).
Middle-earth: Shadow of Mordor
Middle-earth: Shadow of Mordor runs well at this resolution, averaging more than 60 frames per second and never dropping below 50. The results between the AMD cards were so similar they would fall within this title's margin of error. Overclocking results in a 4 FPS boost over the 390's stock clock rates, and the Fury is a little more than 10% faster than that.
Full HD testing reveals a sizable gap between the GTX 970 and R9 390. The overclocked 390 nips at the Fury's heels at this resolution, too.
At 4K, the R9 390 never drops lower than 30 FPS, and maintains an average above 37. Nvidia's GTX 980 trails right behind, while the rest of the field isn't capable of keeping up with these settings. After overclocking, the 390 is only 2.5 frames slower than the Fury on average, and it stays above 33 FPS at all times.
Tomb Raider is the oldest game in our suite, but it stays relevant through its demanding settings. Before overclocking, the R9 390 averages just shy of 65 FPS, and it only dips as low as 34. Nvidia’s GTX 980 sustains a higher average frame rate, but overclocking helps Sapphire's R9 390 Nitro close the gap. AMD's Radeon R9 Fury maintains its lead, though the delta is less than 5%.
Sapphire’s R9 390 trails the GTX 970 slightly at 1080p on average. After overclocking the card, it achieves the same average frame rate as the Fury.
4K is incredibly grueling in this game. Not one of the cards we tested managed to stay within playable frame rates through our benchmark. With that said, the gap between the Fury and the Radeon R9 390 Nitro is small.
Sapphire’s Radeon R9 390 Nitro comes overclocked from the factory, but of course we want to try pushing it further. No overclock is guaranteed though, and your results with the same hardware may differ from what we're seeing.
We used MSI's Afterburner software to overclock Sapphire’s Radeon R9 390 Nitro. The utility is compatible with most modern graphics cards, and many overclocking applications are simply derivatives of Afterburner.
Using the stock clock rate of 1010MHz, I ran 3DMark Firestrike to establish a baseline and recorded a score of 12,085. I then increased the GPU clock by 10MHz and re-ran the test. After each successful pass, the score was compared to ensure it kept climbing. This process was repeated until the first artifacts were observed, which happened at 1110MHz. A quick test at 1105 resulted in instability. Even a voltage bump of +6mV didn’t help mitigate the artifacts.
At that point, I settled on 1100MHz and moved on to the memory. Going with 25MHz increments from 1500MHz, the display started showing artifacts at 1550MHz. Due to my issues increasing the GPU clock, I backed it down to 1090MHz and tried tuning the memory again. This worked well; the memory subsequently scaled up to 1700MHz before presenting artifacts.
The high memory speed proved to be unstable, though. The 3DMark scores started to drop slightly after passing 1625MHz, and real-world games ran into issues at higher frequencies.
Ultimately, with a GPU clock of 1090MHz and a memory clock of 1625MHz (6500 MT/s), both settings were up nearly 8%. The extra processing power helped bring the 3DMark score to 12,901, which is over 800 points higher than the baseline. The overclock managed a moderate increase in performance in all games tested as well.
Noise, Temperature & Power
The graph represents measurements taken two inches from the card's I/O bracket. You might consider these numbers to be rather quiet, but to my ears they seemed much louder. My suspicions were confirmed when I moved the meter next to the fans instead. Two inches away from them, my instrument returned a much different result.
Directly behind the card, the meter registered 43 dB(A). That's not particularly loud, but this isn't the quietest card we've tested, either. The reading taken in front of the fans is quite a bit louder, registering 51 dB(A). That's after 10 minutes of game play; at idle, the card is completely silent since its fans don't spin at all.
Temperature measurements are taken over 10 minutes using GPU-Z in a 25-degree room. As you can see, the Tri-X cooler mounted to the R9 390 keeps AMD's Hawaii GPU cooler than the R9 Fury, despite having fewer heat pipes than the Fury.
Sapphire’s R9 390 Nitro peaked at 70 degrees after almost eight minutes in Battlefield 4. This was in an open test bench with no fans, though. In a case with proper circulation, the temperature probably wouldn't reach as high.
The power consumption test that I perform takes measurements at three different times. Power draw is measured while the system is idle, at the end of a 10-minute Battlefield 4 run before quitting and while FurMark is being run.
Sapphire’s Radeon R9 390 Nitro is far from an energy efficient graphics card. At idle, it pulls 16.7W, even though its fans aren’t usually spinning. Gaming is where you'll really see the big power numbers though. It draws as much as 255W in Battlefield 4, which is nearly 85W more than the similarly-priced GTX 970 and 70W more than the comparatively powerful GTX 980. For some reason, FurMark affects AMD's GPUs more than Nvidia's. The R9 390 Nitro needs as much as 323.3W in this test.
After recently reviewing Sapphire’s R9 Fury Tri-X, I’m not sure what I was expecting out of the R9 390 Nitro. This card comes in at a much lower price point, yet offers more GDDR5 memory. Needless to say, I was eager to see what it could do.
After running my tests, I realized I underestimated what a 390 could do. Priced to compete with Nvidia’s GeForce GTX 970, I expected to see approximately the same level of performance. But I was surprised at the numbers it actually achieved. The R9 290X, which is technically a higher-tier card, wasn’t even in the same ballpark at higher resolutions. The 390 walked all over my GTX 970 sample, and consistently traded blows with the GTX 980. In many cases, it was right on the heels of Sapphire’s much more expensive Radeon R9 Fury Tri-X.
Curiously, while running tests in 1080p, the card didn’t perform as consistently. At that resolution, the R9 390 is a perfect match for the GTX 970. But it really stretches its legs at 1440p. Sapphire gives this card a performance rating of 4, which it considers ideal for 1080p. I would say it’s better suited for delivering more pixels than that.
In the end, I was blown away by this card. I really like the Fury Tri-X, and I said so in my review. But had I tested the 390 Nitro first, I would have come to a different conclusion. Sapphire’s Radeon R9 390 Nitro is a tremendous value for the money, making it difficult to justify higher-priced boards. One of these is perfect for resolutions up to 2560x1440; you'd need two to approach 4K.
Considering that the performance gap between the GeForce GTX 970 and this Radeon R9 390 is bigger than the gap between the 390 and our Radeon R9 Fury, Sapphire's solution earns an easy recommendation.
Kevin Carbotte is an Associate Contributing Writer for Tom's Hardware, covering Graphics.