After the introduction of the Fiji-based Fury cards, there hasn’t been all that much talk about the new R9 300 series other than criticism directed towards their rumored use of previous-generation architectures. That’s a shame since these cards, for the time being at least, represent AMD’s best hope to compete against the GTX 980, GTX 970 and GTX 960 while Fury series targets substantially higher price points. In addition, the high end market where AMD’s flagships play in can never hope to match the volumes generated by lower priced products. Simply put, the Radeon lineup needs a good blend of enthusiast and regular mid-tier offerings if it hopes to make inroads against GeForce’s slice of the GPU market pie.
Headlining the R9 300 series is the R9 390X, a card picks up from where the R9 290X left off. This is also where the cries of “rebrand!” will likely be heard the loudest since its price / performance equation is supposed to compete against NVIDIA’s GTX 980, a card which boasts an arguably more efficient architectural design. Meanwhile, the 390X uses the same GCN 1.1-based architecture as the outgoing Hawaii-based cards but in an updated core which has been code named Grenada XT.
On the surface at least there’s absolutely nothing to distinguish Grenada from Hawaii but there’s supposedly more here than what first meets the eye. While the SIMD array, asynchronous compute engines and generalized compute units haven’t been touched, the 28nm manufacturing process used for these GPUs has come a long way since they were first introduced almost two years ago. The end result is higher overall efficiency and a general maturation of processing pathways. In theory this could lead to Grenada-based cores which reach higher Boost speeds than their predecessors while also exhibiting slightly higher performance when identically clocked. We’ll go into this a bit later in the article.
From a features perspective the R9 390X has all the hallmarks of AMD’s latest architectures, including Fiji. That means Virtual Super Resolution (though limited to Hawaii levels), FreeSync support and Frame Rate Target Control have all been rolled into one neat package. There may be some differences in feature level DX12 support between GCN 1.1 and the Fury parts but those details haven’t been made public yet. Expect to hear more about that in a week or so.
Despite there being no core layout differences, there have been several tertiary modifications rolled into the Grenada XT-based R9 390X. The core operates about 5% faster than the reference R9 290X did which isn’t all that much but AMD has still managed to lower typical board power to 275W, down from about 300W on its predecessor.
Memory is one area where this new SKU exhibits serious benefits over the 290X. While Hawaii launched with 4GB of GDDR5 memory operating at 5Gbps Grenada XT will be paired with 8GB of 6Gbps, making it a veritable 4K powerhouse compared to similarly priced cards. Granted, it wasn’t all that hard to find R9 290X cards with an 8GB framebuffer but they were costly and their default memory clocks never hit the 6GHz mark.
One area where buyers will likely have to take a cold shower is in the pricing department. While 4GB R9 290X cards could be found for as little as $350 just a few months ago, the R9 390X will start at $429. This is actually right in line with a few of the 8GB Hawaii-based products still available but the few which are less (Sapphire’s Tri-X OC is just $389) tend to throw a wrench into the works. In addition, AMD hopes this price point will help their new card compete on a more level footing with NVIDIA’s GTX 980…at least on the price / performance front.
Rather than a rebrand or rebadge, the R9 390X is more like the end-of-life refreshes car models usually go through before they’re replaced in the next model year. There’s certainly not enough here to call the R9 390X a “new” card but branding consistency and a few very minor, almost transparent changes point towards this being something a bit more than AMD slapping a different name onto an old card and calling it a day.
Finally we come to the star of this particular review: the R9 390X itself. Sapphire’s R9 390X Tri-X OC is a blast from the past since, other than slightly faster memory, it looks and feels like a literal clone of their R9 290X 8GB Tri-X OC. Remember, that’s the one we mentioned above which goes for $40 less than their new card.
Other than the obvious similarities, it receives a paltry 5MHz overclock from AMD’s reference clocks. With a custom cooled card it may be impossible to find out exactly what “baseline” R9 390X performance is like but consider this: Sapphire is selling this thing for $429. That’s not a dime more than AMD’s SRP for reference-based examples.
This is one good looking card. The Tri-X has an extensive, triple-section heatsink which uses a stunning black / orange color scheme but with a length of 11.75”, it shouldn’t have too many issues fitting into most standard issue ATX cases. Underneath that heatsink is a custom designed 6-phase PWM alongside a dual BIOS button.
With its three 80mm cooling fans, an extensive aluminum fin array and a vapor chambered contact plate the cooler itself is an impressive piece of engineering. The build quality here is immaculate even though the shroud is manufactured out of plastic. Sapphire hasn’t seen the need to place a secondary heatsink over their memory modules, though the VRMs do get some attention with the inclusion of a form-fitting aluminum plate that’s actively cooled by the Tri-X fans.
The PCB’s back is left bare though we can see how large the cooler is; it exceeds the PCB by a good 1.5”. Without any memory modules here there really isn't any point to add a backplate, though the added cleanliness would be nice.
In order to accommodate as much overclocking headroom as possible, Sapphire has added a dual 8-pin power connector layout. Meanwhile, the rear I/O port houses a trio of DisplayPort outputs as well as HDMI and DVI connectors.
Headlining the R9 300 series is the R9 390X, a card picks up from where the R9 290X left off. This is also where the cries of “rebrand!” will likely be heard the loudest since its price / performance equation is supposed to compete against NVIDIA’s GTX 980, a card which boasts an arguably more efficient architectural design. Meanwhile, the 390X uses the same GCN 1.1-based architecture as the outgoing Hawaii-based cards but in an updated core which has been code named Grenada XT.
On the surface at least there’s absolutely nothing to distinguish Grenada from Hawaii but there’s supposedly more here than what first meets the eye. While the SIMD array, asynchronous compute engines and generalized compute units haven’t been touched, the 28nm manufacturing process used for these GPUs has come a long way since they were first introduced almost two years ago. The end result is higher overall efficiency and a general maturation of processing pathways. In theory this could lead to Grenada-based cores which reach higher Boost speeds than their predecessors while also exhibiting slightly higher performance when identically clocked. We’ll go into this a bit later in the article.
From a features perspective the R9 390X has all the hallmarks of AMD’s latest architectures, including Fiji. That means Virtual Super Resolution (though limited to Hawaii levels), FreeSync support and Frame Rate Target Control have all been rolled into one neat package. There may be some differences in feature level DX12 support between GCN 1.1 and the Fury parts but those details haven’t been made public yet. Expect to hear more about that in a week or so.
Despite there being no core layout differences, there have been several tertiary modifications rolled into the Grenada XT-based R9 390X. The core operates about 5% faster than the reference R9 290X did which isn’t all that much but AMD has still managed to lower typical board power to 275W, down from about 300W on its predecessor.
Memory is one area where this new SKU exhibits serious benefits over the 290X. While Hawaii launched with 4GB of GDDR5 memory operating at 5Gbps Grenada XT will be paired with 8GB of 6Gbps, making it a veritable 4K powerhouse compared to similarly priced cards. Granted, it wasn’t all that hard to find R9 290X cards with an 8GB framebuffer but they were costly and their default memory clocks never hit the 6GHz mark.
One area where buyers will likely have to take a cold shower is in the pricing department. While 4GB R9 290X cards could be found for as little as $350 just a few months ago, the R9 390X will start at $429. This is actually right in line with a few of the 8GB Hawaii-based products still available but the few which are less (Sapphire’s Tri-X OC is just $389) tend to throw a wrench into the works. In addition, AMD hopes this price point will help their new card compete on a more level footing with NVIDIA’s GTX 980…at least on the price / performance front.
Rather than a rebrand or rebadge, the R9 390X is more like the end-of-life refreshes car models usually go through before they’re replaced in the next model year. There’s certainly not enough here to call the R9 390X a “new” card but branding consistency and a few very minor, almost transparent changes point towards this being something a bit more than AMD slapping a different name onto an old card and calling it a day.
Sapphire's R9 390X Tri-X OC
Finally we come to the star of this particular review: the R9 390X itself. Sapphire’s R9 390X Tri-X OC is a blast from the past since, other than slightly faster memory, it looks and feels like a literal clone of their R9 290X 8GB Tri-X OC. Remember, that’s the one we mentioned above which goes for $40 less than their new card.
Other than the obvious similarities, it receives a paltry 5MHz overclock from AMD’s reference clocks. With a custom cooled card it may be impossible to find out exactly what “baseline” R9 390X performance is like but consider this: Sapphire is selling this thing for $429. That’s not a dime more than AMD’s SRP for reference-based examples.
This is one good looking card. The Tri-X has an extensive, triple-section heatsink which uses a stunning black / orange color scheme but with a length of 11.75”, it shouldn’t have too many issues fitting into most standard issue ATX cases. Underneath that heatsink is a custom designed 6-phase PWM alongside a dual BIOS button.
With its three 80mm cooling fans, an extensive aluminum fin array and a vapor chambered contact plate the cooler itself is an impressive piece of engineering. The build quality here is immaculate even though the shroud is manufactured out of plastic. Sapphire hasn’t seen the need to place a secondary heatsink over their memory modules, though the VRMs do get some attention with the inclusion of a form-fitting aluminum plate that’s actively cooled by the Tri-X fans.
The PCB’s back is left bare though we can see how large the cooler is; it exceeds the PCB by a good 1.5”. Without any memory modules here there really isn't any point to add a backplate, though the added cleanliness would be nice.
In order to accommodate as much overclocking headroom as possible, Sapphire has added a dual 8-pin power connector layout. Meanwhile, the rear I/O port houses a trio of DisplayPort outputs as well as HDMI and DVI connectors.
Last edited:
