Just when we thought that we’d seen everything from AMD this year, the company sprung the Radeon R9 380X on us, a more powerful version of the Radeon R9 380 with a greater number of stream processors, texture mapping units, and higher clock speeds.
This has been a busy year for AMD. After rolling out a series of long overdue graphics cards – including cards based on the brand new Fiji architecture and other rebadged products – and debuting the first 3D memory architecture in High Bandwidth Memory (HBM), the company is hoping to reassert its place in the GPU market. Following a restructuring of its graphics division, it also announced that it would be ditching its Catalyst brand of drivers, opting instead for a broad umbrella called “Radeon Software”.
With the Radeon R9 380X, AMD looks to be hoping to offer even more choice to mainstream consumers, who arguably comprise the largest section of the market. By using a “fully-enabled” version of the Antigua GPU (rebranded Tonga) that we saw in the Radeon R9 380, AMD has given us a slightly more capable card, intended to woo consumers who desire even more performance.
NVIDIA actually took the reverse tack with the GeForce GTX 950, which featured a cutdown version of the GM206 GPU used on the GeForce GTX 960. It sure seems like both AMD and NVIDIA are milking their different GPU chips for all they are worth, in an attempt to provide an option for every conceivable budget.
AMD hasn’t released a reference version of the card, so we’ll be using the ASUS Strix Radeon R9 380 to assess the performance characteristics of the card. Incidentally, we used an ASUS card for our review of the Radeon R9 380 as well, so the comparisons between the two should be relatively fair.
While the Radeon R9 380 utilized an Antigua chip based on the Radeon R9 285’s Tonga GPU, the new R9 380X will feature Antigua XT, indicative of a fully-enabled Antigua chip. Indeed, it is equipped with 2,048 stream processors, around 14% more than the 1,792 on the Radeon R9 380. The number of texture mapping units has also been bumped up to 128 from 112, a similar 14% increase.
As per most custom cards, the core clock has been factory overclocked to 1,030MHz, up from 990MHz on the Strix R9 380. This also represents a decent boost above the reference speeds of around 970MHz for a typical Radeon R9 380X class graphics card. A dedicated OC mode raises this to 1,050MHz for a slight performance boost. It also comes with 4GB of GDDR5 memory – unlike the R9 380 which came in 2GB and 4GB versions (we never knew why), this is the only flavor offered.
In addition, it has the same 256-bit memory bus width but the memory clock speeds have been bumped up to 1,425MHz, which translates into 182.4GB/s of available memory bandwidth, putting it just above what a reference Radeon R9 380 is specced for.
And despite its fully-enabled Antigua XT GPU, the Radeon R9 380X retains similar power characteristics (this is indeed the case in our power consumption tests) as the R9 380 with the same TDP of 190 watts.
Here’s a snapshot summarizing the card’s specifications:
When AMD announced its Radeon 300 series cards back in June, it singled out the Radeon R9 380 and above as being suited to gaming at 1440p resolutions. This time, AMD was more specific with the Radeon R9 380X’s placement in its product line-up – the R9 380X will be a 1440p gaming card, while the R9 390 and up will be targeted more at 4K gaming.
The ASUS Strix Radeon R9 380X features all the bells-and-whistles that we’ve come to expect from a custom card. It relies on ASUS’ familiar dual-fan DirectCU II cooler to keep cool, complete with a stylized cooling shroud that resembles the watchful eyes of some red-eyed owl.
There are also no surprises when it comes to the fans, which use a semi-passive design that allows them to power down entirely when the GPU’s temperature falls below the 55-57°C range. And considering that the GPU idles at around 38°C, you shouldn’t see the temperature rise above the specified range while mucking around with light tasks like Web browsing or Microsoft Word, which means that you get to work, read, and surf the Web in relative quiet.
The Strix Radeon R9 380X also uses a triple heatpipe design, comprising a single extra-thick 10mm heatpipe and dual 8mm ones to help channel heat away from the GPU die.
In addition, it sports an extra-dense array of heatsink fins to increase the area available for heat dissipation. ASUS puts the total heat dissipation surface area at 2,842.53cm2 , 1.5 times more than the 1,942.49cm2 available on the reference design. In theory, this should mean less thermal throttling because of the card’s ability to dissipate more heat.
The card relies on a digital 8-phase power design and upgraded capacitors, chokes, and voltage regulation components over the reference design. These supposedly help create a cleaner power signal and widen the voltage modulation tolerance, which looks like it might help with more stable overclocking. But as we’ll see later, the card doesn’t take too well to being overclocked, which is unsurprising given that overclocking headroom wasn’t impressive on the Radeon R9 390 series cards either.
As expected of custom cards today, there's also a metal backplate for further structural reinforcement (and it looks good too). At 273mm long, this is not exactly a small card, so the extra support is a welcome feature.
The card is powered by two 6-pin PCIe power connectors, a change from the single 8-pin connector on the Strix Radeon R9 380. As per the previous Strix card we’ve reviewed, there are two LEDs above the connectors that will glow white when a successful connection has been established. They will glow red if there’s a problem, so you’ll know if it is a loose connection that’s the culprit.
Round the back, the card has a fairly standard selection of display connectors, including one DVI-I port, one DVI-D port, one HDMI connector, and one DisplayPort output.