Conclusion; Competitive, But Is It Enough?
Conclusion; Competitive, But Is It Enough?
It’s been a long time since I’ve been so torn about a graphics card review. I went into this thinking the exact same thing that many of you who actually made it this far probably did: the R9 390X would be a waste of time, simply a rebrand without much going for it. On one hand, that’s true since it doesn’t bring any few features to the table but despite using an older architecture, this is still an excellent card.
At this point in time you might think I’ve either gone off the deep end or I might have started drinking AMD-flavored Kool-Aid but hear me out. With higher core clocks and an epic amount of memory bandwidth, the R9 390X is able to leave the reference R9 290X in its dust and even manages to outmuscle the GTX 980 in nearly every test.
All of this has been achieved without a massive increase in power consumption or even switching to a completely different architecture. To me, that shows AMD has achieved their goals while still living within their somewhat limited financial means. They just don’t have the money to introduce a completely new volume-focused architecture on a manufacturing process that will soon be phased out in favor of 14nm. The high-cost, low volume Fiji design is a perfect testing ground for AMD’s bold vision of the future and that’s what it is being used for.
From a raw numbers perspective it’s obvious that the R9 390X’s 8GB of 6Gbps memory pays dividends in certain situations, but only those which aren’t already bottlenecked by the processing core itself. Naturally, that leads to some impressive wins in 4K and a few at 1440P. If anything, the results show just how versatile this card can really be; great performance now and just enough resolution-proofing if that jump to 4K is somewhere in your future.
There will likely be some bones of contention to pick here as well. Our results against a reference R9 290X are a bit skewed since any of the countless custom Hawaii XT-based cards would provide better clock speed consistency and thus more competitive performance than our results show. The R9 290X in its most basic form is a bit embarrassing when you put it up against Sapphire’s titanic Tri-X. It’s just important to remember that other than a meager 5Mhz overclock on its core, the R9 390X Tri-X is indeed clocked at AMD’s default speeds.
While the R9 390X throws up an awesome price / performance bulwark against a stock GTX 980, it goes through an identity crisis when compared against similar R9 290X cards. For example, the Sapphire R9 290X Tri-X 8GB has a similar core clock and a 500MHz lower GDDR5 speed but it costs $40 less before generous rebates are factored into the equation. If we had that 290X here for comparison purposes, I guarantee I wouldn’t be able to find justification for the 390X’s premium. It just boggles the mind.
There are some other hiccups here as well. Actually getting the 15.15 driver to install on a fresh version of Windows 8.1 resulted in hours of frustration (post on our forums if you encounter any problems!) but after turning off Windows Update and running DDU, things finally turned around. In addition, despite its relatively good power consumption improvements over its predecessor, the R9 390X still lags far behind NVIDIA’s GTX 980 in the performance per watt department.
Regardless of whether you want to call this a rebrand or refresh (I’m firmly on the refresh side), the R9 390X is an undeniably appealing card for anyone who can’t justify spending over $450 for a GPU. It is truly amazing to see that a Hawaii-based derivative can be so competitive this far into its life. I’m just not sure if that represents a ringing endorsement for the versatility of AMD’s GCN 1.1 architecture or an honest critique about how the graphics performance yardsticks haven’t moved all that much in almost two years. Maybe it’s both.
