Today we're checking out the new-ish Radeon RX 590 and on hand for testing we have the XFX Radeon RX 590 'Fatboy'. This is technically a new GPU, but kind of not new at the same time. What we have here is Polaris 20 XT on the 12nm FinFET process, and this small change translates into slightly better clock speeds.
There are no additional stream processors, no fancy GDDR6 memory, or even GDDR5X memory for that matter. So in essence we're getting the same RX 580 with a 15% overclock, at least when comparing the reference clock speeds. However since most entry-level RX 580 models are clocked at 1380 to 1410 MHz, this gives the RX 590 a 10-12% clock frequency advantage out of the box. Then there's the higher-end models like the Gigabyte Aorus RX 580 XTR, which we use for testing, and that particular model comes clocked at 1425 MHz, reducing the RX 590's clock speed advantage to 8%.
It's important to clarify this context because for testing we felt it would be misleading to compare the RX 590 with a base model RX 580 using the AMD reference clocks. AMD has sullied this launch by pricing the RX 590 at $280, which is a considerably premium for this slightly overclocked RX 580. Meanwhile, the RX 580 comes in at around $230 (MSRP), and right now can be found for as little as $200.
AMD claims up to a 12% performance boost over a reference clocked RX 580, yet they think that means they can increase the MSRP by 22%, so this should be interesting, or upsetting depending on your position.
For testing we're using the XFX Radeon RX 590 Fatboy and it will be compared to the Gigabyte Aorus RX 580 XTR and MSI GeForce GTX 1060 Gaming X+. We were unable to test the usual big battery of GPUs for this review, so you'll notice the absense of more affordable models such as the RX 570 and 3GB GTX 1060, though we included as many results as we could. If you don't mind extrapolating however, previous tests will give you plenty of straight GPU comparisons if that's what you want (RX 590 excluded of course).
Our benchmarks were ran using a Core i7-8700K clocked at 5GHz and 32GB of Vengeance DDR4-3400 memory. In total we have results for 21 games at two resolutions, 1080p and 1440p but the commentary will be focused on 7 of the more recently released titles.
First up we have Battlefield V results (don't miss our recent RTX ray tracing test!) and here the XFX 590 was 7% faster than the Aorus 580 XTR, taking the average frame rate from 85 fps to 91 fps. This is the kind of margin we were expecting, though that difference is reduced to just 3% at 1440p. On the upside, the RX 590 was a full 20% faster than our GTX 1060 card.
Strange Brigade results at 1080p have the RX 590 performing 5% above the 580 and 16% faster than the GTX 1060. At 1440p we see the same 5% margin, here the 590 provided an additional 3 fps over the Aorus RX 580 XTR, not exactly exciting stuff is it?
The Assassin's Creed series has been a bit brutal for AMD recently and Odyssey is no different. Here the 580 and 590 are easily beaten by the GTX 1060 at 1080p. We find a similar story at 1440p. The RX 590 was 7% slower than the GTX 1060 and just 6% faster than the 580, which equates to two extra frames.
This is our first time benchmarking with Hitman 2... shame it's not a more exciting occasion. Here the RX 590 offered a 5% performance bump over the 580 and that meant it was 11% faster than the 1060 at 1080p. Then at 1440p the margin between the 580 and 590 is reduced to just 2fps, a 4% performance advantage going the way of the 590.
Another 5% performance gain is seen in Shadow of the Tomb Raider, rendering 63 fps on average opposed to the 580's 60 fps.
At 1440p we see just a single frame advantage going to the 590, that said it was 14% faster than the GTX 1060.
Forza Horizon 4 played really well on the RX 590 at 1080p as frame rates never dipped below 80fps. Oddly we did see a 10% performance boost at 1440p, I suspect with a reference clocked 580 you'd see more double digit gains and I'll look into this in a moment.
Of the 21 games we've tested the last one we're going to discuss is Monster Hunter World. At 1080p we see another unexciting 4% performance gain over the 580, though the 590 was 10% faster than the GTX 1060. Then at 1440p we see a single extra frame and I don't know about you, but that's about all the single digit gains I can handle in a review.
Overclocking the XFX Fatboy landed us a core clock speed of 1620 MHz with a GDDR5 memory frequency of 2250 MHz, resulting in a transfer speed of 9 Gbps. This boosted Battlefield V 1080p performance by 4% with similar gains seen at 1440p.
You will note that the Aorus RX 580 XTR is a power hungry 580 model, using ~10% more power than a more standard 580. With that said, we see XFX's RX 590 pushing total system consumption slightly higher, and that resulted in a 42% increase over the GTX 1060 configuration, which is no small difference.
Realistically, your power supply won't have an issue with this increase and you're not exactly going to notice it on the power bill either, unless you game 24/7. The only real issue is dealing with the extra heat that's dumped in your case. If you have a well ventilated case then this is less of an issue, but for budget builds this will heat things up.
Under load for an hour, the XFX Radeon RX 590 Fatboy peaked at 79 degrees which is reasonable, but also starting to get up there. The good news is at this temperature it was relatively silent as the fans were only spinning at up to 1380 RPM.
Overclocking actually reduced the temperature to 78 degrees as the fans were now spinning at up to 1600 RPM and here they could be heard over the case fans. The operating volume wasn't unacceptably loud and certainly nothing like a Vega reference cooled card.
Putting It All Together, Cost Per Frame
Coming in first place by a country mile is the Radeon RX 580. You can currently buy 8GB models for $200 which is a super purchase. Not that long ago gamers would have sacrificed a finger for a price like that on a mid-range graphics card.
The GTX 1060 is also down at the MSRP, though the 9Gbps models are likely going to cost a bit more. Honestly they are a gimmick for the most part, we never saw much of a performance uplift over the standard models. In any case, the RX 580 costs 20% less per frame so it's the obvious value choice. The RX 590 is far less appealing at $280, coming out at a cost of $4 per frame or 32% more expensive on a cost per frame basis. Instead we'd rather spend $100 more and get a GTX 1070 or the 1070 Ti.
Performance-wise the Radeon RX 590 was always going to be unexciting -- a mildly overclocked RX 580 is only going to take you so far -- but pricing is what's killed this offering. We're just hoping this is a situation where AMD introduce it at $280 then axe pricing down to a $250 street price before the year is over. At $250 it matches the GTX 1060 while offering superior performance. Given AMD's current position in the GPU market they need to offer compelling options and that's what a $250 RX 590 would be.
The RX 590 might consume more power and while not ideal, it makes up for that with a healthy ecosystem of affordable FreeSync monitors and an extra 2GB of VRAM. But as a mid-range value offering, its biggest problem is having to be sold alongside a $200 RX 580. And the biggest problem the RX 580 faces is the $150 RX 570, so it appears AMD is AMDying themselves...
Bottom line, the RX 590 is not a bad product, it's just got the wrong price sticker. If you're in the market for a new graphics card these holidays, there are plenty of GPUs on offer at attractive prices which is a welcome change from where we were a year ago.