Launching today is AMD’s new Radeon RX 500 series. As we’re covering in our companion launch article, the RX 500 series is a refresh of Polaris, bringing about newer, faster SKUs based on the existing Polaris 10 and 11 GPUs. Also joining the family is a newer, smaller GPU, Polaris 12, which will be the basis of the Radeon RX 550. AMD is using an updated revision of Polaris for all of these products, so there are some minor clockspeed improvements and a new memory state that have been baked into the RX 500 family that is not present in the RX 400 family, which makes the new RX 500 parts a bit more interesting.

The first Radeon RX 500 cards out of the gate are the Radeon RX 580 and the Radeon 570, which we’re reviewing today. These SKUs are pretty straightforward: take the new Polaris 10 GPU revision, plug it into more powerful boards, turn up the clockspeeds a bit, and you have a new SKU. AMD hasn’t done anything wild here – the configurations haven’t changed, and in fact TBPs have gone up – so relative to the RX 480 and RX 470, at the end of the day it’s a set of slightly more powerful cards for the same price as before.

Because these are just minor performance improvements over the existing RX 480 and RX 470 cards, these newer cards replace the RX 400 cards in AMD’s product stack, but they aren’t intended as upgrades for existing owners. Instead they’re meant to tempt owners of cards like the R9 280 and R9 380 series who didn’t already upgrade to Polaris. For those owners who did hold off, their reward is a slightly more powerful upgrade option in 2017 than they would have gotten in 2016.

Otherwise for AMD, this is a chance to partially clean the slate for 2017. The RX 500 series’ job isn’t to radically alter the competitive landscape – that’s Vega’s job – but rather it’s to push out a bit more performance and help close the gap in the midrange market with NVIDIA, while giving partners something new for 2017. NVIDIA has seemingly already made their move for 2017 in the midrange market with the optional 9Gbps factory overclocked GTX 1060 SKUs, so AMD would appear to be setting the stage for what should be much the rest of the year. When the dust settled from the launch of AMD’s Polaris and NVIDIA’s Pascal architectures in 2016, NVIDIA generally prevailed, so this is AMD’s chance to rethink their gameplan and continue trying to grab market share from NVIDIA.

AMD Radeon RX Series Specification Comparison
	AMD Radeon RX 580 (8GB)	AMD Radeon RX 570	AMD Radeon RX 480 (8GB)	AMD Radeon RX 470
Stream Processors	2304 (36 CUs)	2048 (32 CUs)	2304 (36 CUs)	2048 (32 CUs)
Texture Units	144	128	144	128
ROPs	32	32	32	32
Base Clock	1257MHz	1168MHz	1120MHz	926MHz
Boost Clock	1340MHz	1244MHz	1266MHz	1206MHz
Memory Clock	8 Gbps GDDR5	7Gbps GDDR5	8 Gbps GDDR5	6.6Gbps GDDR5
Memory Bus Width	256-bit	256-bit	256-bit	256-bit
VRAM	8GB	4GB	8GB	4GB
Transistor Count	5.7B	5.7B	5.7B	5.7B
Typical Board Power	185W	150W	150W	120W
Manufacturing Process	GloFo 14nm	GloFo 14nm	GloFo 14nm	GloFo 14nm
Architecture	GCN 4	GCN 4	GCN 4	GCN 4
GPU	Polaris 10	Polaris 10	Polaris 10	Polaris 10
Launch Date	04/18/2017	04/18/2017	06/29/2016	08/04/2016
Launch Price	$229	$169	$239	$179

At the high end is AMD’s new midrange contender, the Radeon RX 580. Like the RX 480 before it, this is a fully enabled Polaris 10 GPU. Taking advantage of their manufacturing gains, AMD is bumping up the boost clock by 6%, from 1266MHz to 1340MHz. Meanwhile the base clock – which has proven somewhat arbitrary on RX 480 since it rarely throttles anywhere near that much – is increasing by 12%, from 1120MHz to 1257MHz. As we’ll see further in this review, expect the performance gains to closely mirror the boost clock changes.

Meanwhile the memory clock is not changing for the 8GB cards. AMD is holding fast at 8Gbps GDDR5 on a 256-bit memory bus. It should also be noted though that while the default configuration of the RX 580 is 8GB, like the RX 480, some 4GB cards are also expected to be produced.

Joining the RX 580 in today’s launch is the Radeon RX 570. Like its more powerful sibling, this is an enhanced version of its RX 400 series predecessor, the RX 470. We’re looking at the same cut-down Polaris 10 GPU with 32 of 36 CUs enabled, but again clockspeeds are increased. RX 570 goes from 1206Mhz to 1244MHz on the boost clock, a 3% gain, while the base clock is increased from 926MHz to 1168MHz, a gain of 26%.

One thing RX 570 gets that RX 580 does not is a memory speed bump. On RX 470 AMD set the SKU standard at the somewhat odd 6.6Gbps; for RX 570, this is now a full 7Gbps, for a 6% increase in memory bandwidth. Polaris 10 in general likes memory bandwidth, so as you can see in our companion RX 570 review, this works out well for the RX 570. The standard memory configuration here will be for 4GB of VRAM, however AMD has mentioned that we should expect to see some 8GB cards as well, though none of these are on the launch list they’ve provided.

Otherwise the big change here is on power consumption. RX 580 is a 185W card, while RX 570 starts at 150W. This is a 30-35W increase in TBPs over the RX 400 series, and given the expected prevalence of factory overclocked cards, the TBP of the average retail SKU is probably a bit higher still. Manufacturing improvements in the last year have allowed Polaris to clock higher and/or reduce power consumption slightly at a given clockspeed, however in AMD’s case they’ve opted to spend all of these gains (and then-some) on clockspeed improvements, hence the TBPs we’re seeing today.

While this is a natural consequence of cranking up Polaris’s clockspeeds, what’s not really being said by AMD is why. And while not putting word’s in AMD’s mouth, from an outside perspective it’s pretty easy to see what’s going on. Polaris greatly improved AMD’s energy efficiency, but then Pascal did much the same for NVIDIA. As a result while the RX 400 series cards delivered good performance, they weren’t very competitive with NVIDIA’s GTX 10 series in the realm of power consumption. Consequently AMD’s no longer trying to compete on power efficiency on the Polaris 10 series: they are going right to the bend in the power/performance curve (and quite possibly past it) to deliver the best performance per dollar that they can. Competing on price is ultimately how they can best grab more of NVIDIA’s market share.

Moving on, since these new SKUs are just higher clocked Polaris 10 configurations, this means that for retail cards partners are hitting the ground running with custom cards. In fact AMD isn’t even shipping a reference design. The cards we’ve been sampled for today and the cards you’ll see on retailer/e-tailer shelves today are all customized in some form or another. For many board partners, this means they can just take their existing factory overclocked RX 480 cards – which were built for higher TBP operation in the first place – and use them as the basis for their RX 580 and RX 570 cards. Meanwhile other vendors are pushing out new custom designs to handle the new TBPs and to balance board costs with AMD’s prices.

Both the RX 580 and RX 570 are launching at $10 below their RX 400 series counterparts’ launch prices from last summer. This puts the 8GB RX 580 at $229 and the RX 570 at $169. The odd man out is the RX 580 4GB, which is launching at the same $199 price as the RX 480 4GB. This keeps AMD spot-on the $199 sweet spot, while perhaps more importantly it does a better job of differentiating the card from the next card down, the $169 RX 570. The smaller $20 price gap between the 4GB RX 480 and RX 470 meant that the cards didn’t always stand apart in a useful manner, especially hurting the RX 470.

The competition for these new cards, like the RX 480/470 before them, are NVIDIA’s GeForce GTX 1060 series. The RX 580 goes squarely up against the GTX 1060 6GB, while the RX 570 is very close in price to the GTX 1060 3GB. NVIDIA card prices have come down in recent weeks – when I took inventory last month, $229 GTX 1060s were not nearly as common as they are right now – so it would appear that NVIDIA and its partners have trimmed prices a bit in anticipation of the RX 580 and RX 570. AMD can’t compete with NVIDIA on power efficiency here, but if performance improves enough, they can certainly compete on performance-per-dollar. Meanwhile NVIDIA’s latest bundle is ending today, so we’re looking at a straightforward match-up in terms of pricing/value has been extended into next week rather than ending today, so NVIDIA is throwing in a bit extra to try to tilt the value proposition in their favor.

Finally, AMD’s partners and retailers are in the process of unloading their existing RX 480 and RX 470 cards. As a result these cards are going for a relatively good deal, but like past inventory clean-outs, this won’t last. However I will be curious to see where the dust settles here in a month or so when the clean-out is done; if AMD’s prices on the RX 500 series hold steady, then it would mean that their price/performance ratio will not have drifted by too much in the last 10 months since the RX 400 launch.

Spring 2017 GPU Pricing Comparison
AMD	Price	NVIDIA
Radeon RX 580 (8GB)	$229	GeForce GTX 1060 (6GB)
Radeon RX 580 (4GB)	$199
Radeon RX 570	$169/$179	GeForce GTX 1060 3GB
	$129	GeForce GTX 1050 Ti
Radeon RX 460	$99/$104	GeForce GTX 1050

Polaris Refined: Better Fab Yields & a New Memory State

One of the components of AMD’s marketing angle for the Radeon RX 500 series is that it’s Polaris Refined. These are still the same Polaris 10 GPUs as in the Radeon RX 400 series, but AMD wants to highlight the small improvements that they have made/gained in the past year. For RX 400 owners this doesn’t amount to much – your cards are still shiny and chrome – but it serves to help differentiate the new cards from the old cards. And for the owners of cards like the R9 280 and 380 series that AMD is trying to reach and convince them to upgrade, it’s the justification for why AMD thinks they should want to upgrade now after having passed on the RX 400 series.

There are two elements to Polaris Refined: silicon improvements, and a new memory clock state. The former in turn is comprised of both the benefits in improving fab yields and quality that AMD has enjoyed over the past year, and a new revision of Polaris 10.

In the case of fab yields, all of the revised Polaris chips are being manufactured on what AMD is calling the “Latest Generation FinFET 14” process. This is a bit of a mouthful, but in short it’s AMD calling attention to the improvements partners GlobalFoundries and Samsung have made to their 14nm LPP processes in the last year. Yields are up and overall chip quality is better, which improves the average performance (clockspeed & power) characteristics of the chips. Both foundries have also been making other undisclosed, small tweaks to their lines to further boost chip quality. It’s not a new fab process (it’s still 14nm LPP) but it’s an improvement over where Polaris 10 production started nearly a year ago.

Typically these kinds of yearly gains would simply be rolled into a product line without any fanfare – these improvements are gradual over time anyhow, not a binary event – but for the RX 500 series AMD wants to call attention to them to explain why clockspeeds are improved versus the RX 400 series cards released last year. Though to be clear here, the difference isn’t dramatic; the gains from a year’s optimization to a manufacturing line are a fraction of a full node improvement.

Meanwhile AMD is also releasing a new revision of Polaris 10, which is being used in the RX 580/570 launch. These revised chips have received further tweaking to reach higher clockspeeds, allowing AMD to reliably clock up a bit higher and/or reduce power consumption a bit. The new revision also fixes a couple of minor issues with the GPUs. Specifically, AMD is adding a new mid-power memory clock state so that applications that require memory clocks faster than idle – primarily mixed-resolution multi-monitor and video decoding – no longer cause the memory to clock up to its most power-demanding speeds, keeping overall power consumption down.

One thing to note here is that while AMD’s chip quality has improved here though the combination of manufacturing improvements and revised silicon, for the desktop AMD is investing all of those gains into improving clockspeeds. This is why the TBPs have gone up by 30-35W over the RX 480 and RX 470.

Power Consumption: By the Numbers

Since all of AMD’s optimizations are focused on bringing down power consumption, let’s take a look at that now. There are a few different things we can look at here, and I’ll start with what’s probably the most burning question: just how much better is the new revision of Polaris 10 over the old revision?

To test this, I’ve taken the Radeon RX 580 sample AMD sent over – PowerColor’s Red Devil RX 580 – and underclocked it to the same clockspeeds as the RX 480. It should be noted however that this process is a bit more complex than just underclocking to the RX 480’s official boost clock of 1266MHz. Because the RX 480 power throttles under both FurMark and Crysis 3, it’s necessary to match the RX 480’s specific clockspeeds in those scenarios.

After doing so, what we find are mixed results.

Load Power Testing, Normalized GPU Clockspeeds (Power Draw at Wall)
	FurMark (740MHz)	Crysis 3 (1230MHz)
Radeon RX 480	231W	301W
Radeon RX 580	205W	314W

Even after dialing the RX 580 down to 1230MHz for Crysis 3 to match the reference RX 480, power consumption at the wall is still 11W higher than the RX 480. Performance is the same, so the RX 580 isn’t doing more work, but none the less power system at a system level is still a bit higher.

On the other hand, turning the RX 580 down to 740MHz to match the RX 480 on FurMark (power viruses cause significant throttling), we find the RX 580 ahead by a rather shocking 26W. Power consumption at the wall is 205W, versus 231W for the RX 480.

Broadly speaking, although FurMark isn’t always the best tool for load power measurement on cross-vendor cards, it has proven to be very reliable when looking at cards based on the same architecture. It suffers from a very specific limitation: it will push a card to its TDP limit, and this can vary among manufacturers, but even with that it typically gives you a consistent and sane metric to compare like-cards.

Consequently I tend to favor the FurMark numbers here. However it doesn’t change the fact that power consumption numbers under Crysis 3 are wildly different, and paint the RX 580 as being worst. So they can’t both be right, can they?

As it stands, I suspect we’re getting into the area of random variation – with a sample size of 1 on each Radeon card, the random variations in quality from GPU to GPU are downing out the actual data. It’s entirely possible we’re looking at a worse-than-average RX 480 and a better-than-average RX 580, especially as the latter has been binned for factory overclocking. However I’m not ready to rule out that something more complex may be going on here: that the improvements Polaris 10’s power curve aren’t linear/consistent. It may be that AMD’s greatest gains are at lower clockspeeds and voltages, and that those improvements taper off at higher clockspeeds and voltages.

But for the moment, I’m ruling it a push. The FurMark data is interesting, but without Crysis 3 being in agreement it’s not enough to say anything definitive.

That New Memory State

Finally, let’s take a look at the specific benefits AMD is touting for the new memory state that the company has included with the new Polaris 10 revision. The new mid-power state allows the memory to be clocked at 4Gbps GDDR5 on the RX 580. The other power states on the RX 580 (and the RX 480) are 1.2Gbps (idle) and 8Gbps (full load), so on the RX 480 if AMD ever needed to increase the memory clocks above idle, their only option was to go to full clocks, which on GDDR5 is relatively expensive.

The two scenarios AMD is looking to address with this new memory clock state are multi-monitor configurations and video playback. In the case of the former, mismatched monitors would require the RX 480 to go to its full memory clocks even when idling. Due to the timing differences, the higher memory clock is needed to avoid flickering. Matched monitors avoid this problem, as they have identical timings. Otherwise in the case of video playback, while AMD has their fixed function decoder to offload most of the work, it still generates a lot of video data, which can require the memory to jump to a higher clock state to keep up. Though the video playback scenario is particularly complex as the GPU clock itself can also jump up if the video decoder needs a higher performance state for itself.

Putting this to the test, I ran both the RX 480 and RX 580 through a mix of multi-monitor and video playback scenarios.

Multi-Monitor Power Testing (Power Draw at Wall)
	Single Monitor (1080p)	Multi-Monitor Matched (1080p+1080p)	Multi-Monitor Mismatched (1080p + 1440p)
Radeon RX 480	76W	76W	100W
Radeon RX 580	74W	74W	100W
GeForce GTX 1060 6GB	73W	73W	73W

Starting with the multi-monitor testing, the results were not what I was expecting. While AMD tells me that this should trigger the new mid-power state, I haven’t been able to successfully trigger it. With matched monitors the RX 580 can go to full idle, just like the RX 480. Otherwise with mismatched monitors, it always goes to 8Gbps, skipping past 4Gbps and never returning. Even with a few different monitors, the results were always the same. Due to the quick launch I haven’t had time to further debug the issue, so I’m not sure if it’s related to the monitors or if it’s something specific to the Red Devil RX 580.

Video Playback Power Testing (Power Draw at Wall)
	Idle	High Bitrate H.264	High Bitrate HEVC
Radeon RX 480	76W	125W	125W
Radeon RX 580	74W	90W	93W
GeForce GTX 1060 6GB	73W	96W	96W

On the plus side however, AMD’s new memory state worked as expected with video playback. Whereas the RX 480 would have to settle for an 8Gbps memory clock when playing back high-biterate H.264 and HEVC video in Media Player Classic – Home Cinema, the RX 580 would settle at 4Gbps. In fact the RX 580 actually performed a bit better than expected; the RX 480 would typically have to go to higher core clock speeds as well, compounding the power cost. As a result power consumption at the wall was notably lower on the RX 580 than the RX 480.

And just for reference, this is actually a bit better than the GeForce GTX 1060 6GB. NVIDIA’s midrange card goes to its maximum memory clock in the same tests, and as a result power consumption at the wall was a few watts higher than the RX 580.

Meet the Cards: PowerColor & Sapphire

For today's launch, AMD sampled us a pair of partner cards: PowerColor's Red Devil Radeon RX 580 8GB, and Sapphire's Nitro+ RX 570. The Red Devil RX 580 sports a modest factory overclock, while the Nitro+ 570 takes comes with a more aggressive factory overclock.

Radeon RX 500 Series Cards
	PowerColor Red Devil RX 580	Radeon RX 580 (Reference)	Sapphire Nitro+ RX 570	Radeon RX 570 (Reference)
Boost Clock	1380MHz	1340MHz	1340MHz	1244MHz
Memory Clock	8Gbps	8Gbps	7Gbps	7Gbps
VRAM	8GB	8GB	4GB	4GB
TBP	220W?	185W	195W	150W
Length	9.5"	N/A	10.5"	N/A
Width	Double Slot	N/A	Double Slot	N/A
Cooler Type	Open Air	N/A	Open Air	N/A
Price	$249	$229	$199	$169

Both of the cards are relatively typical for a custom, factory overclocked design. Along with their overclocks, they feature a dual fan open air cooler. PowerColor is charging a $20 premium on their card, while Sapphire makes that $30 for the Nitro+ RX 570.

PowerColor Red Devil Radeon RX 580

Sapphire Nitro+ Radeon RX 570

The Test

Before diving into our tests, I want to quickly touch upon the test setup. Since AMD isn’t making any reference RX 580 or RX 570 cards, they instead sent over the PoworColor and Sapphire cards listed on the previous page. However both of those are factory overclocked, so both needed to be underclocked to stand-in for the baseline RX 580 and RX 570 cards.

The trick with underclocking cards like this isn’t the clockspeeds, but rather the power consumption. Factory overclocked cards are frequently built and configured for higher TDPs to support their frequencies, which can throw off our results, especially if a baseline card would power throttle in the same situation. So it’s sometimes not enough to simply underclock a card to represent the baseline performance.

In the case of today’s cards, thankfully both of them ship with a second, lower power BIOS. PowerColor calls this Quiet OC on the Red Devil RX 580, and along with reducing the max GPU power by 20W, it reduces the GPU boost clock to 1355MHz, a 15MHz overclock. Sapphire does one better on their Nitro+, as the second BIOS reduces the GPU power by 25W and brings the card down to AMD’s reference clocks.

PowerColor RedDevil RX 580's "Quiet OC" BIOS

Unfortunately the power limit coded into the BIOS don’t perfectly correlate with TBP – the value is just for GPU power – so it’s difficult to precisely tell if these BIOSes match AMD’s 185W and 150W TBPs. However if these values are off, they should still be close to what a real baseline card would get, as they’re in the ballpark of what I’d expect for AMD’s TBPs to begin with. So our results here should be reasonably accurate here for both total power consumption and for accounting for any power throttling during testing.

For our review of the Radeon RX 580 & RX 570, we’re using AMD’s “Crimson Press” driver, version 17.10.1030. Going by the build number, this driver appears to be between the latest 17.3.1 and 17.4.1 Crimson public drivers.

CPU:	Intel Core i7-4960X @ 4.2GHz
Motherboard:	ASRock Fatal1ty X79 Professional
Power Supply:	Corsair AX1200i
Hard Disk:	Samsung SSD 840 EVO (750GB)
Memory:	G.Skill RipjawZ DDR3-1866 4 x 8GB (9-10-9-26)
Case:	NZXT Phantom 630 Windowed Edition
Monitor:	Asus PQ321
Video Cards:	PowerColor Red Devil Radeon RX 580 Sapphire Nitro+ Radeon RX 570 AMD Radeon RX 480 (8GB) AMD Radeon RX 470 AMD Radeon R9 380 AMD Radeon R7 370 NVIDIA GeForce GTX 1070 Founder's Edition NVIDIA GeForce GTX 1060 Founder's Edition NVIDIA GeForce GTX 1050 Ti NVIDIA GeForce GTX 960 NVIDIA GeForce GTX 950
Video Drivers:	NVIDIA Release 381.65 AMD Radeon Software Crimson Press Beta 17.10.1030
OS:	Windows 10 Pro

Rise of the Tomb Raider

Starting things off in our benchmark suite is the built-in benchmark for Rise of the Tomb Raider, the latest iteration in the long-running action-adventure gaming series. One of the unique aspects of this benchmark is that it’s actually the average of 4 sub-benchmarks that fly through different environments, which keeps the benchmark from being too weighted towards a GPU’s performance characteristics under any one scene.

 

DiRT Rally

For the racing game in our benchmark suite we have Codemasters’ DiRT Rally. Codemasters continues to set the bar for graphical fidelity in racing games, delivering realistic looking environments with layered with additional graphical effects. Based on their in-house EGO engine, DiRT Rally includes a number of DirectCompute based compute shader effects, and while it’s not the most punishing game in our suite, it still takes a very good card to sustain the 60fps frame rate that driving games are best played at.

 

Ashes of the Singularity: Escalation

Sorely missing from our benchmark suite for quite some time have been RTSes, which don’t enjoy quite the popularity they once did. As a result Ashes holds a special place in our hearts, and that’s before we talk about the technical aspects. Based on developer Oxide Games’ Nitrous Engine, Ashes has been designed from the ground up for low-level APIs like DirectX 12. As a result of all of the games in our benchmark suite, this is the game making the best use of DirectX 12’s various features, from asynchronous compute to multi-threaded work submission and high batch counts. What we see can’t be extrapolated to all DirectX 12 games, but it gives us a very interesting look at what we might expect in the future.

 

Battlefield 4

One of the older games in our benchmark suite, DICE’s Battlefield 4 remains a staple of MP gaming. Even at its age, Battlefield 4 remained a challenging game in its own right, as very few mass market MP shooters push the envelope on graphics quality right now. As these benchmarks are from single player mode, based on our experiences our rule of thumb here is that multiplayer framerates will dip to half our single player framerates, which means a card needs to be able to average at least 60fps if it’s to be able to hold up in multiplayer.

 

Crysis 3

Still one of our most punishing benchmarks 3 years later, Crysis 3 needs no introduction. Crytek’s DX11 masterpiece, Crysis 3’s Very High settings still punish even the best of video cards, never mind the rest. Along with its high performance requirements, Crysis 3 is a rather balanced game in terms of power consumption and vendor optimizations. As a result it can give us a good look at how our video cards stack up on average, and later on in this article how power consumption plays out.

 

The Witcher 3

The third game in CD Projekt RED’s expansive RPG series, The Witcher 3 is our RPG benchmark of choice. Utilizing the company’s in-house engine, REDengine 3, The Witcher makes use of an array of DirectX 11 features, all of which combine to make the game both stunning and surprisingly GPU-intensive. Our benchmark is based on an action-heavy in-engine cutscene early in the game, and Hairworks is disabled.

 

The Division

The final first person shooter in our benchmark suite, The Division is a multiplayer-only game powered by Ubisoft’s Snowdrop engine. The game’s design focuses on detailed urban environments and utilizes dynamic global illumination for parts of its lighting. For our testing we use the game’s built-in benchmark, which cycles through a number of scenes/areas of the game.

 

Grand Theft Auto V

The latest edition of Rockstar’s venerable series of open world action games, Grand Theft Auto V was originally released to the last-gen consoles back in 2013. However thanks to a rather significant facelift for the current-gen consoles and PCs, along with the ability to greatly turn up rendering distances and add other features like MSAA and more realistic shadows, the end result is a game that is still among the most stressful of our benchmarks when all of its features are turned up. Furthermore, in a move rather uncharacteristic of most open world action games, Grand Theft Auto also includes a very comprehensive benchmark mode, giving us a great chance to look into the performance of an open world action game.

On a quick note about settings, as Grand Theft Auto V doesn't have pre-defined settings tiers, I want to quickly note what settings we're using. For "Very High" quality we have all of the primary graphics settings turned up to their highest setting, with the exception of grass, which is at its own very high setting. Meanwhile 4x MSAA is enabled for direct views and reflections. This setting also involves turning on some of the advanced redering features - the game's long shadows, high resolution shadows, and high definition flight streaming - but it not increasing the view distance any further.

Otherwise for "High" quality we take the same basic settings but turn off all MSAA, which significantly reduces the GPU rendering and VRAM requirements.

Hitman

The final game in our 2016 benchmark suite is the 2016 edition of Hitman, the latest title in the stealth-action franchise. The game offers two rendering paths: DirectX 11 and DirectX 12. While the DX12 path was added after the fact, in the intervening months both the game and driver developers have continued to optimize their software, and at this point the DX12 path is faster than the DX11 path for all current-generation video cards. As with past Hitman games, the latest proves to have a good mix of scenery and high model counts to stress modern video cards.

 

Compute

Shifting gears, let’s take a look at compute performance on new Radeon RX 500 cards.

Starting us off for our look at compute is LuxMark3.1, the latest version of the official benchmark of LuxRender. LuxRender’s GPU-accelerated rendering mode is an OpenCL based ray tracer that forms a part of the larger LuxRender suite. Ray tracing has become a stronghold for GPUs in recent years as ray tracing maps well to GPU pipelines, allowing artists to render scenes much more quickly than with CPUs alone.

For our second set of compute benchmarks we have CompuBench 1.5, the successor to CLBenchmark. CompuBench offers a wide array of different practical compute workloads, and we’ve decided to focus on face detection, optical flow modeling, and particle simulations.

 

Synthetics

As always we’ll also take a quick look at synthetic performance. As the new Radeon cards are just higher clocked versions of the Polaris 10 configurations we already know and love, the cards shouldn't offer too many surprises here.

 

Power, Temperature, & Noise

As always, last but not least is our look at power, temperature, and noise. Next to price and performance of course, these are some of the most important aspects of a GPU, due in large part to the impact of noise. All things considered, a loud card is undesirable unless there’s a sufficiently good reason – or sufficiently good performance – to ignore the noise.

AMD RX Series Video Card Voltages
	Boost	Idle
Red Devil RX 580	1.2063v	0.7625v
Radeon RX 580	1.1625v
Radeon RX 480	1.0625v
Nitro+ RX 570	1.1625v	0.725v
Radeon RX 570	1.1v
Radeon RX 470	1.0125v

As you can likely infer from the earlier discussion on power consumption and TBPs, in order to reach these higher clockspeeds AMD and their partners had to increase their GPU voltages. Relative to both our RX 480 and RX 470, the differences are quite significant. Overall voltages have increased by around 0.1v when using AMD’s reference clocks, and closer to 0.15v for the full factory overclocks. As a result the highest frequencies on these two cards are very expensive in terms of power, and it explains a great deal about why AMD needed to increase TBPs by 30-35W just to add another 40-80MHz to the boost clock.

On the plus side, idle voltages are down for both RX 500 cards. Our RX 400 series cards idled at 0.8v, whereas the RX 580 and RX 570 idle at 0.7625v and 0.725v respectively. This has a minimal impact on a desktop card (especially wall power measurements), but if this is consistent for all AMD chips, it bodes well for the laptop-focused Polaris 11 and Polaris 12 GPUs.

Moving on, let’s take a look at average clockspeeds. Functionally speaking, AMD’s boost mechanism is closer to a fine-grained throttle mechanism: the card always tries to run at its full, advertised boost clock, and will pull back if there’s not enough power available or it triggers thermal throttling. In practice, both the RX 480 and RX 470 regularly power throttled to a small degree; this allowed AMD to keep the cards closer to their optimal point on the clockspeed/power curve.

Radeon Video Card Average Clockspeeds
Game	RD RX 580	RX 580	RX 480	N RX 570	RX 570	RX 470
Tomb Raider	1380MHz	1280MHz	1230MHz	1340MHz	1244MHz	1190Mhz
DiRT Rally	1380MHz	1340MHz	1266MHz	1340MHz	1244MHz	1206MHz
Ashes	1360MHz	1250MHz	1200MHz	1330MHz	1230MHz	1150Mhz
Battlefield 4	1380MHz	1340MHz	1266MHz	1340MHz	1244MHz	1206MHz
Crysis 3	1380MHz	1300MHz	1250MHz	1340MHz	1244MHz	1190Mhz
The Witcher 3	1370MHz	1260MHz	1220MHz	1340MHz	1230MHz	1170Mhz
The Division	1375MHz	1290MHz	1230MHz	1340MHz	1244MHz	1180Mhz
GTA V	1380MHz	1340MHz	1266MHz	1340MHz	1244MHz	1206MHz
Hitman	1365MHz	1250MHz	1200MHz	1330MHz	1230MHz	1130Mhz

Besides supporting higher clockspeeds overall, the higher TBPs of the RX 580 and RX 570 mean that these cards power throttle less often than their predecessors. To be clear, they still throttle, but the average degree of throttling across our game set is lower than with the earlier cards. This means that the RX 580 and RX 570 should be running closer to their maximum clockspeeds more often. It removes a bit of headroom, but it will improve performance.

Adding the fully unlocked factory overclocks into the mix, and we find that throttling is further reduced. The factory overclock BIOSes on these cards have even higher power limits, so even with their higher clockspeeds, they throttle less often. The PowerColor Red Devil RX 580 never averages below 1360MHz in a game, and the Sapphire Nitro+ RX 570 only shaves off all of 10MHz in two of our games. This is also why the factory overclocked cards are as fast as they are; the higher boost clocks are part of the story, but the reduced throttling further boosts performance over the baseline cards.

After AMD fixed their Polaris idle power driver bug last year, AMD’s idle power numbers have been rather consistent. Earlier Polaris 10 cards averaged around 75W at the wall, and so do these newer generation cards.

As for load power consumption, this is where AMD pays the piper, so to speak. Roughly in-line with AMD’s TBPs, power consumption at the wall has increased by a bit over 20W for both the RX 580 and RX 570 relative to their predecessors. At this point the RX 570 is approaching 300W, and the RX 580 is just shy of 325W. This puts the power consumption of the RX 570 at 10W under the GeForce GTX 1070, while the RX 580 is 17W above it. It goes without saying both are well above the GTX 1060 cards that AMD is competing with in terms of performance.

Throwing in the factory overclocks further pours on the power. The Nitro+ system needs 330W here, and the Red Devil system 360W, each around 35W more than their reference-clocked configurations. Bear in mind that this is total system power, so part of the increase comes from the higher CPU power consumption that results from higher framerates, but given the limited framerate difference from the factory overclock, the bulk of the power increase here does come from the cards themselves.

FurMark gives us a more focused view of GPU power consumption, and it tells a similar tale as Crysis 3. We’re looking at a 32W increase in power at the wall for the RX 570, and a 19W increase for the RX 580, the latter actually being a bit less than I was expecting. The new RX 500 series cards do look better against NVIDIA’s GeForce cards, but as I’ve previously mentioned in other reviews, in this generation FurMark only seems to be consistent between cards from the same GPU vendor. Cross-vendor comparisons are more accurate under Crysis 3.

Meanwhile we get a second point of view for the power consumption of the factory overclocked cards. All-told, the higher factory overclocks cause FurMark power consumption to jump by 40W or so. FurMark is a pathological case of course, so games rarely (if ever) draw the same amount of power, but this shows why the factory overclocked cards don’t throttle as much. In their factory overclocked configurations, both cards have very high power limits. These limits are significantly higher than the original reference RX 480.

For idling, both RX 500 cards implement zero fan speed idle. As a result their temperatures are a degree or two warmer than most of the pack. But idle power consumption is so low that these cards have little trouble dissipating that heat with just their heatsinks.

Judging from their temperatures under Crysis 3, both the PowerColor and Sapphire cards are tuned for a balance of noise and temperature. The open-air styled cards reach just shy of 70C when underclocked to AMD’s reference clocks, and 75C with their respective factory overclocks. With their massive coolers, neither card has any trouble with this amount of heat, they just aren’t spinning up the fans by too much to keep noise levels down.

With FurMark the story is much the same as Crysis 3. As it turns out, both cards have a 75C soft cap; once the GPU reached that temperature, the fans will further spin up as necessary to keep temperatures from going any higher. It should be noted that neither card appears to temperature throttle, even under FurMark, as the power throttle is more than sufficient.

Finally with idle noise levels, both cards are silent thanks to their zero fan speed idle implementations. The only noise that’s left comes from the rest of the GPU testbed.

Moving to load noise, both cards continue to impress. The Sapphire RX 570 card, even when it’s running in its full factory overclock condition, barely gets above the noise floor; it’s dissipating 150W (or more) of heat in near silence. PowerColor’s Red Devil RX 580 fares similarly well; it’s under 40db(A) at AMD’s reference clocks, and only finally hits 42dB(A) when fully factory overclocked. Open air coolers have their strengths and weaknesses, but one thing is for sure: manufacturers have increasingly honed their hardware and fan speed algorithms, and these days are producing consistently awesome results.

As for FurMark, noise levels do pick up as you’d expect. When underclocked to AMD’s specifications, both cards are still below 40db(A). It’s only once their factory overclocks and higher power limits kick in that noise starts to become meaningful. The Sapphire RX 570 holds to 42.5dB(A) here, while the Red Devil RX 580 only finally becomes a meaningful source of noise at 48.2dB(A). Though as the Red Devil has a rather high power limit, I doubt it will come anywhere close to this noise level under any gaming workload.

Final Thoughts

While product refreshes have their ups and downs, they’re not without their utility. For AMD and their partners this means a chance to run through the rest of the year with a fresh lineup of cards, and maybe change consumer perceptions a bit in the process. Meanwhile for potential video card buyers, performance has gone up over the previous lineup, and you get more bang for your buck.

So to wrap things up, where do AMD’s new Radeon RX 500 midrange cards stand? For AMD and its partners the picture is better, though I’m not sure it’s quite where they would like to be.

To get the elephant in the room out of the way first, power efficiency has taken a noticeable hit with the Radeon RX 580 and Radeon RX 570. AMD has opted to pour everything into performance rather than fighting a war with NVIDIA they can’t win. As a result their performance and their pricing will dictate their success in the market. Otherwise if you need a power efficient card for a mITX build or smaller air conditioning bill, look elsewhere.

Relative to the RX 480 and RX 470 then, the performance gains we’re seeing with the RX 580 and RX 570 are nothing spectacular, but then AMD has kept their promises similarly small. In practice this means that the RX 580 is only averaging 3% faster than the RX 480 it replaces in AMD’s product stack, while the RX 570 looks better, picking up 7% over the RX 470. Based on these numbers, I feel it’s fair to say that while both cards are faster than their earlier incarnations, I wouldn’t fault anyone for lumping the two generations of cards together. The small gains don’t enable the newer RX 500 cards to do anything the RX 400 cards couldn’t always do; though even a few percent can make all the difference in a game right on the bubble of sustaining 60fps.

Looking at the configurations of the cards, I suspect that the RX 580 would really like some more memory bandwidth, which is why the real-world gains are only about half of what we’d expect looking solely at the boost clocks. The RX 470 on the other hand did get a small increase in memory bandwidth, and it ends up being the stronger card for it. However I don’t know if faster memory is a viable option for AMD and its partners, as none of the factory overclocked cards are shipping with overclocked memory.

As for the competitive landscape then, AMD’s situation has improved, though I fear by not enough. Across the full spread of games in our benchmark suite, the RX 580 and GTX 1060 6GB change lead a few different times, so the RX 580 is able to best NVIDIA’s best in absolute performance in the right games. The problem for AMD is that those games appear to be too few; as a result the RX 580 trails the GTX 1060 by an average of 7% at both 1080p and 1440p. AMD has narrowed the gap somewhat – this was an 11% deficit with the RX 480 – but not by enough. And coupled with AMD’s worse power efficiency, this puts AMD in a tough spot. The biggest challenge right now is that GTX 1060 prices have come down to the same $229 spot just in time for the RX 500 series launch, so AMD doesn’t have a consistent price advantage. That’s the one thing AMD can change, and it’s likely to be where they need to look next.

As for the RX 570, the story is similar. It puts up good numbers for a 1080p card priced at $169, but it faces the continuous threat of a GTX 1060 3GB that starts only $10 higher. Though for bargain hunters looking to stay in the AMD ecosystem, the RX 570 offers a whole lot of value relative to the RX 580, especially if you are willing to make the sacrifices that come with 4GB of VRAM.

Shifting gears a bit, let’s take a closer look at the factory overclocked cards AMD sampled us with in a bit greater detail. AMD and its partners are banking hard on factory overclocked cards, and some of these are going to ship with 1400MHz+ boost clocks. The good news for AMD and its partners is that these cards are indeed faster, helping to further narrow the gap with NVIDIA. The downside is that they’re also more expensive, running headlong into the NVIDIA partners’ own factory overclocked cards.

But regardless of the overall position of the RX 580 and RX 570, both PowerColor’s Red Devil RX 580 and Sapphire’s Nitro+ RX 570 left me impressed. Their hulking size feels a bit out of place, but I can’t argue with the build quality. Both cards are very well built, and both cards are nearly whisper silent while gaming, even with their full factory overclocks in place. Those factory overclocks in turn add around another 5% to their framerates, coming as a mix of improved clockspeeds and reduced/eliminated power throttling. The combination of build quality and improved performance means that, if nothing else, PowerColor and Sapphire have earned their price premiums within the Radeon RX 500 series. This is everything I like to see in terms of a customized, open air cooled card.

Finally, let’s turn our gaze towards the future. While the Radeon RX 500 series gives AMD a welcome chance to reset their lineup for the year, today’s launch is largely just setting the table for more important things to come. The main event for AMD this year is going to be Vega, which is due this quarter. As a complement to Polaris, Vega will mark AMD’s first foray into high-end video cards for this generation. I don’t know how it’s going to turn out, but few things are as exciting as a new architecture, so for both techies and gamers, it should be a fun trip.