Home AMD’s Radeon RX 470 graphics card reviewed

AMD’s Radeon RX 470 graphics card reviewed

Renee Johnson
In our content, we occasionally include affiliate links. Should you click on these links, we may earn a commission, though this incurs no additional cost to you. Your use of this website signifies your acceptance of our terms and conditions as well as our privacy policy.

If you’re just reading this review for the first time, you may not be aware that our initial impressions and conclusions about the Radeon RX 470 were based on incorrect data in a number of the titles we tested, thanks to a problem with our test system. We investigated and explained those issues in a separate blog post. Be sure to read that post before continuing.

The short version of the story is that we had to retest Grand Theft Auto V, Hitman, and Rise of the Tomb Raider for this review, thanks to excessive DPC latency on our testing PC during our initial round of benching. As consolation, we added benches of the DirectX 12 versions of Hitman and Rise of the Tomb Raider to our results, and we also benched the OpenGL and Vulkan renderers available with Doom. This review has changed significantly from its initial form, and some parts of it won’t make sense if you’re not up to speed on the issues we noted and the steps we took to resolve them. The revised review continues below.

As I write these words, it’s just about dawn outside. Yesterday morning, a Radeon RX 470 showed up on my doorstep, and after endless technical difficulties, I finally got around to collecting data on the thing and four competing graphics cards at about 9 PM last night. Since that time, I’ve forgone sleep and consumed more caffeine than is probably healthy for the average horse, never mind the average sedentary hardware reviewer. That’s alright, however, since we have a Radeon RX 470 review to share with you for the trouble.

R9 280X 1000 32 128/64 4.1 2048 384 6 288 250W
RX 470 926 1206 32 128/64 4.9 2048 256 6.6 211 120W
RX 480 1120 1266 32 144/72 5.8 2304 256 7 224 150W
R9 290 947 64 160/80 4.8 2560 512 5 320 290W?
GTX 950 1024 1188 24 48/48 1.8 768 128 6.6 106 90W
GTX 960 1126 1178 32 64/64 2.4 1024 128 7.01 112 120W
GTX 970 1050 1178 56 104/104 3.9 2048 256 7.0 224 145W

As a refresher, AMD revealed the full specs of the Radeon RX 470 a couple days back. This card is the middle child between the e-sports-oriented RX 460 and the VR-ready RX 480. The RX 470’s Polaris 10 GPU has 2048 stream processors and 128 texture units, down slightly from the 2304 SPs and 144 texture units on the RX 480. AMD also hobbled the RX 470 a bit by clocking its 4GB of GDDR5 RAM at 6.6GT/s, slightly slower than the RX 480’s guaranteed 7 GT/s speed floor and typical 8GT/s pairing. The RX 470 maintains the full 32 ROPs and 256-bit path to memory of its slightly better-endowed sibling.

If you’re thinking those changes aren’t that drastic compared to the RX 480, it may or may not be surprising that the Radeon RX 470 is priced at $179.99 and up—just $20 less than the RX 480 4GB reference card. As we’ll soon see, there’s just not that much air between this card and the 4GB Radeon RX 480 at $200 (assuming you can find one at that price). The RX 470 will need to perform quite well to keep buyers from simply jumping ship to the more powerful Polaris card.

The hot-rodded Radeon RX 470 graphics card we have on hand for testing comes courtesy of XFX. This “RS Black Edition” RX 470 claws back most of the cut-down Polaris chip’s clock speed deficit with a 1256MHz boost clock. XFX also used GDDR5 running at 7GT/s instead of the reference 6.6 GT/s spec.

We didn’t have time to tear down this card to the bare PCB before publishing, but the RS cooler is certainly beefier than the reference blower AMD introduced on the RX 480. I count at least three copper heatpipes in there. XFX also outfits this card with a sturdy-feeling backplate and a handy warning LED above the PCIe power connector that glows red if the proper power cable isn’t plugged in. In normal operation, this LED burns blue. Should one of the RX 470 RS’ fans fail, XFX uses a clever installation system that lets users pop out the dead fan and replace it at home instead of sending the whole card in for service. 

The one roadblock this card might face is its price tag. We’re still waiting on official confirmation from AMD on this point, but Newegg already has a listing up for the RX 470 RS for an eyebrow-raising $220. At that price, we’re kind of baffled why someone would choose this card over the perfectly serviceable Radeon RX 480 reference card and the extra computing resources it offers. We’ve already got a reference RX 480 on hand for testing (albeit in 8GB, not 4GB, form), so let’s find out how the RX 470 stacks up.


Our testing methods
As always, we did our best to deliver clean numbers. We ran our graphics card tests on the following test system and with the following graphics cards:

Processor Intel Core i7-6700K
Motherboard ASRock Z170 Extreme7+
Chipset Intel Z170
Memory size 16GB (2 DIMMs)
Memory type Corsair Vengeance LPX
DDR4 SDRAM at 3200 MT/s
Memory timings 16-18-18-36
Chipset drivers Intel Management Engine
Intel Rapid Storage Technology V
Audio Integrated Z170/Realtek ALC1150
Realtek drivers
Hard drive OCZ Vector 180 480GB SATA 6Gbps
Power supply Corsair RM850
OS Windows 10 Pro with Anniversary Update


  Driver revision GPU base
core clock
GPU boost
XFX Radeon RX 470 RS Radeon Software 16.8.1 beta  1256 1750 4096
Radeon RX 480 Radeon Software 16.8.1 beta 1120 1266 2000 8192
Sapphire Nitro Radeon R9 380X Radeon Software 16.8.1 beta 1228 1329 1753 4096
MSI GeForce GTX 970 Gaming 4G GeForce 368.81 1114 1253 1753 4096
Gigabyte Windforce GTX 960 4GB GeForce 368.81 1216 1279 1753 4096

For our “Inside the Second” benchmarking techniques, we now use a software utility called PresentMon to collect frame-time data from DirectX 11, DirectX 12, OpenGL, and Vulkan games alike. We sometimes use a more advanced tool called FCAT to capture exactly when frames arrive at the display, but our testing has shown that it’s not usually necessary to use this tool in order to generate good results for single-GPU setups.

You’ll note that aside from the Radeon RX 480, our test card stable is made up of non-reference designs with boosted clock speeds and beefy coolers. Many readers have called us out on this practice in the past for some reason, so we want to be upfront about it here. We bench non-reference cards because we feel they provide the best real-world representation of performance for the graphics card in question. They’re the type of cards we recommend in our System Guides, so we think they provide the most relatable performance numbers for our reader base. To make things simple, when you see “GTX 970,” “GTX 960,” or “Radeon RX 470” in our results, just remember that we’re talking about custom cards, not reference designs.

With that exposition out of the way, let’s talk results.


Grand Theft Auto V
Grand Theft Auto V kicks off our newly revised tests at 1920×1080. We cranked almost every setting we could to “Very High,” and we even dialed in 2x MSAA plus FXAA to make Los Santos and San Andreas that much more free of jaggies. You can ignore the Vsync and full-screen settings in the screenshots below—they’re artifacts of the way we had to capture those clips. We’ve included buttons beneath some of the graphs to allow readers to get an impression of the changes in our testing data since our initial RX 470 review went live.

If you read our first iteration of this review, the numbers above represent a major change in performance for the RX 480. Originally, both cards were able to push about 60 FPS on our test system. Now that our test system isn’t suffering from insidious DPC latency issues, however, the fully-enabled Polaris chip distances itself considerably from its lesser cousin. Our 99th-percentile frame times are all much lower, too.

The somewhat furrier frame-time graph for the RX 470 versus the RX 480 in GTA V is one case where we might be observing the difference that 8GB of memory makes on the beefier Polaris card, as well. With GTA V‘s “extended distance scaling” maxed, the RX 470, the R9 380X, and the GTX 960 all appear to be swapping data in from main memory often, while the RX 480 appears to be able to keep all the assets it needs in its pool of GDDR5. I’m guessing that keeping that data local has a noticeable impact on smoothness and performance.

These “time spent beyond X” graphs are meant to show “badness,” those instances where animation may be less than fluid. The 50-ms threshold is the most notable one, since it corresponds to a 20-FPS average. We figure if you’re not rendering any faster than 20 FPS, even for a moment, then the user is likely to perceive a slowdown. 33 ms correlates to 30 FPS or a 30Hz refresh rate. Go beyond that with vsync on, and you’re into the bad voodoo of quantization slowdowns. And 16.7 ms correlates to 60 FPS, that golden mark that we’d like to achieve (or surpass) for each and every frame.

In our new results, none of the cards we tested spend any time beyond the 50-ms mark working on tough frames, and only the GTX 960 spends a notable amount of time on frames that take more than 30 ms to render. Past the critical 16.7-ms threshold this time around, however, the RX 480 practically never runs into frames that take more than 16.7 ms to render. The GTX 970 is close behind, while the RX 470’s trouble with our test settings shows itself as a couple of seconds spent past 16.7 ms. We may have to dial back the “extended distance scaling” setting in GTA V for future tests.


Crysis 3
Though it’s an older game, Crysis 3 can still put the hurt on modern graphics cards. Aside from anti-aliasing, we cranked every setting at 1920×1080 to give the RX 470 and friends a hard workout. (Ignore the fullscreen setting in the screenshots below—that’s an artifact of the way we had to capture them.)

In this test, the RX 480 and the RX 470 are within a hair’s breadth of one another in our results, and their frame-time plots are nearly indistinguishable from one another. The GTX 970 rules the roost here, though, while the R9 380X and the GTX 960 4GB bring up the rear.

Just like peas in a pod, neither the RX 470 nor the RX 480 spend any time mired past 50 ms or 33 ms, and they spend about the same amount of time chugging away on tough frames that drop the frame rate below 60 FPS. Moving on.


Rise of the Tomb Raider (DX11)
Rise of the Tomb Raider is a demanding and gorgeous game. We ticked the “high” preset at 1920×1080 and ran through a portion of the “Soviet Installation” level that’s thick with flames, particles, and lighting effects to see how the RX 470 handles Lara Croft’s latest adventures.

RoTR has always favored Nvidia cards in our tests, and even in our revised results, the GTX 970 takes a decisive victory in both our 99th-percentile and average FPS numbers. The RX 480 puts more light between itself and the RX 470 than it did in our poisoned initial set of test results,  and it improves its 99th-percentile frame time numbers, too.

The first major signs of “badness” for these cards running Rise of the Tomb Raider in DX11 mode crop up at the 16.7-ms threshold in our new set of data. In keeping with its high frame rate and low 99th-percentile frame time, the GTX 970 spends an imperceptible amount of time working on some difficult frames past the 16.7-ms mark. That’s great performance, and we can conclude the card is delivering a solid 60 FPS or better for the vast majority of our one-minute testing period.

The Radeon RX 480 spends about a second working on tough frames past this threshold, while the RX 470 needs about three. That’s still much better performance than the cards we formerly might have recommended in this price class, however. The R9 380X and GTX 960 fare much worse.


Rise of the Tomb Raider (DX12)
Since we already needed to go back and retest Rise of the Tomb Raider, we decided to deploy our brand-new Inside the Second methods (as seen in our Radeon RX 460 review) to gather some DX12 data with these more powerful cards for the heck of it. Aside from flipping on RoTR‘s DX12 mode, we kept settings identical to our DirectX 11 testing.

The switch to the DX12 API has practically no effect on these powerful graphics cards. The Radeon R9 380X does benefit from an improvement in its 99th-percentile frame time, but that figure itself isn’t all that impressive for the Tonga part.

In our measures of “badness,” the RX 470 spends almost five seconds of our one-minute test period working on frames that take longer than 16.7 ms to complete. That may be enough time spent churning to produce noticeable slowdowns and roughness in the course of Lara Croft’s adventures. The RX 480 slashes that figure by 80%, so perhaps the RX 470’s issues stem from the fact it’s a fresh card with immature drivers. We’ll have to see whether that’s the case with time.


The Witcher 3
To test the latest adventures of Geralt of Rivia, we maxed out every graphics setting in The Witcher 3‘s menus aside from Nvidia’s HairWorks tech. We then followed a winding road through a part of the Northern Kingdoms with plenty of water, vegetation, and non-player characters to spare.

Just a single frame per second separates the RX 470 and the RX 480 in this test, and the cards provide practically identical performance in our 99th-percentile frame time metric.

Our measures of “badness” show that the RX 470 has a little more trouble than the RX 480 if we use the 16.7-ms threshold as our yardstick. The GTX 970 remains the most fluid among the cards we tested, but both Polaris Radeons and the GTX 970 provide a plenty playable Witcher experience.


Hitman (DX11)
Hitman is among the hardest games for graphics cards to run smoothly among those we’ve tested of late. Let’s see how the RX 470 handles it maxed-out at 1920×1080. Just as we did with Rise of the Tomb Raider, we collected and crunched DirectX 12 data for Hitman during the course of revising our results, too.

If Rise of the Tomb Raider tends to favor Nvidia cards, here’s a win for the red team. The Radeon RX 480 handily beats out the GTX 970 in our average frame rate results, and the RX 470 takes a slight lead over the Maxwell card. Both Polaris contenders are neck-and-neck with that card in our 99th-percentile frame-time results, too.

Our measures of “badness” show that both Radeons have little trouble maintaining a solid 60 FPS or better in this game. The GTX 970 is equally capable. Pour one out for the GTX 960, though. Perhaps Nvidia has a GTX 1050 waiting in the wings to avenge it.


Hitman (DX12)
Here’s some next-gen API results for Hitman using the same graphics settings as we chose for DX11. AMD prominently features Hitman‘s DX12 numbers in its marketing materials for Polaris cards, so we were curious to see how the numbers would shake out for the Radeons when we flipped the API switch over to Microsoft’s latest.

Moving to DX12 with Hitman gives both of our Polaris cards a minor performance boost, along with an accompanying minor decrease in 99th-percentile frame times. The Radeon R9 380X also benefits from the move. Both GeForces perfom worse in DX12 mode than they do in DX11, though, so it’s probably best to stick with that API if you’re a Maxwell owner.


None of our cards except for the GeForce GTX 960 have any trouble running Hitman at a solid 60 FPS. Let’s see how Doom treats our test subjects.


Doom (OpenGL)
To get an idea of how id Software’s latest Doom runs on our test cards, we set the game to use the Ultra preset with 16x anisotropic filtering and 8x TSSAA. We then put the chainsaw to some demons guarding the yellow keycard in the early stages of Doom‘s Foundry level. Like we did with the rest of our tests, we ran Doom at 1920×1080. 

As we discovered in our Radeon RX 460 review, Doom‘s OpenGL renderer is not kind to Radeons, and the story doesn’t change with Polaris 10. The RX 470 and RX 480 both deliver admirable average frame rates, but they trail the super-fast, super-smooth GTX 970 in both average FPS and 99th-percentile frame time numbers.

No card we tested with Doom‘s OpenGL renderer spent any time past 50 ms or 33 ms, a testament to how smooth and well-optimized Doom is in general. In fact, for the GTX 970, we have to click all the way over to our 8.3-ms chart to see any truly troublesome frames. For perspective, those numbers mean that the GTX 970 spends just 11 seconds of our one-minute test run working on frames that drop its average below 120 FPS.

The Polaris Radeons aren’t quite that good with OpenGL, but they do keep the time spent past 16.7 ms to a minimum. Doom‘s Ultra settings even let the GTX 960 shine a bit, for once, but the R9 380X spends a significant amount of time working on frames that drop its FPS average below 60. That time translates into noticeable slowdowns during gameplay.

Next, let’s see what happens when we unleash the fury of Vulkan in Doom.


Doom (Vulkan)
As with our other next-gen API tests, we kept Doom‘s graphics settings identical to those for OpenGL. The only change we made was to turn on Vulkan.

Yow. In our average FPS measure, the Radeon RX 480 gets a 45% boost from the move to Vulkan at 1920×1080, and it comes darn close to delivering 99% of its frames at 90 FPS or better. Recall that we’re running the game at the Ultra preset with 8X TSSAA .

The RX 470 isn’t quite as fast, but it beats the GeForce GTX 970 in both performance potential (as measured by average FPS) and ties it for smoothness (as measured by our 99th-percentile frame-time metric). The GeForces end up just a bit worse off than they were with OpenGL, so there’s no real advantage to enabling this next-gen API for Maxwell GPUs, at least. Seems we need to get one of them newfangled GTX 1060s for our test bench to conduct a fair fight.

After a second round of retesting, the Radeon R9 380X also got an appreciable performance boost from Vulkan. For some reason, our first test run pegged the card at 59 FPS, while a second test with the Radeon Software 16.8.2 hotfix driver raised that figure to a quite-respectable 87 FPS. The card’s 99th-percentile frame time also improved from a troubling 28.4 ms to a much happier 15.7 ms. To keep the playing field equal, we re-ran every Radeon card above using the same drivers, though neither the RX 470 nor the RX 480 demonstrated any meaningful changes in performance from that update.

After our re-testing of the Radeon cards above, none of the cards spend any time beyond the critical 50-ms or 33.3-ms thresholds, and only the GTX 960 spends a noticeable amount of time working on frames that drop its average under 60.

To see any meaningful struggles from our top three cards, we have to click over to that 8.3-ms plot again. The Radeon RX 480 8GB card comes darn close to delivering a perfect 120-FPS experience at 1080p with Doom‘s Ultra settings. It spends just three seconds of our one-minute test run on frames that take more than 8.3 ms to render, and that translates into an amazing Doom experience.

The RX 470 spends about three times as much time working on tough frames past 8.3 ms, so it’s in a close race with the GTX 970. The Radeon R9 380X does slightly worse still, while the GTX 960 brings up the rear by a wide margin.


Power consumption
Let’s take a look now at the effects the Radeon RX 470 has on our system power draw. Our “under load” tests aren’t conducted in an absolute peak power draw scenario. Instead, we run Crysis 3 on each card to show us power draw with a typical gaming workload.

Here’s another spot where the RX 470 isn’t much different from the RX 480. AMD specifies a board power of 120W for this card, but even without more precise testing equipment, we can see that the overclocked XFX RX 470 is drawing about as much power as the reference RX 480 8GB board. No matter how you slice it, the RX 470 lets our system draw less power than it does with an R9 380X installed, and it delivers much greater performance while doing it. Not bad. Meanwhile, our factory-boosted GeForce GTX 970 consumes the most power of the lot.

Noise levels

At idle, all of these cards save the Radeon RX 480 can shut off their fans at idle, making the roughly 32 dBA produced by the rest of my test system the lower limit to quietness in this test. Crank up Crysis 3 on the RX 470, though, and it becomes the loudest card among those we tested. Even the Radeon RX 480 reference card is quieter. The XFX card isn’t unpleasant-sounding, at least. Its twin fans mostly sound like moving air instead of an unpleasant tonal or growly noise.

GPU temperatures

The XFX cooler keeps the RX 470 the coolest of the Radeon cards we have on hand to test, so at least its aggressive fan profile has a good payoff. We did discover an interesting behavior with this card despite its cool-running nature. Even with its 1256-MHz boost clock specification, this RX 470 seemed more content to hang out in the 1200-MHz zone after being loaded with Crysis 3 for a while, rather than boosting up to or near its on-paper maximum. We expected higher clock speeds given this card’s cool-running nature.

Strangely, the RX 480 reference card seemed plenty able to run at or near its specified 1266-MHz maximum in our informal testing. We’re guessing AMD’s reference board engineers were willing to let the Polaris 10 GPU run hotter in exchange for the boostier behavior, given its 80° C load temperature. Perhaps some tuning in Radeon WattMan (née OverDrive) could let us extract more performance from XFX’s custom RX 470. We’ll have to look into this matter further.


Since we’ve completely revamped a large portion of our test data for the RX 470 using the same methods we introduced with the Radeon RX 460, we can also crunch our price-to-performance data three different ways. Our value scatters now let you see each card’s price-to-performance using DirectX 11 and OpenGL data only, DirectX 12 and Vulkan data only, or a “best API” chart that accounts for the best result each card deliverd using either current-gen or next-gen APIs.

Let’s start off with average FPS per dollar. In our initial review, we concluded that the Radeon RX 470 delivered RX 480-class performance for similar or higher prices. It’s certainly not on par with the bigger Polaris chip in our results any more, though. The RX 480 8GB card turns in especially strong showings when it’s not shackled by DPC latency on our test rig, and it takes a commanding lead in our “best API” chart. That result doesn’t help the value proposition of the $230 XFX Radeon RX 470 we were sent for testing, though. For just $10 more, the RX 480 8GB reference card delivers about 10 more FPS, plus twice the RAM. We initially advised system builders to step up to the RX 480 or seek out a cheaper RX 470 if possible, and our updated test data just puts an exclamation point on that advice.

Next, let’s look at each card’s delivered performance per dollar using our advanced 99th-percentile frame time metric. To make this measure work, we’ve converted the 99th-percentile frame-time numbers for each card into FPS.

Our 99th-percentile frame-numbers still tell us more or less the same story as we gleaned from them in our initial review. The XFX RX 470 RS card is quite good at smooth frame delivery, and it would be an amazing value at AMD’s $180 suggested price. At $230, its value proposition is just so-so. The problem, as we’ve already noted, is that an extra $10 nets you a big leap in smoothness from the RX 480 8GB card, and $30 in the pocket leaves you with an RX 480 4GB card, if you can find one for that price. Neither of those steps help the XFX RX 470 RS’ value proposition much.

In isolation, the XFX card we tested is a nice product, but it’s not a perfect representative of the RX 470 breed. While the card boasts solid core and memory clock boosts from the factory, its fan profile favors performance over politeness, and it seems to have trouble keeping up its 1256 MHz boost clock all the time. Asus, Gigabyte, Sapphire, PowerColor, and MSI all have custom RX 470s in the works, so we’ll have to see how those cards measure up in due course. We’re especially curious whether any of them will hit AMD’s suggested price point, and how they’ll get there.

At the end of the day, the RX 470 is a great performer with an awkward price tag, at least going by the card we tested. It brings many of the same benefits we saw from the RX 480—including smoother frame delivery than its forebears—to a potentially wider audience. The problem for AMD, if it can be called that, is that it has too many good products huddled near similar price points right now. If the company can draw a brighter line between the RX 470 and the RX 480 with pricing as its pen, the lesser Polaris may yet find a home in the market. Our testing results still suggest buyers will find a better value in the Radeon RX 480, however, and that’s the card we’d be after were we in the market for a Radeon today.

The Tech Report - Editorial ProcessOur Editorial Process

The Tech Report editorial policy is centered on providing helpful, accurate content that offers real value to our readers. We only work with experienced writers who have specific knowledge in the topics they cover, including latest developments in technology, online privacy, cryptocurrencies, software, and more. Our editorial policy ensures that each topic is researched and curated by our in-house editors. We maintain rigorous journalistic standards, and every article is 100% written by real authors.

Latest News

EU Probes into Meta’s Facebook and Instagram for Not Doing Enough to Ensure Child Safety

EU Probes into Meta’s Facebook and Instagram for Not Doing Enough to Ensure Child Safety

Tesla Fires 600 Employees in California

Tesla Rolls Out Another Round of Mass Layoff: 600 Employees Fired in California

Tesla decided to cut an additional 600 jobs at the company’s manufacturing plant and engineering headquarters in California – at its Fremont and Palo Alto locations to be precise. The...

Bitcoin Whales Accumulate Over 3,400 BTC Tokens, Signaling a Possible Price Increase Ahead
Crypto News

Bitcoin Whales Accumulate Over 3,400 BTC Tokens, Signaling a Possible Price Increase Ahead

Bitcoin has broken above the $66,000 price, confirming rising bullish sentiment. Recent reports reveal that Bitcoin whales have accumulated over 3,400 coins amid the improving market sentiment. This increase in...

Pink Drainer Packs Up Following Attacks On 20K Victims And $75M Crypto Theft
Crypto News

Pink Drainer Packs Up Following Attacks On 20K Victims And $75M Crypto Theft

Avalanche (AVAX) Set to Rally to $40 Based on Bullish Patterns
Crypto News

Avalanche (AVAX) Set to Rally to $40 Based on Bullish Patterns

Ethereum (ETH) Trades Above $3,000 Price Mark, What's Next?
Crypto News

Ethereum (ETH) Trades Above $3,000 Price Mark, What’s Next?

OpenAI Strikes Deal with Reddit to Use Its Content to Train ChatGPT

OpenAI Strikes Deal with Reddit to Use Its Content for ChatGPT Training