Nvidia fans have had their fun with their recently introduced cards but in the middle of all of that, we have also been hearing more and more about Polaris, AMDs latest GPU architecture. Most of it was rumors, but at the beginning of this month AMD did host a Webcast at Computex and finally gave us some more information about it. The big news was that the new RX 480 was coming on June 29th, it will have impressively smooth VR, and its price would be $199. Well, I’ve had the RX 480 in the office for the last few days and today's the day that we can finally sit down and talk about how it performs. At its price point, we know that AMD isn’t looking to shake up the top of the charts, but it's clear they want to bring good performance to the mainstream gaming market. The high-end cards might be exciting, but this is the price range people are buying the most, I’m excited to find out if the RX 480 is worth all of the hype.
Product Name: AMD Radeon RX 480 8GB
Review Sample Provided by: AMD
Written by: Wes
Pictures by: Wes
Amazon Link: HERE
|Clock Speeds (Boost / Base)||1266 MHz / 1120 MHz|
|Peak Performance||Up to 5.8 TFLOPS|
|Memory Size||4/8 GB|
|Memory Bandwidth||224 GB/s or higher|
|Memory Interface||256 bit|
|AMD FreeSyncTM Technology||Yes|
|DirectX® 12 Support||Yes|
|DisplayPort Version||1.3 HBR / 1.4 HDR Ready|
The RX 480 is going to be available in both 4GB and 8GB models and AMD reflected that in their sampling by sending out a random number of each out to the press. To make sure you know what our testing is with, I’ve also included a screenshot of our GPUz showing that our sample is an 8GB card.
During AMD's webcast at Computex they started off their discussion with a focus on gamers who are buying video cards that range in price from $100 to $300, they say that 84% of gamers are buying cards in that range. Because of that, it’s no big shock that the first card they introduce in the new RX 400 series of cards would be focused directly in the middle of that range.
To get good performance at that price point they did have to make a few big changes with the new 4th generation GCN architecture. Like Nvidia with the new 1000 series cards, a lot of those improvements are focused on improving memory performance. They started with improvements in their shader efficiency for a 15% performance improvement. This includes turning the L2 cache behavior, increasing per wave instruction buffer size, instruction prefetch, as well as native support for FP16 and Int16.
They also made improvements in the memory and delta color compression engines. Adding support for up to 8GBps memory to speed things up.
Moving to FinFET 14, or in other words, 14nm production size. This helps shrink the size of memory and also makes power circuitry smaller to help improve the efficiency. It might seem like things get smaller every year, but take a look at the dates for the die shrink moves in the photo below. The last shrink was all the way back in 2011, going to 14nm was a long process. Especially when compared to the gaps between the other shrinks.
So what are the introducing with Polaris? Well, there is, of course, the RX 480 that I will be taking a look at today. But in addition to that, they have put out details on the RX 470 and RX 460 as well. They don’t include any information on clock speeds or power usage for those cards like they do with the RX 480. But you can get a good idea of how they compare to the RX 480 by looking at the number of CUs. The RX 480 has 36, the RX 470 32, and the RX 460 has 14. Unlike Nvidia though with Polaris AMD isn’t making a mad dash to crazy core clock speeds. The RX 480 is running at 1120MHz with a boost clock of 1266MHz, this is a lot more like previous generation cards. The 36 Compute Units translates to 2304 stream processors. Funny enough this is actually less than the R9 390 with its 2560 but a little under 300 more than the R9 380X’s 2048.
Card Layout and Photos
With our RX 480 sample being purely a reference card this time around there wasn’t any need for a packaging section. I did want to point out though that the cards with the reference cooler are coming completely wrapped up in plastic. You get that fresh feeling when pulling the sticker off from four different sides!
I haven’t had any AMD reference cards in the office for a while. The cooler on the RX 480 looks completely fresh and new to me even though it does match the reference cards from the high-end R9 300 series cards as well as the Fury cards. The design is simple and clean, just how I like things. It's all black with the Radeon logos on the left and on the fan in red to stick with the team red theme. Most of the outside of the card has this textured panel covering it and the pinhole like texture give the RX 480 a little style. Beyond that, it used a plastic casing unlike the reference and now “Founders Edition” cards from the green side. At first glance, the textured area looks almost like they have bolted on a rubber pad but it is made of plastic just like the rest of the fan shroud.
The PCB and the top of the fan shroud also push the limits for height compared to other reference cards.
Being a reference card it does have the normal blower design. This means that it pulls air in from this one fan and blows it across the card and out the back of your PC. Cards with aftermarket coolers pack in two and sometimes three large fans and drop the blower design to push air directly down over the heatsink and against the PCB. Blower cards aren’t preferred for overclockers, but a lot of OEMs like them and SFF builders prefer them as well normally. This is because in cases with low airflow like a Dell or a SFF case you don’t want to add to the heat. Blowing the heat out of the case helps keep temps down in those situations. Cases with a lot of cooling on the other hand benefit from cards with aftermarket coolers because those cooler cool the card better and then the case can get the heat out.
Being a blower cooling design the reference cooler on the RX 480 also has a sealed top, bottom, and end. The top of the card does have another Radeon logo on it, but those looking to show off their team red love might be disappointed it's not backlit.
To handle the RX 480’s 150-watt TDP they only had to use a single 6-pin power connection. That said it is in a bit of a strange place. Due to a shorter PCB, the power connection is about 3 inches towards the middle of the card.
Flipping the card around we can better see what is going on. Much like a few of the mid-range Nvidia cards the RX 480 has a PCB that is shorter than the length of the cooler. The cooler lines up perfectly with the PCB and the backplate are of the cooler is actually metal as opposed to the rest of the cooler. They do this when going shorter on the cooler would mean bad cooling performance. The shorter PCB actually helps with the cooling because they are able to pull fresh air in from the back of the card as well as the front. In addition to that, this design means the fan doesn’t have to sit above anything. When laying everything out on a PCB they normally have to put the fan above things that aren’t going to get warm because there isn’t room for a heatsink between the fan and the PCB. This shorter PCB design also means we will most likely see aftermarket cards that are shorter, so SFF guys (like myself) will have a good option on the AMD side of things to pick from.
For display connections, AMD did change things up with the RX 480. They have completely dropped the DVI connection. It still has three full sized DisplayPort connections as well as an HDMI. These are the same as what I have been seeing on most cards recently but dropping the DVI connection could mean problems for some people. I suspect we will be seeing aftermarket cards with it still. It seems they dropped it to open up more room for cooling. The entire second PCI slot has a vent in it now as well as a shorter vent above all of the other connections.
Typically, I only take a look at software for aftermarket cards that have specific programs needed to take full advantage of them. But I haven’t taken a look at AMD's software for a while and they have now have their new Crimson Edition software. Being completely honest the software side of things has always been one of the downsides for using an AMD and even before than ATI product. The competition seemed to always be ahead and innovating and for a long time they seemed stagnant on the software side. Developing good software and good drivers is expensive and takes a lot of time but it’s an investment that does pay off. Well a year ago this past May AMD started this long process with their Omega driver, stepping back and focusing on just fixing bugs and cleaning up performance. The Crimson Edition was a similar move, but with the focus being on usability and software speed.
The new software is completely different when you open it up. Up top, you have five different tabs and down on the bottom you can get at your preferences, notifications, and you can finally update inside the software. The first page is the gaming tab, this is where most of the action is. The tab opened up instantly, you can tell the focus on speed worked. This page shows up with every game we have installed on our testbench. By clicking on a specific game you can get into the settings and force the game to run with or without any setting you would like. You can also even jump into the profiles Wattman page, this is where you can overclock your card for any specific game. This is great if you don’t want to put the card under the load all of the time but maybe need a small bump in a specific game.
Along with the games, there is also a global page. You can get into the same settings as the games here, but all of these changes will force the settings to be on in everything you do. So, for example, you really like V-Sync, you could turn it on here and be done with it.
The global page is also the standard page for overclocking using what they call Wattman. Ignoring the weird name, it’s great that the overclocking options are right in the software. AMD has always done this where Nvidia always has you download one of their partner's programs to do the same. Up top the software graphs things like the memory and GPU clocks over time, temperatures, fan speed, and GPU activity to help you get an idea of how your overclock is doing. All of the adjustment settings come set to auto but if you click on them you can start to dive in. You can set power or thermal limits for temperatures. They even have a new cool acoustic limit for fan speeds. For actual overclocking, I was surprised to see the new state setup. Basically, when gaming depending on the load your card will run in different states, each having a different clock speed. This is a lot like a boost clock but broken down even more. So we can actually get in and overclock each one individually.
The second main tab (the tabs from up top move to the bottom when you have them open) gives us tuning options for display color. The main options are just for sports and movies but you can also jump into the custom settings.
The Display page shows all of your hooked up displays. In our case, we have just the one monitor. Funny enough our monitor supports AMDs FreeSync but using the HDMI connection did give me the Not Supported on the option to turn on FreeSync. Beyond that, you can get into Virtual Super Resolution where you can encode at 4k or other high resolutions then scale down to your resolution to get more detail.
There is the Eyefinity tab but with us just using one display there is nothing going on there. Then the last tab is the system tab. This lets us take a look at a few details similar to GPUz showing our clock speeds, memory, software versions, etc.
Our Test Rig and Procedures
|Our Test Rig|
|CPU||Intel i7-5960X||Live Pricing|
|Memory||Kingston HyperX FURY Black 32GB Quad Channel Kit 2666 MHz||Live Pricing|
|Motherboard||Gigabyte X99-SOC Champion||Live Pricing|
|Cooling||Noctua NH-U12S Cooler||Live Pricing|
|Power Supply||Cooler Master V1000 Power Supply||Live Pricing|
|Storage||Kingston Hyper X Savage 960GB SSD||Live Pricing|
|Case||Dimastech Test Bench||Live Pricing|
|Our Testing Procedures|
|3DMark||The same goes for the most current version of 3DMark using the Fire Strike benchmark in normal, extreme, and ultra settings|
|Unigine Valley Benchmark 1.0||Using the Extreme HD preset to get an average FPS|
|Catzilla 4k||Default tests for 1080p, 1440p, and 4k resolutions using the overall score for each as our result|
|SteamVR||Default SteamVR test using Average Quality score|
|HITMAN 2016||Fullscreen with V-Sync turned off Detail, Texture Quality, Shadow Maps, and Shadow Resolution all set to their highest settings. We test using both DX11 and DX12 at both 1080p and 1440p resolutions.|
|Ashes of the Singularity||Built-in benchmark ran at 1080p and 1440p with graphics settings set to the “Crazy” setting with the exception of turning off V-Sync Mode. The benchmark scenario is set to GPU Focused and we use the Average Framerate for All Batches as the result. Tests are run both in DX11 and DX12|
|The Division||Built-in benchmark ran at 1080p and 1440p with graphics settings set to the default “Ultra” setting with the exception of turning off V-Sync Mode|
|Bioshock Infinite||Using the Adrenaline Action Benchmark Tool we run Bioshock Infinite on the “Xtreme” quality setting. This has a resolution of 1920x1080, FXAA turned on, Ultra Texture detail, 16x Aniso Texture Filtering, Ultra Dynamic Shadows, Normal Postprocessing, Light Shafts on, Ambient Occlusion set to ultra, and the Level of Detail set to Ultra as well. We also run this same test at 2560x1440 using the same settings as mentioned above.|
|Tomb Raider||Using the Adrenaline Action Benchmark Tool we run Tomb Raider on the “Xtreme” quality setting. This has a resolution of 1920x1080, Exclusive Fullscreen turned on, Anti-Aliasing set to 2xSSAA, Texture Quality set to Ultra, Texture Aniso set to 16x Aniso, Hair Quality set to TressFX, Shadow set to Normal, Shadow Resolution on High, Ultra SSAO, Ultra Depth of Field, High Reflection quality, Ultra LOD scale, Post-Processing On, High Precision RT turned on, and Tessellation is also turned on. We also run this same test at 2560x1440 using the same settings as mentioned above.|
|Hitman: Absolution||Using the Adrenaline Action Benchmark Tool we run Hitman: Absolution on the “Xtreme” quality setting other than the MSAA setting is turned down from 8x to 2x. That setting puts the resolution at 1920x1080, MSAA is set to 2x, Texture Quality is set to High, Texture Aniso is set to 16x, Shadows are on Ultra, SSA is set to high, Global Illumination is turned on, Reflections are set to High, FXAA is on, Level of Detail is set to Ultra, Depth of Field is high, Tessellation is turned on, and Bloom is set to normal. We also run this same test at 2560x1440 using the same settings as mentioned above, except on the “high” setting.|
|Sleeping Dogs||Using the Adrenaline Action Benchmark Tool we run Sleeping Dogs on the “Xtreme” quality setting. That means our resolution is set to 1920x1080, Anti-Aliasing is set to Extreme, Texture Quality is set to High-Res, Shadow Quality is High, Shadow Filter is set to high, SSAO is set to High, Motion Blur Level is set to High, and World Density is set to Extreme. We also run this same test at 2560x1440 using the same settings as mentioned above.|
|Total War: ROME II||Ultra-setting tested at 1920x1080 and 2560x1440, built in forest benchmark|
|Middle-earth: Shadow of Mordor||Using the built-in benchmark we test with ultra settings at 1440p|
|Sniper Elite 3||Ultra-setting tested at 1920x1080 and 2560x1440, built in benchmark|
|Thief||Tested using the “Very High” setting at 1920x1080 and 2560x1440|
|Folding at Home 2.2||Using the Folding at Home benchmark 2.2.5 set to OpenCL, WU set to dhfr, and run length set to the default 60 seconds. We test at both double and single precision and use the score at the result|
|CompuBenchCL||Video Composition and Bitcoin tests|
|Unigine Valley Benchmark 1.0 heat testing||We run through Unigine Valley using the “Extreme” preset for 30 minutes to test in game cooling performance with the fan speed set to auto then again with the fan set to 100%.|
|Power Usage||Using Unreal Valley Benchmark 1.0, we get our “load” power usage number from the peak power usage during our test. We get our numbers from a Kill-A-Watt connected to the test benches power cord.|
|Noise Testing||Our Noise testing is done using a decibel meter 3 inches away from the video card on the bottom/fan side of the card. We test an idle noise level and then to get an idea of how loud the card will get if it warms all the way up we also turn the fan speed up to 50% and 100% and test both speeds as well. The 100% test isn’t a representation of typical in-game noise levels, but it will show you how loud a card can be if you run it at its highest setting or if it gets very hot.|
As always to start off my testing I always go with our synthetic benchmarks. They don’t do the best job of letting you know what kind of FPS to expect in game, but they are consistent tests that help me compare the performance between cards. To start off my testing I always use 3DMark. I test using the Fire Strike benchmark and I run it using all three of its settings. The three tests run at 1080p, 1440p, and 4k to help give us a view of how the RX 480 performs at all three resolutions. In the performance setting, I was really happy to see that the RX 480 fell right in with the overclocked R9 390’s and not far from the GTX 980. In fact, the GTX 970 that is its closest competitor right now had a score 1000 points less and that was with a heavily overclocked GTX 970. The Extreme and Ultra tests paint a similar picture as well with their 5460 and 2699 scores respectively.
In Unigine’s Valley Benchmark 1.0 things look a little different. Here the R9 390’s are about 10 FPS faster. The RX 480’ is still significantly faster than the R9 380.
Catzilla is a new test for me, but like 3DMark I like that it gives us a chance to test at three popular resolutions. Here the RX 480 came in extremely close to the GTX 780. When compared to the R9 380 the R9 380 you can start to see just how much of an improvement the RX 480 is, though.
The last test was using the SteamVR test to see just how well the RX 480 will perform with VR. This one had me especially curious because when the RX 480 was announced they pushed that it would be bringing VR performance to a price point that people could afford. After running this test we can confirm that the RX 480 does fall into the VR Ready category.
So synthetic benchmarks are well and good, but as I’ve always said at the end of the day it’s all about the in-game performance. That’s why we build badass computers right? So the most of my testing with the RX 480 8GB was spent testing its performance in game. In fact, in addition to retesting more cards on the new test bench to recover those results, I also continued to work on expanding and tweaking our test suite. I dropped Grid Autosport due to performance issues across all of the AMD cards. I also added in the new Hitman in both DX11 and DX 12. In addition to that, I also added Ashes of the Singularity in DX12 and retested the cards I added during the GTX 1080 and GTX 1070 reviews.
With 11 different games being tested (plus the two games tested twice for both versions of DirectX) it can get a little crazy to read, so I do like to condense all of our testing down into two graphs to break down the overall playability. This breaks the results down into below 30 FPS that I consider to be unplayable, over 30 FPS but less than 60 that is playable but not ideal, and then everything over 60 FPS is perfectly playable. I do this for both 1080p and 1440p resolutions. The RX 480 8GB performed well, especially at 1080p. At that resolution 9 of the tests came in over 60 FPS with two in the 30 to 60 range and one under 30 FPS. The under 30 FPS result was in Ashes of the Singularity using DX11, a very demanding game at the settings I’m running. The 30-60 FPS results were the DX12 Ashes of the Singularity and also The Division. The Division was very close to 60 with its 57.6 FPS, I don’t think we can be too upset about that.
At 1440p the RX 480 had six games in the 60+ range, 5 in the 30 to 60 Range, and 2 under 30. Once again Ashes of the Singularity was both under 30 results with its DX11 and DX12 results. The 30 to 60 games at 1440p were The Division, Thief, Tomb Raider, Hitman: Absolution, and Sleeping Dogs. Keep in mind these are all games running at their highest settings, so beyond them still being playable you can get them up over 60 but you are going to have to turn things down a little.
Diving into the results a little more. I was excited to finally take a look at our two DirectX 12 games and see if there are any performance differences compared to DirectX 11. In Hitman there was about a 1 FPS jump in performance at 1080p across all of the cards but the RX 480 didn’t stand out more than any of the others. In Ashes of the Singularity though I did see a jump in performance in DX12 with the RX 480 were some of the Nvidia cards actually dropped in speed or at best stayed the same.
In the rest of our in-game benchmarks, things were mostly smooth sailing. The RX 480 performed well ahead of the GTX 780 that outperformed in 3DMark. In Thief the RX 480 even pulled ahead of the GTX 980 Ti!
For compute testing I have finally expanded our out testing with two new tests from CompuBenchCL to test Video Composition performance and also BitCoin hash performance. As for Folding at Home, I have started to move over to the new FaH 2.2.5 benchmark from our older 1.2 bench. That said I did test both for now. In Folding at Home, I was impressed to see the RX 480 performing spot on with the GTX 980 in the Single Precision testing. This is worlds above the R9 390 as well. Of course in the Double Precision testing, Nvidia has cut out their FP64 cores so the Nvidia cards that performed well in the Single test drop here. The RX 480 performs almost as good as the GTX 1080 here but the R9 390’s still rule double precision performance.
In CompuBenchCL I ran all of the tests but cut it down to just these two results to keep things simpler to read and to test two compute areas that a lot of people care about, video composing and BitCoin mining. In the video composing test, the Rx 480 is running at 108 FPS, putting is about 14 FPS below the GTX 980 Ti. This, however, is 18 FPS higher than the R9 380. In bitcoin mining, there is a huge range in performance between the cards tested. The RX 480 8GB performs okay here but the Nvidia cards rule the top of the charts. That is, of course, ignoring that you could nearly buy two of the RX 480’s for the price of one of those cards, but I will get into that here shortly.
Cooling, Noise, and Power
The last batch of testing isn’t as exciting as the in game testing, but it is all of the other important things to keep in mind when looking at video cards. To start off I tested the overall power usage of the RX 480. Well, to be specific I tested the power usage of our entire X99 based test bench with the RX 480 on it and compared it to a few other cards on the same bench. These are all peak wattage loads when running Valley Benchmark. Even though the GTX 1070 has the same 150-watt TDP as the RX 480, the RX 480 did pull less wattage. Part of this is most likely related to the 1070 card I tested also being overclocked. That said I am loving that AMD is starting to compete on power usage as well. Lower power usage is important in smaller builds where high power PSUs don’t fit and it also saves you money on the electric bill.
Next, I tested the reference RX 480 cooler’s noise performance at both 100% and 50% fan speed. Ideally, I would also test while in game here, but fan speeds while in game aren’t consistent. This gives us a better idea of the in-game performance at 50% and then 100% shows us how loud it can get when things get hot. The RX 480 wasn’t too bad at 50% fan speed but when I turned it up everything else in the office seemed quiet. That blower pumps out a LOT of air when it needs to and when that happens it gets a little noisy. Several decibels louder than even the Nvidia reference coolers.
Last I tested the cooling performance of the RX 480. I did this with two tests. The first test is running Valley Benchmark on loop and seeing where the RX 480 levels off at with the fan settings untouched. The next test I turned the fan up to 100% (put in my earplugs) and then tested again. This shows us how the card will perform while keeping in mind fan noise and how cool it can run if uncapped. At stock settings the RX 480 ran inline with all of the other reference cards. With the fan turned up though it did start to improve gaming one more degree between it and the GTX 1080. That said it is a little concerning if the GTX 1060 ends up with the same Founders Edition cooler while using less wattage than the 1070 and 1080 I have a feeling it is going to outperform the RX 480 given how close it is to the GTX 1080 here.
Overall and Final Verdict
So now that we have put the new RX 480 through the paces, taken a closer look at it, and had a quick run through on what Polaris is all about, what do I think of the card? Well for starters, for those of you who were expecting Polaris to come in and blow away everything on the charts, I’m sorry you are disappointed. The RX 480 is a great performing card for gaming at 1080p and it is even very capable at 1440p. It’s not a chart-topper, though, what is exciting about it is its performance for the price. The RX 480 does outperform a lot of cards that are priced well above it including getting very close to the GTX 980 in performance. It also does exactly what AMD promised with Virtual Reality as well, it brings VR Ready video cards to a price point that people can afford. Sadly getting the VR kit itself is still going to break the bank, though. I'm also liking the clean look of the reference cooler, it might be plastic and frankly, I’ve been getting spoiled with the metal fan shrouds, but it's going to look good in any PC.
It’s not all perfect. I wouldn’t recommend running the RX 480 with the fan cranked up. At 100% fan speed, the fan was loud enough that my wife and I had trouble talking to each other across the room.
What it all comes down to though is the price point. We know that AMD announced the RX 480 with a $199 price point for the reference design. They did confuse things a little bit, though. The same they sent me at least was an 8GB model. The $199 price point was only for the 4GB model, so we can’t really compare the performance tested here today to that price point. The RX 480 8GB has a price point of $239. I spent a little time browsing Newegg. The RX 480 8GB is currently positioned above the GTX 970 and just below the GTX 980 in performance from Nvidia and in line in a lot of my tests with the overclocked R9 390’s. Currently, the lowest price on a GTX 970 is $245 after rebate and the GTX is $334 with a rebate. With that in mind, we know the Rx 480 is a great value compared to what Nvidia currently has going on. I also double checked on the R9 390 as well because we know it performs closely just to see if any of the discounted cards would be slipping into the same price range of the RX 480, they are as low as $259.
The RX 480 ends up being a great value card. With that, it also looks like it could offer even more value when you look at running them in Crossfire but I’m hoping to get another card in and test that out before commenting on it. What I do know though is that people who use to have to make performance sacrifices when building low budget PCs now have a great option that is going to get them great performance at 1080p, something that wasn’t possible in the past without diving into game settings and turning things down.
Live Pricing: HERE