XFX RX 480 4GB Review: AMD Navigating in the Right Direction
Am I late to the Party? Maybe, but this is our Retail Review of the Radeon RX 480 4GB model from XFX. We really wanted to review the $199 card as it was what was shown off by AMD. As of June 29th 2016 AMD released the RX 480 to the world in mass quantities with both reviews and retail units hitting at the same time. Even with plenty of quantity ready to go we saw numerous retail and e-tail outlets selling out rather quickly. Not all sold out showing that the early claims of mass quantities being available on day one turned out to be true, which is a nice turnabout from previous launches where quantities were limited and restocking took forever. Well, I guess it’s time to get into the RX 480, Polaris 10, and what AMD’s strategy is with this interesting little entry in the mid range fray.
Instead of launching a high end product to compete with NVIDIA’s 16nm monster, the GTX 1080 or even going after the upper mid range GTX 1070, AMD shot right for the sweet spot of the market, the dead center mid range. According to AMD’s research in graphics card sales they found that 84% of all gamers are buying cards that range from $100 to $300 leaving only 16% to buy the ultra enthusiast line. While I can’t validate their claim I do know that the majority of people at the local LAN party that I attend regularly are sporting something along the lines of a R9 270x or GTX 760-960. So all in all, I think it’s fairly consistent with what I see out in my world. But when a person buys a graphics card in that price range they typically understand that they’re going to be getting middle of the road 1080p performance and not get all the bells and whistles of the higher end cards. Not quite true this go around as we’ll see later on with the gaming results. AMD is marketing this card as a high powered card capable of VR, now without a VR headset of my own I can’t quite validate these claims outside of the Steam VR Performance Test.
Improvements for the Polaris architecture go beyond just a move from 28nm down to the new 14nm FINFET architecture with GCN 4.0. Starting with the Primitive Discard Accelerator that takes the triangles early on as they move through the graphics rendering pipeline and removes the ones with no ‘inclusive sample points’ thus increasing performance when using MSAA.
Efficiencies to the core architecture have resulted in a 15% performance increase per compute unit when compared to the R9 290, allowing for similar performance with fewer stream processors. With the move from a 512bit bus down to a 256bit bus a lot of people were concerned with that move’s impact on performance as that wide bus has been a fighting point from Radeon since the launch of Hawaii. But, thanks to a very heavily tuned L2 caching behavior the cache and memory access has improved to the point that it shouldn’t prove impactful. Along with the new L2 cache system and the new memory controller the Polaris GPU supports up to 8Gbps GDDR5 natively and up to 256GB/s of memory bandwidth.
The display controller on Polaris received a much needed upgrade, but brings with it features that are ready for the future. The Fury line received much criticism for only having DP 1.2a and HDMI 1.4 support, but the Polaris line of GPUs have been updated to have Display Port 1.3 HBR3 Ready and 1.4 HDR Ready as well as HDMI 2.0b
AMD RX 400 Series Specifications
|Graphics Card Name||AMD Radeon RX 480||AMD Radeon RX 470||AMD Radeon RX 460|
|Graphics Core||Polaris 10 XT||Polaris 10 Pro||Polaris 11|
|Process Node||14nm FinFET||14nm FinFET||14nm FinFET|
|Peak Compute||5.83 TFLOPs||4.9 TFLOPs||2.2 TFLOPs|
|Memory||4/8 GB GDDR5||4/8 GB GDDR5||2/4 GB GDDR5|
|Memory Speed||8 GHz||6.6 GHz||7 GHz|
|Memory Bandwidth||256 GB/s||211 GB/s||112 GB/s|
|MSRP||$199 (4 GB)|
$239 (8 GB)
|$179 (4 GB)||$109 (2 GB)
We’ve updated our testing methodology to no longer include just averages and minimum frame rates for each game. Still including a minimum of three runs and averaging the results, we now include the average frame rates as well as the 1% and .1% low frame rates to better exemplify the smoothness of the game rather than relying on absolute minimums as they might only occur in a transition scene and go completely unnoticed to the gamer. We are also including the 99th percentile and 99.9th percentile frame times to show the latency at which the game plays at its worse, as always with that one the lower the better. Thermal testing is all done within the Cooler Master MasterCase 5 Pro with the two included Cooler Master intake 140mm fans set to 1000rpm and thermal readings were taken after 30 minutes of running Unigine Heaven at maximum detail settings at 1440p. Power Readings were taken as a total system power draw using a Kill-A-Watt with only the test system plugged into it, readings for idle were taken with nothing running on the desktop and load were taken while running the Futuremark 3DMark Firestrike Stress Test so focus as much of the load onto the GPU as possible.
Due to the concerns regarding power draw over the PCIe slot we did not include overclocking results in this review. I did manage a stable overclock of 1370mhz on the GPU core but found power draw to be entirely too high and the card nearing thermal throttle, even with a custom fan curve. I ran through Firestrike and found such an insignificant gain I wasn’t comfortable taking the risk on my personal test system. At this time I do not recommend overclocking this GPU until further notice or aftermarket cards are available. I do however recommend following our guide for lowering the voltages and improving the efficiency of the RX 480.
Intel Core i7 6800k (stock)
32GB CORSAIR Vengeance LPX DDR4 2400MHz
Cooler Master V1200 Platinum PSU
Cooler Master MasterCase 5 Pro
Crucial MX100 512gb SSD
Seagate 4tb SSHD
Monoprice 4k CrystalView Monitor
Cards used for testing
We have decided that going forward we will keep our GPU performance reviews primarily targeted on actual gaming performance, but we are keeping these two synthetic test for comparative purposes.
Running 3DMark Firestrike in Performance, Extreme, and Ultra yielded excellent results right out of the gate for the RX 480 matching almost perfectly our XFX R9 390 Black Edition
We ran Unigine Heaven at the highest settings at 1080p and 1440p. While the RX 480 falls short of what the R9 390 achieved it ran off and left behind the current cards in its price bracket
DX 11 GAME TESTING
Now that we have synthetic testing out of the way, we’re going to take a look at how the RX 480 holds up to gaming and where it falls in the current lineup of graphics cards.
Ashes of the Singularity
AotS was tested at all resolution with the ‘Crazy’ default setting using the in game benchmark. We could have easily gone with a lower setting to achieve higher FPS, but we wanted to beat these cards and see just what separates them.
The RX 480 came very close to the R9 390 in terms of performance but offered much smoother frame times resulting a better gaming experience, even with half the VRAM
We tested Battlefield 4 across 1080p and 1440p at the Ultra Preset during the opening driving sequence of the ‘Tashgar’ campaign mission to ensure consistency across all GPUs.
Interestingly enough our RX 480 fell quite a bit behind the R9 390 in this game, but still managed a healthy lead over the other cards in its pricing category.
Doom was tested with a 60 second manual run through the first part of the level ‘Foundry’. At 1080p and 1440p we used the Ultra quality preset and enable SMAA.
Prior to launch of the RX 480 you needed at least the R9 390 to get a great experience playing DOOM at these settings. The RX 480 even bests the R9 390 at this title and feels better while doing so.
DiRT Rally was run at 1080p and 1440p using the Ultra preset and using 2xMSAA. We used the lenght of the in game benchmark for our results.
The RX 480 fell behind the R9 390, but maintained a healthy lead over the class that it falls in.
Grand Theft Auto V
GTA V is a game that seemed to give other reviewers issues with the RX 480, but we didn’t experience any of the issues they reported, so we’re including screenshots of all of the settings we used in for our results.
We did however have an issue with the game crashing when trying to launch, with usually taking three tries to get the game to launch. We tested at 1080p and 1440p taking our results from the entirety of the last sequence of the in game benchmark utility.
While the RX 480 sits dead between the GTX 960 and R9 390 it’s not exactly impressing me here, but there could be room for improvement if other’s are experiencing technical issues.
Running Hitman at 1080p and 1440p at a mix of High/Ultra detail settings with SMAA enabled and using the built in benchmarking utility.
The RX 480s performance in Hitman was much better than I expected, coming within spitting distance of the R9 390 with ease.
Rise of the Tomb Raider DX11
Running RotTR was a little interesting, we could have gone with maximum settings here, but we stuck with the ‘High’ preset for consistency sake as going higher requires more than 4GB of VRAM. We took our measurements using the in game benchmark run at 1080p and 1440p.
Quite interestingly in this game the RX 480 matches the monstrous R9 390, even at 1440p while sporting half the total VRAM. We’re noticing as an interesting pattern begins to develop here.
The Witcher 3
We run The Witcher 3 at 1080p and 1440p with the all settings under processing and graphics set to ‘High’ and disabled ‘hairworks’. Our custom benchmark includes a horse ride along the hillside, through the forest, and a jog though Crow’s Perch.
The RX 480 Held up very well here coming just shy of the R9 390’s performance level, but leaving the 380 and 960 sitting far behind.
For DX12 gaming results we’re only able to focus on average frame rates as I am still working to better understand PresentMon so that I can deliver the same detailed results for DX12 as I am able to with DX11. We are also only focusing on 1080p results by comparison as they show the greatest variation when going from DX11 to DX12
Ashes of the Singularity DX12
AotS was run on the same setting for DX12 as we ran for DX11.
Hitman was run on the same setting for DX12 as we ran for DX11.
Rise of the Tomb Raider DX12
RotTR was run on the same setting for DX12 as we ran for DX11.
Total War: WARHAMMER
For TW:W we only have DX12 results. At the moment we are relegated to the benchmarking utility provide to us by AMD for testing this game. In the future we will have the full game so that we can test in both DX11 and DX12
VR TESTING, POWER DRAW and THERMALS
Our ability to test VR is very limited at the time as I do not own a HMD, so we’re reliant on the Steam VR Performance index, so we wanted to share where the RX 480 falls in that category.
Power Draw and Thermals
Yes, we do agree that the power draw is a little on the high side for this GPU, but see how you can change that here.
At the end of the day AMD has created something stellar, fitting that they named it after a star. No longer is the $199 price point something that comes with the understanding of turning down the details to achieve a great gaming experience. I know there are going to be lot of people out there who were expecting more. But let’s be honest here, this is a $199 graphics card that is performing nearly on par with a R9 290x, a card that cost $549 when it launch and spat out 94*C thermals while sounding like a Cessna having a fit while funneling power. When comparing it to cards currently at this price, well…you can’t its way faster than both the R9 380 and the GTX 960 so we simply have to compare it to cards like the R9 390. There are legitimate concerns when referring to the power draw of the card, specifically through the PCIe slot, but it seems that AMD is confident they can fix that for the first wave of cards through the driver, and hopefully a permanent fix will be in place. We’ve shown that a little tweaking can even increase the efficiency of the card and get a slight performance bump in the process.
Basically, this card is an Out-of-the-Box dream, but an overclocker’s nightmare. Don’t expect much gain from overclocking this card, other than gains in how much power you’re drawing for so very little return. The small gains that overclocking yields interest me in waiting to see how things change when the AIBs start launching their fully custom version of the RX 480.