RX580 8GB Red Dragon Review
Introduction
Greetings everybody I know stuff from me is like waiting for busses right now isn’t it? You spend ages waiting for one then two come along at once. Today for this peoples review I thought we would take a look at the PowerColor RX580 8GB Red Dragon. Usually I’d never bother with a review for a graphics card that’s the age the RX580 is now but I’ve had this thing sitting around forever and always been curious about it and with the RX590 being nothing but an overclocked marginally die shrunk RX580 I figured what the hell, let’s take a look anyway. So here we are, let us welcome the PowerColor Radeon RX580 8GB Red Dragon!
Gallery
Let’s get to it and see what this card looks like.
Well, first impressions are quite positive. The packaging is compact and offers good protection, the cooler looks fairly capable on first glance and there’s a backplate. Unfortunately the latter is only there for rigidity and serves no purpose beyond that and we can already see the cooler does not do anything to cool the memory at all beyond airflow. You might notice PowerColor (TUL) have chosen to put a warranty sticker right over one of the screws meaning you won’t be able to do basic maintenance like replacing thermal materials or a thorough de-dusting, that’s not cool PowerColor, not cool at all and I’ve already decided I’m docking you points for this. In terms of looks the Red Dragon is doing ok for itself but considering the red and black theme I’m surprised there’s no red on the card itself at all. Come on, would a bit of red on the card somewhere really cost THAT much PowerColor to keep the theme consistent?
Having a look at the IO plate shows the usual suspects, DisplayPort connections, HDMI and DVI are all here however I would have liked to see both DVI-I and DVI-D outputs. A bit of a disappointing ending to a fairly bright start to be honest, let’s dig in and strip this puppy down.
The fans utilised on the heatsink are dual ball bearing 87mm Power Logic PLA09215B12H with a 0.55 amp current and a max RPM of around 3800. Obviously by the stats of these fans they are meant to pick up the slack of the fin-starved heatsink – we’ll get to that in a moment, which means this cooler has the potential to get bloody loud, and it will start to do just that at around the 65% fan speed mark.
Meanwhile taking a look at the heatsink straight away there’s problems, fin density is really rather low despite the three large heat pipes due to the fact this heatsink isn’t just going to have to cool the GPU itself but the VRMs as well. This card is going to get like a little blast furnace extremely quickly particularly when you start overclocking, it comes as no surprise to me PowerColor didn’t add the burden of cooling the memory as well to this heatsink after seeing this, I’m also really not fond on “built-in” VRM cooling to the main heatsink this just makes putting the card under water, LN2, or hell even a good 3rd party heatsink a real pain in the arse. Despite these things if that fin density were to be improved by a firm 10% I’d say the cooler would be a fairly competent one.
Sigh. When will manufacturers start making good heatsinks? Often it would require only small changes here and there to turn these very mediocre solutions in to something really rather good.
Moving on to the PCB its.. actually well made, which makes the rather poor cooler design / choice by PowerColor all the more bizarre. The memory is Samsung K4G80325FB -HC25, rated for 8Gbps, while the MOSFETs are On-Semi 4C08N YEGH70 rated for 52A with an operating temperature between -55c to +150c and On-Semi 4C03N RFL1U (NTMFS4C03N?) rated for 136A with an operating temperature between -55 to +150c while the voltage regulator is the Infineon IR3537/CHL8510 (Chill 8510) the PCB also sports a 6 phase design and I didn’t see any phase doubling going on each phase has a driver for itself. There’s even a space to add a 6 pin PCIe connector so I’m going to bet this PCB is a slightly cut down version which is used on PowerColors Red Devil line. Lastly you might notice some black resin around a couple of the Samsung memory ICs, I’ve not seen any other manufacturer do this but I have seen PowerColor do it on several occasions my guess would be PowerColor think a few ICs might get toastier than they should do so use the resin to help prevent any BGA joints breaking that would otherwise require the card to be RMA’d for a reflowing.
In summary, crap cooler, good PCB. PowerColor are weird.
Technical Specifications
Now we have taken a look at the cooler and card itself it is time to have a peek at what this RX580 is packing under the hood.
Nothing out of the ordinary to see here compared to every other 8GB RX580 in the world but I will take the chance to comment on the low pixel fillrate, it’s only really important for anti-aliasing these days but that fillrate is on the low side rendering (yes pun intended) the RX580s anti-aliasing capabilities rather limited. Also, I still don’t like “Adrenaline” not being spelt properly. These are the people we have to trust have competent coders, everybody.
Test Setup
We should all be familiar with the test system now but here we go again...
CPU: AMD Zen 1700 @ 3.9GHz
Mainboard: MSI X470 Gaming Pro Carbon
RAM: 2x8GB G.Skill Ripjaws V 3200MHz 15-15-15-35 @ 3466MHz CL16
GPU: Pwercolor RX580 8GB
Storage: 250GB Hynix SL301 SATA SSD, WD 120GB M.2 SATA SSD (OS drive), 2TB Seagate Barracuda
Opticals: 24x Lite-On iHAS324 DVD-RW, 16x HP BH40N Blu-Ray
Sound: Xonar DX 7.1, Realtek ALC1220
PSU: EVGA 1000w Supernova G2
OS: Windows 10 Pro x64 (latest ISO) and all updates
Case: NZXT Phantom 530
Drivers:
AMD Chipset Drivers v18.10.0830
Adrenaline 19.5.2
Realtek Nahimic HD Audio Driver v6.0.1.8627
Xonar C-Media 1825 Driver
Radeon Settings
Radeon Settings really doesn’t get enough attention or even a mention in reviews and it should if you ask me as you can't have good hardware without good software, for all the bugs and glitches Radeon Settings has (especially WattMan) it is still a fully featured comprehensive piece of software so let’s have a few screenshots of it.
As you can see, Radeon Settings is well organised, compact, and all integrated very nicely far better than what nvidia are doing with their control panel and GFE. I do have to mention though that Radeon Settings is doing an awful job of detecting games installed on the test system, most of them are on Steam and the only game from Steam RS is detecting is Metro 2033.
Benchmarks
That was quick wasn’t it? We’re already at the benchmarks. This isn’t a case study or motherboard review, there’s a lot less to go over. All game benchmarks have been run at 1920x1080 as this is undoubtedly the most common resolution for this card with maximum quality game presets being used. Where a preset is not available all in-game graphical settings have been set to maximum. OK, now it’s off to the races.
Swings and roundabouts is how we start off for the Fire Strike test the RX580 takes the overall and graphics results while the GTX980 has the physics and combined results. I expected the RX580 to have a bit stronger start if I’m honest.
Squeaky bum time for the RX580 as it manages to sneak ahead of the much lower clocked GTX980 for Time Spy. There’s not much in it at all and would be indistinguishable gaming but a win is a win so give the RX580 a pat on the back for that.
Swings and roundabouts again with the theoretical API overhead results showing the RX580 being ahead in Vulkan and DX12, while the GTX980 beats the RX580 in DX11 tests. The short answer here is that AMD appear to have single and multi-threaded dynamic batching problems under DX11, not good news if you have an AMD GPU for current and upcoming DX11 titles or ones that use an API wrapper to convert DX11 calls to DX12. API overhead is going to be something I will be including for all GPU reviews I write from this point on due to this issue.
The last stock synthetic test is LuxMark, those extra compute units give the RX580 an advantage here but the RX580 does have 125% more CUs than the GTX980 but only manages a score that is about 23% higher. That’s really not very impressive for all those extra CUs.
Time to check out some games hopefully we’ll start to see a change of pace.
Performance here isn’t bad for the RX580 but ultimately the card doesn’t give you anything to get excited about either while at the same time falling short and losing out to the much older GTX980 in Alien Isolation which on PC is an AMD Gaming Evolved title has to be considered a disappointing showing but at least the RX580 manages to respond in kind for Tomb Raider which is an nvidia “the way its meant to be played” title. Metro Last Light is another title where the RX580 disappointingly once again falls short, does anyone know if AMD RTG has ever bothered optimising for the Metro titles? I certainly don’t recall anything of that nature.
Trading blows has been the theme of things for these stock results hopefully some overclocking will bring a nice surprise, because there certainly hasn’t been any so far.
Overclocking
As expected the stock cooler began to scream once overclocking came in to play GPU and VRM temperatures rocketed it was like having an oven in the system there’s just not enough of the cooler to adequately control the cards temperatures when you start stressing the card for prolonged periods even with the stock GPU voltage of 1.15v but let's see what we were able to get.
This particular RX580 has been a somewhat backward affair the core didn’t want to realistically do more than 1.425GHz even with 1.2v but was happy to undervolt from the stock 1.15v to 1.1v at the overclock frequency, that puts the core in average OC territory so this result wasn’t completely terrible at least. No amount of coaxing would get the memory past 2075MHz (8300MHz effective) error free though, given the quality of memory on this sample that leaves only one culprit for such a poor memory overclock; a weak IMC. Suffice to say marginal clock speed bumps like this are going to provide about as much affect on performance as painting red stripes on the card would, I needed to get more creative. This is where GDDR5 memory timing optimisation came in to play. The IMC may have scraped the RX580 grade in the same manner the Titanic scraped that iceberg but I wasn’t about to let this card go down without a fight fortunately with Samsung memory on this sample prospects here seemed promising as it’s not very likely there’s a weak Samsung IC here. In the end I managed to rather substantially tighten timings and keep the maximum memory clock that was attainable with the spectacularly terrible stock timings. There’s only so much time I could spend sorting out AMD RTGs rather, well, poor attempt at configuring memory timings but I am confident about as much performance as there is available has been eeked out in order to put the cards best foot forward.
Let’s find out how all of that memory tweaking and overclocking helped the RX580.
The synthetic tests show a nice uplift for the RX580 the card is able to just barely keep up with a highly overclocked GTX980 but let’s see what real games have to say first before we start singing any praises.
Moving to the gaming tests the RX580 manages to pick itself up a bit with an improvement of around 9% for Metro, 7.2% for Alien Isolation, 10% in Deux Ex and 5% in Tomb Raider. That’s not a bad improvement to see from what is mostly memory timing optimisation as this sample is almost a complete lemon for overclocking but ultimately the RX580 can’t compete at all in the tested DX11 titles losing by 10% in Metro, or 9.4FPS, and about 10% again in Alien Isolation, or 15.4FPS, that’s certainly no small margins and is indicative again of AMD RTG having much work to do for performance in DX11 titles. If only AMD RTG optimised their drivers as well for games as they seem to do for benchmarks.
Finally, at last we get a result that is much more befitting of the hardware the RX580 is packing, it only took until the very final test to show there is good performance lurking in the card somewhere. Those optimised memory timings are showing a much better balanced result for the extra CUs the RX580 has those 125% additional CUs are now pulling a performance improvement of about 45% compared to the overclocked GTX980 that’s certainly nothing to scoff at in the slightest.
Conclusion
The curtain starts to come down on another peoples review so with that curtain it is time to wrap things up.
Let’s begin by talking in a broader perspective about the RX580, and by extension RX590 which not only struggles to pull away from the RX580 it’s meant to replace but only has an increased GPU speed and tiny node shrink from 14nm to 12nm while being even more of a power hog.
The AMD RTG Polaris architecture does use a substantial amount of power compared to nvidias Pascal and Turing architectures making an out of the box Polaris card somewhat of a hard sell however you get extra on-board vRAM to help balance this a bit which with texture sizes growing by the game now is something to consider there is nothing worse than running out of frame buffer and seeing micro stutter and / or performance outright tanking as a result. If you are a tinkerer then the RX series will also offer you a more appealing option with vBIOS tools available to fine tune voltages to help bring power draw down and to adjust GDDR5 memory timings that you can then directly set and apply to the card along with a whole host of other abilities which make the RX cards rough diamonds and a tinkerers playground if you are willing to put the time in. Radeon Settings is also nice and easy to use with lots of control available through it and if you do anything compute heavy you will see a good jump in performance by tightening those GDDR memory timings. Unfortunately for the Radeon Technologies Group this is where the positives end as the RTG drivers leave considerable room for improvement which undoubtedly hold all Radeon products back particularly in DX11 titles and also apparently lacking any kind of driver level multi-threading capabilities as witnessed in the API overhead test.
That brings us to this particular PowerColor RX580, there is absolutely no doubt in my mind this particular sample should never have qualified as a passable RX580 the IMC is already pushed to the limit right out of the box which could have an impact on component longevity on the card and the cooler PowerColor have used really is rather poor being obnoxiously loud and not particularly good at the job it’s meant to be doing turning this poor card in to a little blast furnace in your system with any prolonged or demanding load. At this point you’re probably picking up on my unenthusiastic indifference for the PowerColor RX580 Red Dragon, not once did this toothless dragon do anything to impress or even try to stand out I’ve not seen a more underwhelming card since the HD5830.
Now, as a responsible reviewer I did put PowerColor as a company to the test, it is an incognito test that varies but something I do to every manufacturer for something I review. I contacted PowerColor both through email and social media (after not getting any reply to my email after two weeks) with a simple question asking if there was an updated vBIOS for the card. A simple test to act as a gauge to if the company potential customers would be buying from actually care about said customers. PowerColor failed this simple test spectacularly. I know somebody at the company saw my query through social media but there was never an answer. From this the only conclusion that can be drawn is that PowerColor don’t give a damn once they have your cash, what happen to you PowerColor? You used to be a respected manufacturer highly spoken and thought of. Good luck getting a response out of PowerColor should you need to RMA through them as the Red Dragon does supposedly come with a 3 year warranty. A warranty I would suspect that is not worth the paper it’s printed on.
Well, it’s time to get to scoring and I can tell you the Red Dragon is going to fall well short of even a Bronze award.
Aesthetics: 5 / 10
Hardware Quality: 15 / 20
Cooling: 10 / 20
Performance: 6 / 10
Overclocking:: 5 / 10
Warranty Period & Product Support: 12 / 30
Final Score: 53%
My final words then, typically AMD RTG drivers are mediocre at best in any title the driver has not specifically been optimised for. Polaris is a great architecture if compute tasks is your main focus it’s just not much of a pure gamers card there are better alternatives from a gaming performance perspective you could pick up second hand for similar money such as the GTX980TI that still has a reasonable 6GB of frame buffer. All of that is not to say an RX580 or 590 is a bad choice, it’s just not the first choice.
--------------------------------------------------------------------------------------------------------
If you liked this review then consider checking out more of the work I've done Right Here and consider dropping me a paypal donation doing so shows your love and helps me to write the articles and reviews you enjoy. Spreading these articles around will also help me and if there is anything you would like me to check out and write about contact me at ket26[at}hotmail[dot}co[dot}uk.