Final Assembly Photos
We learned that the RX 480 has a 6-phase VRM and uses Nichicon 2K capacitors. The VRM seems to be of better design than the rest of AMD's card can accommodate, mainly its supply of power. With only a 6-pin header, we're completely maxing-out GPU power when overclocking (hitting 192W GPU power), and the card is definitely capable of more than that. We'd speculate that AMD may have over-designed the VRM so that the reference PCB and VRM could be used by AIB partners, who then would improve cooling and offer a better product. That wouldn't resolve the power supply, though; not unless board partners replace the 6-pin with an 8-pin header or 6+8-pin setups, and the latter would likely accompany custom board designs.
We also learned that the reference design has solder points for DVI, allowing board partners to make modifications as desired.
And, more importantly, we learned that the reference cooler is weak. That's what we're here to fix. Without any heatpipe or vapor chambers, the reference cooler is just an aluminum finned heatsink (with wide fin spacing, for that matter, so there's little surface area) with a blower fan. The VRM is cooled by thermal pads secured to the heat plate, which has its conducted heat dissipated by the blower fan.
Taking apart the RX 480 is trivial, and can be done by removing a set of screws on the back-side. The cooler and shroud fall off the PCB after this, and top screws (near the Radeon branding) can be removed to reveal the heatsink and separate the heat plate from the shroud.
We ended up mounting the pivotal VRM push fan over the VRM & single unsinked VRAM module. A second fan was clipped to the test bench and pointed down at the VRM and VRAM modules, to further dissipate heat.
RECAP: Previous Clock vs. Temperature Tests (Stock Cooler)
Here's a look at our original clock-rate vs. temperature (& fan RPM) charts that we produced for the RX 480 review.
Text is also pasted from the review, for this brief section:
The clock-rate is a little shaky at first as dynamic clock fluctuations range from 300MHz to the boosted 1266MHz, but this is stabilized as the card ramps into its workload.
Let's zoom-in on one of the spikes to better detail the “amplitude” of our line:
This chart is cropped to 4200-4600 seconds of the test. We see fluctuations from ~1150MHz to ~1240MHz, which is nearly a 100MHz range. That's enough to cause a very brief dip in 0.1% low and 1% low frametimes (both of which improved with our core-locked overclocking), but is otherwise easily overlooked during gameplay. These fluctuations happen so rarely that the impact is only occasionally noticed – we're talking a few times per hour, if that – and only if the user is looking for such fluctuations.
Here is a look at some of the data from the above slice:
Most of this fluctuation is a result of AMD's new dynamic clock, which adapts to pre-defined states in WattMan as workload changes. The rest is resultant of thermals – we're seeing some throttles once temperatures exceed 80~82C, but this can be overridden in Wattman by increasing the max temperature. Of course, fan RPMs can also be increased.
AMD's clock-rate stability is reliable and steady, something for which we commend the company's work on its new process and power management.
And for the curious, here is a chart comparing the fan RPM to the temperatures. Again, this is with all settings stock – so you're looking at out-of-box fan RPM & temperature target performance:
The red line represents temperature (plotted against the right axis, absolute temperature *C). The blue line represents frequency in MHz (plotted against the left axis). The orange line represents fan RPM (plotted against the left axis, in RPM).
In order to sustain the card's ~82C resting area and 1200~1300MHz stock clock-rate (boosted – so it's dynamically fluctuating), the blower fan must spin at approximately 2000~2300RPM. The specific sound of this fan is demonstrated in our video review that's embedded on the first page, but we've also got some objective metrics.
We tested using our GPU test bench, detailed in the table below. Our thanks to supporting hardware vendors for supplying some of the test components.
The latest AMD drivers (16.6.2 RX 480 press) were used for testing. NVidia's unreleased 368.39 drivers were used for game (FPS) testing. Game settings were manually controlled for the DUT. All games were run at presets defined in their respective charts. We disable brand-supported technologies in games, like The Witcher 3's HairWorks and HBAO. All other game settings are defined in respective game benchmarks, which we publish separately from GPU reviews. Our test courses, in the event manual testing is executed, are also uploaded within that content. This allows others to replicate our results by studying our bench courses. In AMD Radeon Settings, we disable all AMD "optimization" of graphics settings, e.g. filtration, tessellation, and AA techniques. This is to ensure that games are compared as "apples to apples" graphics output. We leave the application in control of its graphics, rather than the IHV.
Windows 10-64 build 10586 was used for testing.
Each game was tested for 30 seconds in an identical scenario, then repeated three times for parity.Average FPS, 1% low, and 0.1% low times are measured. We do not measure maximum or minimum FPS results as we consider these numbers to be pure outliers. Instead, we take an average of the lowest 1% of results (1% low) to show real-world, noticeable dips; we then take an average of the lowest 0.1% of results for severe spikes.
|GN Test Bench 2015||Name||Courtesy Of||Cost|
|Video Card||This is what we're testing!||-||-|
|CPU||Intel i7-5930K CPU||iBUYPOWER
|Memory||Corsair Dominator 32GB 3200MHz||Corsair||$210|
|Motherboard||EVGA X99 Classified||GamersNexus||$365|
|Power Supply||NZXT 1200W HALE90 V2||NZXT||$300|
|SSD||HyperX Savage SSD||Kingston Tech.||$130|
|Case||Top Deck Tech Station||GamersNexus||$250|
|CPU Cooler||NZXT Kraken X41 CLC||NZXT||$110|
For Dx12 and Vulkan API testing, we use built-in benchmark tools and rely upon log generation for our metrics. That data is reported at the engine level.
Video Cards Tested
- AMD RX 480 8GB ($240)
- NVIDIA GTX 1080 Founders Edition ($700)
- NVIDIA GTX 980 Ti Reference ($650)
- NVIDIA GTX 980 Reference ($460)
- NVIDIA GTX 980 2x SLI Reference ($920)
- AMD R9 Fury X 4GB HBM ($630)
- AMD MSI R9 390X 8GB ($460)
Thermal Test Methodology
We strongly believe that our thermal testing methodology is the best on this side of the tech-media industry. We've validated our testing methodology with thermal chambers and have proven near-perfect accuracy of results.
Conducting thermal tests requires careful measurement of temperatures in the surrounding environment. We control for ambient by constantly measuring temperatures with K-Type thermocouples and infrared readers. We then produce charts using a Delta T(emperature) over Ambient value. This value subtracts the thermo-logged ambient value from the measured diode temperatures, producing a delta report of thermals. AIDA64 is used for logging thermals of silicon components, including the GPU diode. We additionally log core utilization and frequencies to ensure all components are firing as expected. Voltage levels are measured in addition to fan speeds, frequencies, and thermals. GPU-Z is deployed for redundancy and validation against AIDA64.
All open bench fans are configured to their maximum speed and connected straight to the PSU. This ensures minimal variance when testing, as automatically controlled fan speeds will reduce reliability of benchmarking. The CPU fan is set to use a custom fan curve that was devised in-house after a series of testing. We use a custom-built open air bench that mounts the CPU radiator out of the way of the airflow channels influencing the GPU, so the CPU heat is dumped where it will have no measurable impact on GPU temperatures.
We use an AMPROBE multi-diode thermocouple reader to log ambient actively. This ambient measurement is used to monitor fluctuations and is subtracted from absolute GPU diode readings to produce a delta value. For these tests, we configured the thermocouple reader's logging interval to 1s, matching the logging interval of GPU-Z and AIDA64. Data is calculated using a custom, in-house spreadsheet and software solution.
Endurance tests are conducted for new architectures or devices of particular interest, like the GTX 1080, R9 Fury X, or GTX 980 Ti Hybrid from EVGA. These endurance tests report temperature versus frequency (sometimes versus FPS), providing a look at how cards interact in real-world gaming scenarios over extended periods of time. Because benchmarks do not inherently burn-in a card for a reasonable play period, we use this test method as a net to isolate and discover issues of thermal throttling or frequency tolerance to temperature.
Our test starts with a two-minute idle period to gauge non-gaming performance. A script automatically triggers the beginning of a GPU-intensive benchmark running MSI Kombustor – Titan Lakes for 1080s. Because we use an in-house script, we are able to perfectly execute and align our tests between passes.
Noise Testing Methodology
Our noise testing methodology is new and still being revised, but has been kept consistent across all tests contained herein. We test noise in a real-world environment and do not presently use an anechoic chamber. The results align with what consumers will encounter in their own rooms.
We use a REED logging dB meter mounted to a tripod, whose mic is positioned 20” from the face of the GPU (mounted in an open bench). The REED meter is approximately 6” above the bench. All open bench fans are disabled. The Kraken X41 CPU cooling fan is configured to its “silent” mode, minimizing its noise output to be effectively imperceptible.
A noise floor measurement is taken prior to each test's execution to determine ambient without any systems running in the room. We then take an idle measurement (GPU & CPU at idle). Our noise floor has a fluctuation of approximately +/-0.6dB.
Noise levels are logarithmic, and are therefore not as simple to perform delta calculations as thermals or framerates. Noise percent differences are calculated using dB=20*log(V2/V1) (where V is amplitude). You cannot perform a simple percent difference calculation to determine the delta. For an example, a 10dB range (50dB vs. 40dB) is not equal to a 22% delta.
After the noise floor is determined, we log idle fan dB, 50% speed dB, and 100% speed dB (configured in Afterburner). We also measure auto fan dB at an identical stepping for every test; we do this by running Kombustor for exactly 5 minutes prior to beginning dB logging, which is useful for fans which use two push fans. Some dual-push fan cards will only trigger the second fan if the VRM is under load.
DIY RX 480 'Hybrid' Averaged Peak Thermals
The Arctic Accelero Hybrid III brings the RX 480 down to 23.02C delta T load, a 59% thermal reduction from the reference cooler (or an 83.96% difference, depending on what comparative method you want to use). The reference cooler operates at 56.33C load. This mod brings the RX 480 close to what our 18C GN Hybrid GTX 1080 was capable of, the main difference being that the 1080 Hybrid used an EVGA Hybrid cooler. That cooler has a copper extrusion in the coldplate, which helps sink heat prior to heat whicking and dissipation. It's also a different architecture that behaves differently under load and idle than the AMD Polaris / 14nm architecture. Idle temperatures are low thanks to the massive aluminum heatsink we planted against the rear of the PCB and the two fans blowing directly down on the card (plus the normal case fans).
Absolute values with the Hybrid III solution are within the 40s to 50s when running at the stock clock, rather than the 80 to 85C area of the original cooler – and that's maintained with much lower fan speeds. That means our noise levels are lower than with air.
Let's test that theory.
This shows the thermals of the RX 480 Hybrid at 100% and at 30% fan speed. At 100% with the radiator fan, we're operating at the 23.02C value mentioned previously. Reducing fan speed to 30% (and thereby reducing noise – we'll show that impact below) yields a temperature output of 38.66C. That's still well below the reference RX 480 cooler – nearly 50 degrees, actually – and allows us to reduce our noise emissions.
DIY RX 480 'Hybrid' Thermals Over Time
Here's two charts visualizing thermals over time. The first is a bit more flooded with cards. The second simplifies things with the GTX 1080 & 1070 FE cards and the Hybrid mods.
GN's RX 480 Hybrid vs. Reference VRM Fan Speed & Noise
At 30% fan speed (remember – we're at 38.66C here, so about 48C lower than reference), our noise output is ~38.5dB (delta vs. environment). At 100% fan speed, where we maintain delta T of 23.02C, we're at 42dB. That's quieter than the RX 480 stock at 42.8dB.
If you're overclocking the RX 480, there's a good chance that the VRM blower fan hits ~3800-4100RPM. That's more than 50% speed (max speed is 5200RPM, so 50% is 2600RPM) and produces a loud, stressful sound with a shrill whine from wind passing through the card. The liquid cooling solution fixes that, and so would using an AIB partner cooler with push fans and a real heatsink, not the scrap metal that was mounted to the RX 480.
Here's a chart with some other cards, for comparison:
Original RX 480 Reference Overclocking
Here's our original overclock stepping progression for the RX 480 Reference card:
|Core Clock (MHz)||Mem CLK (MHz)||Mem Offset (MHz)||Power Target (%)||Voltage||vCore Offset||5m Test||60m Endurance||Fan target|
We hit 1340MHz core and 2200MHz memory, and could not progress further.
New DIY RX 480 'Hybrid' Overclocking
And here's the progression for the new RX 480 GN Hybrid:
|Core Clock (MHz)||Mem CLK (MHz)||Mem Offset (MHz)||Power Target (%)||Voltage||Max Voltage||5m Test||60m Endurance||Power Draw|
|(switch VRM fan to go to PSU)|
We were hoping that we'd get some additional overclocking headroom out of the liquid cooler – and we did, thanks to overall reduced temperatures. We even tested with the new VRM fan plugged into the board and unplugged, but saw no real difference in maximum OC. The extra few Watts didn't really do much for us.
Here's a look at the new overclock stepping table. Previously, we got stuck with 1340MHz core and 2200MHz memory, and were forced to run the fan at around 4000RPM just to keep the thing at 89C. With the GN Hybrid solution, we ended up at 1390MHz core and stuck at the same 2200MHz memory clock.
We tried pushing the core clock up to 1400MHz and it did survive an initial five-minute test, but it failed right around that five-minute mark. This is when we unplugged the VRM fan from the card, hoping to get some more stability, and even amped-up the RPM of all the fans in the system to improve cooling... but no luck. We can't claim to have hit 1400MHz. Not with stability, anyway.
Now, that said, the RX 480 Hybrid did hit and slightly exceed 1400MHz when we were testing with dynamic frequency ranges and with synthetic applications like FurMark, but we can't make it happen for real-world gaming or applications, and so it doesn't count. 1390MHz is the final clock-rate for this card, which is a 50MHz increase over our original 1340MHz overclock that we burned-in for a few hours with the stock cooler.
This mod did see us peaking voltage sort of crazy high, though. Maximum board power draw was 192.2 Watts, according to GPU-Z, and the maximum voltage was 1.15V. That's pretty damn high for a GPU and isn't healthy for long-term use, so we wouldn't recommend this high of an overclock for extended periods. That is, not if you want your card to live for a while.
RX 480 Overclocked FPS Benchmark
Framerate impact is not much. In GTA V, we're still seeing the bad stuttering issues we discovered with the 16.6.2 press drivers for the RX 480 – but the FPS overall is marginally improved. We're now at 91.5FPS instead of 89.5FPS for 1080 Ultra over four passes, or a gain of about 2.23%. That's not a linear gain from our original OC, which had us at 7.26% gain from stock.
The same non-linear gain is mapped across Shadow of Mordor, where we saw a perf jump of 3.7% versus the stock-to-1340MHz jump of 7.9%, and of Mirror's Edge, where we saw a big fall from the original 10% performance improvements. But there's still improvement overall.
As for why it's non-linear, that's because of our method of overclocking. Going from stock to the first OC, we used WattMan to flatten the clock-rate to a hard 1340MHz (constant). This is much different from the boosting behavior of a stock RX 480, or any stock GPU right now, as it removes variability of the clock-rate and demands a fixed speed. That's where most the speed boost came from.
Conclusion: GN's RX 480 Hybrid
Building the RX 480 Hybrid was a lot of fun. We bought the Arctic Cooling kits out of pocket (several of them, in fact, to make sure one worked), and they were surprisingly fun to work with. We'll probably do more of these in the future.
But it's absolutely not worth doing if you're just seeking extra FPS. The gain is not there. And the extra OC gain is marginal compared to what may potentially be possible on future AIB partner cards that have more power budget.
As for thermals and noise, that's where the argument changes. The thermal reduction is massive – 59% – and allows us to drop the radiator fan to just 30% RPM (~800RPM) for noise reduction while maintaining superior thermals to reference.
It was just a fun project, though. That's all this really was: An experiment. And a good one, too. We've learned a lot about the RX 480 along the way.
Editorial: Steve “Lelldorianx” Burke
Video: Andrew “ColossalCake” Coleman