AMD Radeon R9 285 Review: Tonga and GCN Update 3.0

Test Setup and Benchmark Suite

Power Consumption Measurement Methodology

Tom’s Hardware Germany’s test system for the exact measurement of graphics cards’, CPUs’, and other components’ power consumption was developed in cooperation with HAMEG (Rohde & Schwarz). It was designed for particularly precise measurements with very small time intervals and has a temporal resolution of up to 1 ms.

Only sophisticated technology is able to handle the challenges presented by AMD’s PowerTune and Nvidia’s Boost technologies. These generate changes in core voltage in time frames of under 10 ms, which results in very large and quick voltage fluctuations. Let’s take a look at what can happen in the space of only one single millisecond, using measurement intervals of 10 μs.

This is why we’re evaluating all measured currents and voltages with a 500 MHz four-channel oscilloscope with a data logger, the HAMEG HMO3054, and extremely fast current probes. This setup also allows for unified data storage and remote control.

The measurements provided by the three high-resolution DC current probes, all HAMEG HZO50s, are taken via a riser card for the 3.3 V and 12 V rails, which we constructed specifically for this purpose. It supports PCIe 3.0 and uses short signal paths. The remaining probes are connected to the PCIe power cable that we modified for this.

We measure the voltages directly at their respective rails. We’re now working with a temporal resolution of 1 ms, since this allows us to record and evaluate the fluctuations caused by AMD’s PowerTune and Nvidia’s Boost technologies with confidence.

We’ve limited the measurement duration to 1 minute due to the very high volume of data when we measure all channels. We only shorten the measurement intervals all the way to the minimum that the physical capabilities of our setup allows for our more detailed measurements.

Test Methodology
No Contact Current Measurement at All Rails
Direct Voltage Measurement
IR Real-Time Monitoring
Test Equipment
1 x HAMEG HMO3054, 500 MHz Four-Channel Oscilloscope with Data Logger
4 x HAMEG HZO50 Current Probe
4 x HAMEG HZ355 (10:1 Probe, 500 MHz)
1 x HAMEG HMC8012 DSO with Data Logger
1 x Optris PI450 80 Hz Infrared Camera + PI Connect
Test System
Intel Core i7-5960X
MSI X99 Gaming 7
16GB G.Skill Ripjaws DDR4 2666 (4 x 4GB)
Samsung 850 EVO 512 GB
Raijintek Water Cooling
be quiet! Dark Power Pro 1200W
Microcool Banchetto 101

Benchmarking Hardware And Software

The Radeon R9 285 was tested with the 14.8 beta launch driver, but all other AMD cards were outfitted with the Catalyst 14.7 RC1 for testing. The GeForce cards used the newest option, which at the time of testing was the  340.52 WHQL driver.

We selected a variety of newer game titles with high detail settings at a resolution of 1920x1080 in order to give the Radeon R9 285 and its competitors a solid, real-world workload that this class of card should be able to handle.

The Asus Radeon R9 285's core clock was dropped to the 918 MHz reference specification in order to show what a typical Radeon R9 285 should be able to accomplish. Keep in mind that there is no reference cooler for this card, so all Radeon R9 285 will be unique in this respect.

Some readers will note that the Radeon R9 280 results are those we collected from our Sapphire Dual-X Radeon R9 280 review.  Sapphire's card comes with a 940 MHz core clock, a mere 7 MHz over the reference specification. We took a few benchmarks with the reference clock but quickly realized that the results are within the margin of error, so we're using the Sapphire numbers to represent reference Radeon R9 280 results. Keep in mind that there is no reference Radeon R9 280 cooler, either.

Two of the games we're testing have an option to use a Mantle code path, so we're running those benchmarks (Thief and Battlefield 4) with Mantle enabled and disabled to measure the API's impact.

Graphics cards like the Radeon R9 280 require a substantial amount of power, so XFX sent us its PRO850W 80 PLUS Bronze-certified power supply. This modular PSU employs a single +12 V rail rated for 70 A. XFX claims continuous (not peak) output of up to 850 W at 50 degrees Celsius.

We've almost exclusively eliminated mechanical disks in the lab, preferring solid-state storage for alleviating I/O-related bottlenecks. Samsung sent all of our labs 256 GB 840 Pros, so we standardize on these exceptional SSDs.


Test System
CPU
Intel Core i7-3960X (Sandy Bridge-E), 3.3 GHz, Six Cores, LGA 2011, 15 MB Shared L3 Cache, Hyper-Threading enabled.
Motherboard
ASRock X79 Extreme9 (LGA 2011) Chipset: Intel X79 Express
Networking
On-Board Gigabit LAN controller
Memory
Corsair Vengeance LP PC3-16000, 4 x 4 GB, 1600 MT/s, CL 8-8-8-24-2T
Graphics
Asus Strix Radeon R9 285
954 MHz GPU, 2 GB GDDR5 at 1375 MHz (5500 MT/s)
(underclocked GPU to reference 918 MHz specification for benchmarks)

AMD Radeon R9 280X
850/100 MHz GPU, 3 GB GDDR5 at 1500 MHz (6000 MT/s)

Sapphire Dual-X R9 280 OC

850/940 MHz GPU, 3 GB GDDR5 at 1250 MHz (5000 MT/s)

AMD Radeon R9 270X

1050 MHz GPU, 2 GB GDDR5 at 1400 MHz (5600 MT/s)

Nvidia GeForce GTX 660
980/1033 MHz GPU, 2 GB GDDR5 at 1502 MHz (5008 MT/s)

Nvidia GeForce GTX 760

980/1033 MHz GPU, 2 GB GDDR5 at 1502 MHz (5008 MT/s)

Nvidia GeForce GTX 770
1046/1085 MHz GPU, 2 GB GDDR5 at 1752 MHz (7008 MT/s)
SSD
Samsung 840 Pro, 256 GB SSD, SATA 6Gb/s
Power
XFX PRO850W, ATX12V, EPS12V
Software and Drivers
Operating System
Microsoft Windows 8 Pro x64
DirectX
DirectX 11
Graphics Drivers
Radeon R9 285: AMD Catalyst 14.8 beta
All other Radeon cards: AMD Catalyst14.7 RC 1
All GeForce Cards: Nvidia 340.52 WHQL
Benchmarks
Watch Dogs
Version 1.04.497, Custom THG Benchmark, 90-sec FRAPS, Driving
Arma 3
V. 1.26.126.789, 30-sec. Fraps "Infantry Showcase"
Battlefield 4
Version 1.3.2.3825, Custom THG Benchmark, 90-Sec
Assassin's Creed IV: Black Flag
Custom THG Benchmark, 40-Sec
ThiefVersion 1.6.0.0, Built-in Benchmark
TitanfallVersion 1.0.5.7, Demeter Map, Custom THG Benchmark
Grid Autosport
Version 1.0.101.4672, Built-In benchmark
Far Cry 3
Version 1.05, Custom THG Benchmark, 55-sec FRAPS

Loading...

This thread is closed for comments
64 comments
    Your comment
  • m32
    I wanted to see the GPU die and OCing results. :(
  • JeanLuc
    The idle power consumption numbers are odd, the previous generation cards use less then at idle didn't they? Not that 15 watts is going to break anyone's bank account but its strange nether the less.

    Good to see AMD have tackled the noise and temperature issues that have plagued it's previous 28nm cards as well but it's a bit late in the day given that 20nm shouldn't be to far off now.
  • chaospower
    TL;DR Pay more to get the same performance in a more power efficient form.
  • gear999
    Really nice article guys. I'm impressed by how the 285 actually was able to keep up with the 280. And I'm shocked by the fact that The $250 Nvidia card loses to a $170 AMD card. Thank god I bought a GTX 770 :P

    Also, on the last page, you guys wrote R7 270X instead of R9, and in the chart it says "Relative to Radeon HD 7950 Boost". Oh, and in the Pros section, it says the 285 has R9 260 like performance?

    [EDIT by Cleeve]
    Thanks for the proofread, fixing it now! :)
    [/edit]
  • Mike Stewart
    wow ! at 250$ it actually is a better card even than 280X !! and it was meant for 760....but as it shows here even a 270X is a WAY better card than 760....
  • tomfreak
    Had the tonga 285 come with a 6GHz/7Ghz GDDR5 & 4GB VRAM, the result will be a lot different. Whats with AMD putting on a 5500 memory? facepalm.jpg
  • srap
    While this is really a third GCN iteration, showing it as a version number of 3.0 (as in: "Tonga and GCN Update 3.0") makes no sense for me.
  • Amdlova
    some one write this with a .45 acp on the head. I see some error on numbers models etc...
    I prefer get a r9 280 and downclock get same results. I can't see the point of this heat on graphics. maybe drivers. OR THIS IS HAWAII XT! Too much Heat!
  • Amdlova
    Quote:
    I wanted to see the GPU die and OCing results. :(

    I think the guys see if they hit the OC the room Will burn! maybe a problem with drivers.
    Last time i see that Heat 290x tests. lol!
  • Gillerer
    On the first page, it says "Improvements are always welcome but with the memory interface cut in half compared to the Radeon R9 280,...".

    But in fact, the memory interface was cut by a third (384 bit -> 256 bit), not half.

    [Edit by Cleeve]
    Good point, fixed! Thx.
    [/edit]
  • mister g
    You guys might want to update the first chart of this review; the one comparing the specifications of the 280, 285, and the 280X. The 280X is a Tahiti chip not Tonga.

    [Edit by Cleeve]
    Good catch, fixed but might take a while to populate. :)
    [/Edit]
  • InvalidError
    502873 said:
    Had the tonga 285 come with a 6GHz/7Ghz GDDR5 & 4GB VRAM, the result will be a lot different.

    Faster memory would have helped but more would not have made much of a difference: most of the extra memory on GPUs with more memory channels gets filled with extra copies of resources to improve availability. Without those extra channels, filling more RAM with extra copies would make little difference.
  • west7
    I'd like to see this lossless color compression in 4k gaming cards
  • Someone Somewhere
    That's really dumb numbering...

    The R7 265 is faster than the R7 260X, yet the R9 285 is slower than the R9 280X?
  • MrstimX
    probably AMD's hand was forced due to gsynch, so they had to quickly phase out all non freesynch cards before dec..might expect a r9 285x by end oct
  • logainofhades
    930500 said:
    That's really dumb numbering... The R7 265 is faster than the R7 260X, yet the R9 285 is slower than the R9 280X?


    Yea this should have been named 275 or 275x.
  • Someone Somewhere
    No, because that would imply that it's slower than the 280.

    The 280X probably should have been the 285, and this card should have been released as the 280X. Or it could be next-gen; call it the 380 or 375.
  • logainofhades
    It kinda is slower than the 280. It trades blows with it, but still is not equal. I would say 275x would be fitting as the 280/280x are 384bit/3gb cards, where as the 270/270x are 256bit/2gb cards.
  • InvalidError
    59887 said:
    It kinda is slower than the 280. It trades blows with it, but still is not equal. I would say 275x would be fitting as the 280/280x are 384bit/3gb cards, where as the 270/270x are 256bit/2gb cards.

    The 270/280 are just rehashes of HD7xxx designs while the 285 is a cut-down 290... and the 285 does beat the 280 enough times to earn its place in the 28x range.

    Give the 285 a 6GT/s memory interface and it would slot in more solidly between the 280 and 280X.
  • Cryio
    Quote:
    That's really dumb numbering... The R7 265 is faster than the R7 260X, yet the R9 285 is slower than the R9 280X?


    Indeed, naming schemes are always kind of bogus.

    260< 260X < 265

    280<=285< 280X

    That's just the way it is.
  • Ranth
    Could you make a graph to show other cards power consumption? I find it hard to relate to without them.
  • MrstimX
    probably AMD's hand was forced due to gsynch, so they had to quickly phase out all non freesynch cards before dec..might expect a r9 285x by end oct
  • Traciatim
    With the differences in memory interface and bandwidth I would really have liked to see some resolution and AA scaling tests. It would be nice to see how the different memory speeds and bandwidth change the performance of each card.
  • vertexx
    Overclocking?