AMD Radeon R9 295X2 Performance Review

[SKYMTL]

4K: Far Cry 3 / Hitman Absolution

Please note that due to the R9 295X2’s inability to output 4K over HDMI at 60Hz, it is incompatible with our current DVI-based FCAT setup (which requires HDMI dongles be used). As such, we have switched to FRAPS for these tests.

Far Cry 3

3840 x 2160

Hitman Absolution

3840 x 2160

 

[SKYMTL]

4K: Metro Last Light / Thief

Please note that due to the R9 295X2’s inability to output 4K over HDMI at 60Hz, it is incompatible with our current DVI-based FCAT setup (which requires HDMI dongles be used). As such, we have switched to FRAPS for these tests.

Metro: Last Light

3840 x 2160

Thief

3840 x 2160

 

[SKYMTL]

4: Tomb Raider

Please note that due to the R9 295X2’s inability to output 4K over HDMI at 60Hz, it is incompatible with our current DVI-based FCAT setup (which requires HDMI dongles be used). As such, we have switched to FRAPS for these tests.

Tomb Raider

3840 x 2160

 

[SKYMTL]

Onscreen Frame Times w/FCAT

Onscreen Frame Times w/FCAT

When capturing output frames in real-time, there are a number of eccentricities which wouldn’t normally be picked up by FRAPS but are nonetheless important to take into account. For example, some graphics solutions can either partially display a frame or drop it altogether. While both situations may sound horrible, these so-called “runts” and dropped frames will be completely invisible to someone sitting in front of a monitor. However, since these are counted by its software as full frames, FRAPS tends to factor them into the equation nonetheless, potentially giving results that don’t reflect what’s actually being displayed.

With certain frame types being non-threatening to the overall gaming experience, we’re presented with a simple question: should the fine-grain details of these invisible runts and dropped frames be displayed outright or should we show a more realistic representation of what you’ll see on the screen? Since Hardware Canucks is striving to evaluate cards based upon and end-user experience rather than from a purely scientific standpoint, we decided on the latter of these two methods.

With this in mind, we’ve used the FCAT tools to add the timing of partially rendered frames to the latency of successive frames. Dropped frames meanwhile are ignored as their value is zero. This provides a more realistic snapshot of visible fluidity.

Note that we are ONLY doing 1440P testing as our FCAT setup is incompatible with the by AMD’s R9 295X2 outputs a 4K 60Hzsignal.

 

[SKYMTL]

Onscreen Frame Times w/FCAT (pg.2)

Onscreen Frame Times w/FCAT (pg.2)

 

[SKYMTL]

AMD's Mantle & R9 295X2; A Perfect Match?

AMD's Mantle & R9 295X2; A Perfect Match?

AMD’s Mantle has received quite a bit of page space here at Hardware Canucks since it was announced last year. There has been quite a lot of hype and speculation but thus far it has only been successfully rolled into two games: Battlefield 4 and Thief. It does however promise substantial performance increases in certain scenarios.

Unfortunately, since Mantle uses a customized API, FCAT monitoring tools and FRAPS aren’t compatible so we benchmarked Thief with its in-game tool. Battlefield 4 on the other hand now includes an integrated overlay so FCAT was used for measuring in that instance at 1440P. It should also be mentioned that AMD’s DisplayPort-only 4K output isn’t compatible with our FCAT setup so all 4K testing was done with FRAPS.

The first set of results shows that in high end systems, Mantle certainly shouldn’t be a standout choice for anyone that cares about framerates. While there wasn’t any performance loss, we didn’t receive a massive boost either

These lackluster results highlight why Mantle isn’t an obvious choice for certain system types. Mantle optimizes the communication between a system’s CPU and its GPU which means lower end CPUs can see massive benefits as the API overhead is substantially reduced. Our system sports a 12-thread Ivy Bridge-E operating at 4.7GHz so it can literally power through any software-level bottlenecks without missing a beat. In many ways our setup is likely a close analog to the ones being used by the R9 295X’s intended clientele so expect similar results with your high end system.

4K and Mantle really aren’t a good match since the CPU is further relegated to spectator status and the graphics card becomes a major system bottleneck. Again there really aren’t any difference here but Mantle is in its infancy and has plenty of room to grow.

 

[SKYMTL]

Temperatures & Thermal Imaging

Temperatures & Thermal Imaging

In this review we decided to go about our temperature analysis in a different way. Due to the Hawaii architecture’s well know habit of throttling when under load as temperatures reach extreme levels, in this section we will be testing its heat production alongside clock speeds.

Since frequencies are tied at the hip to TDP values (core temperatures, power consumption and VRM temperature), there’s a very real risk that like the R9 290X, the R9 295X2 won’t reach its full potential. That is unless its liquid cooling setup is up to the task. In a perfect world, the frequency shouldn’t hit levels that allow frequencies to throttle back.

With temperatures hovering around the 65°C mark, its obvious AMD’s liquid cooling strategy is paying off in spades. At reference settings, the R9 295X2 is actually designed to begin throttling at 75°C which is a level it won’t reach unless the GPUs are heavily overclocked. We can’t stress just how big a step this is considering the horrible experiences we had with the R9 290X’s performance loss when its fans were operating at anything but full tilt.

Low thermals equate perfectly stable performance with clock speeds hitting 1018MHz and remaining there throughout the 15 minute benchmark. Even after an hour of constant load, the frequencies hadn’t budget one iota.

Left: Idle / Right: Load​

As temperatures rise during high load scenarios it becomes quite evident that a ton of heat is being transmitted towards the radiator. As we can see above, the tubing itself gets quite warm, though the various areas on the R9 295X2 itself remain relatively cool. Forgive the insane setup since it was the only way to get the radiator and card into the same shot.

Even after an hour of constant gaming, we never saw visible PCB temperatures above the 65°C mark with an average temperature around 60°C. However, one slightly worrying aspect is the hot-spot around the power inputs. Those things are carrying far more current than they are specified to take and while we don’t have any worries about longevity, it would have been nice to see AMD spread the card’s power distribution over three 8-pin connectors rather than just two. This would have put substantially less load on individual inputs.

Naturally, the backplate and radiator receive the lion’s share of heat but they’re both designed to effectively disperse any extreme temperatures quite effectively. There are certainly no worries here and it looks like the R9 295X2 is well equipped to handle its hot-running cores.

 

[SKYMTL]

Acoustical Testing / Power Consumption

Acoustical Testing

What you see below are the baseline idle dB(A) results attained for a relatively quiet open-case system (specs are in the Methodology section) sans GPU along with the attained results for each individual card in idle and load scenarios. The meter we use has been calibrated and is placed at seated ear-level exactly 12” away from the GPU’s fan. For the load scenarios, Hitman Absolution is used in order to generate a constant load on the GPU(s) over the course of 15 minutes.

Welcome to a bout of shock and awe. We certainly weren’t expecting any AMD dual GPU card to ever display such a low noise profile. Even with a pair of fans (one for the radiator and one for the VRMs) the R9 295X2 happens to be one of the quietest reference cards we’ve ever tested. While some folks may roll their eyes at the need to water cool a graphics cards, the numbers certainly speak for themselves.

System Power Consumption

For this test we hooked up our power supply to a UPM power meter that will log the power consumption of the whole system twice every second. In order to stress the GPU as much as possible we used 15 minutes of Unigine Valley running on a loop while letting the card sit at a stable Windows desktop for 15 minutes to determine the peak idle power consumption.

Please note that after extensive testing, we have found that simply plugging in a power meter to a wall outlet or UPS will NOT give you accurate power consumption numbers due to slight changes in the input voltage. Thus we use a Tripp-Lite 1800W line conditioner between the 120V outlet and the power meter.

While the R9 295X2 may be one of the quietest and coolest running cards we’ve come across in a long time, it also happens to be the single most power hungry GPU ever created. It consumes more than SEVENTY WATTS more than a dual GTX 780 Ti setup and over 150W more than two GTX 780’s. Yikes. We’d recommend nothing less than a 900W PSU for this thing.

It is also noteworthy that the dual R9 290X solutionexhibited lower power consumption simply because it was throttling, even on Uber mode.

 

[SKYMTL]

Overclocking Results; Yes, There's More

Overclocking Results

Contrary to what some believe, when given adequate cooling, AMD’s Hawaii cores do have a good amount of overclocking headroom. The only problem with that equation is the cooling part which needs to be quite substantial to avoid buckling under the extreme thermal load produced by the ASIC. Since the R9 295X2 includes what can best be described as an awesome amount of cooling power, we should be able to push it even further.

There is one small hiccup in the grand scheme of overclocking this card: in order to remain under a predetermined TDP threshold, the cores have been limited to 75°C. Anything higher and they’ll begin to throttle. That’s a far cry from the R9 290X’s stratospheric 90°C operating temperature but we can understand AMD’s hesitation to unleash these cores. Luckily, the water cooling design keeps temperatures around 62°C under normal operating conditions so there’s still nearly 13°C of room before worrying about down-clocking.

Overclocking the R9 295X2 within AMD’s CCC follows the same basic yet convoluted formula seen on previous R9-series cards. The Power Target and Clock Speed must be adjusted together, with one compensating for the other. This maxes fine tuning a bit challenging but with a little trial and error, it becomes second nature.

Actually finding that “sweet spot” between power and clock speeds can be frustrating but we finally settled at 1131MHz for the core and 5.66Gbps on the GDDR5 memory. While neither represent an awe-inspiring increase, these levels were attained without running afoul of PowerTune’s limits so frequencies remained constant throughout testing. Any further and the cores easily hit the 75°C limiter.

Since our 4K testing wasn’t CPU limited in any way, we decided to use it for testing the overclock’s stability. The end result was a good boost in framerates, though one that likely won’t be noticeable to most gamers.

 

[SKYMTL]

Conclusion; Head and Shoulders Above the Rest

Conclusion

These dual GPU reviews typically go the same way for me: I harp on AMD’s solution for being too loud and lacking proper driver support or talk about NVIDIA’s alternative which benefits from a low noise profile and higher performance metrics due to better multi card game profiles. It’s an old story that was getting a bit boring after being repeated from one architectural generation to another. Obviously AMD had enough of hearing the same refrain over and over again since the R9 295X demolishes nearly every preconception I had. It’s that good.

To be completely honest, I had abysmally low expectations for any dual core card derived from AMD’s Hawaii cores. The R9 290X and R9 290 had issues with their extreme TDP values dragging down clock speeds as PowerTune throttled performance in a losing effort to curtail thermals. As evidenced by the HD 7990, cores with rampant temperature output and power consumption don’t make for good bedfellows on a single PCB. While nothing has changed about the Hawaii architecture, AMD’s choice to go with a water cooling setup is an excellent one since the R9 295X2 feels light years more refined than their previous outings.

Let’s take a second to talk about that Asetek-supplied water cooling setup since without it, I doubt AMD would have ever been able to achieve the R9 295X2’s spectacular performance. It uses a single bay 120mm radiator which is compatible with nearly every case on the market and flexible tubing so installation is a breeze. This setup limits the card to an operating temperature of just 65°C while under full load while remaining extremely quiet, effectively eliminating the issues associated with the other cards in the Radeon R9 series.

There just aren’t sufficient words to describe the R9 295X2’s performance. It is obscenely fast and will remain so after hours of gaming since the water loop has thermal overhead to spare. At 2560x1440, the resolution most gamers with disposable income will use, it can’t quite keep up with the GTX 780 Ti SLI but does soundly beat two GTX 780’s. Remember, those solutions are $1400 and $1000 respectively so the R9 295X2’s $1500 asking price doesn’t look like a huge stretch.

Its position against a dual R9 290X setup is a bit less clear since performance was very much similar. However, from a temperature and acoustic perspective, the ball is clearly in the R9 295X2's court.

While the reference R9 290X’s full performance aspects are partially masked by its penchant for throttling (even in Uber mode), AMD’s single to dual GPU scaling is nothing short of spectacular. There were significant performance increases in every game without any of the telltale stuttering that ruined past Crossfire experiences. The frametime results do highlight a few situations where stuttering was apparent but NVIDIA’s dual card solutions suffered the same fate. That points towards game engine problems (no surprise there!) rather than driver or hardware issues.

It goes without saying that Mantle could technically push things even further than what we’ve seen here. However, when playing in ultra high detail environments where the R9 295X2 thrives, Mantle doesn’t make any difference whatsoever when used alongside a fast multi core CPU.

4K and higher resolutions are where the R9 295X2 truly shines. Whereas a single R9 290X couldn’t even finish most of our tests (hence its exclusion from the results), the new Radeon flagship powered through every game without a hiccup. It even managed to completely close the gap with NVIDIA’s GTX 780 Ti SLI, which goes to show that more video memory really comes in handy at 4K.

Throughout the course of reviewing the R9 295X2, I had a constant struggle with how to address its installation foibles. Anyone with an All in One water cooling unit already attached to their CPU may be faced with a dilemma when trying to find a suitable mounting location for another 120mm radiator and its associated tubing. Luckily most modern cases have multiple 120mm locations so that worry is rendered inconsequential.

What really intrigues me about the R9 295X2 is its implications for small form factor mini ITX and micro ATX systems with limited expansion slot options. Its ability to exhaust heat outside the chassis’ confines makes it a perfect companion for space-constrained builds. Alongside a BitFenix Prodigy and ASUS’ Maximus VI Impact, this card would provide extreme levels of full-tower sized performance in a system than has a laptop-sized footprint. That’s hard to argue against. Mini ITX is a direction many enthusiasts are now taking in order to save on desk space and the R9 295X2 can add an awesome amount of punch to an SFF build, provided the case is compatible with 12” long GPUs.

It goes without saying that the R9 295X2 is a power hungry goliath but, unlike previous dual GPU generations, AMD has engineered their way around potential roadblocks. While the use of a water cooling loop may be cause for derision among purists, I’d tend to disagree with any dissenting voices. With the acceptance of Corsair’s Hydro series and other AIO units, closed loop liquid cooling setups no longer have a stigma of ridicule attached to them. In this case, water cooling provides excellent temperatures, low acoustics, room for overclocking, awe-inspiring performance, longevity and adaptability into cases that may not react all that well to cards that dump a ton of hot air into their confines. As they say: it just works, and works well.

I’ve prattled on about the R9 295X2’s benefits simply because finding faults with it is next to impossible. Granted, a dual GTX 780 system exhibits better performance per watt and more bang for your buck at 1440P but when moving to 4K, AMD’s latest card begins to pull away. It may not be the best solution for those with dual card capable systems (though two would make for the most powerful gaming rig available) but the R9 295X2 provides a truckload of drool-worthy potential for gamers looking to downsize their system or simply go home with the fastest graphics card around.