At long last, AMD has filled the gaping hole in its next-generation Radeon lineup. A new pair of graphics cards has come to slot in right between the high-end Radeon HD 7900 series and the decidedly less high-end 7700 series, completing the Southern Islands trifecta—and offering gamers some fresh meat at $249.
If you were expecting exact replacements for the 6800 series, prepare to be disappointed. The new Radeon HD 7850 is the one priced at $249; its big brother, the Radeon HD 7870 GHz Edition, costs a more burdensome $349. That means today’s launch leaves the $199 price point conspicuously devoid of next-generation GPUs. Perhaps that will change in the future—possibly after the arrival of Kepler parts from Nvidia, which are rumored to be coming soon—but for now, we might say the hole in AMD’s lineup is only partially plugged.
Also, it turns out AMD has revived that beloved, time-honored tradition known as the soft launch. In the company’s words, today marks the lifting of the “preview NDA” (or preview non-disclosure agreement). We can tell you everything we know about these cards and post our performance findings, but actual products aren’t due out in volume until two weeks from now, on March 19.
You can look, in other words, but you can’t touch.
Nevertheless, the Radeon HD 7850 and Radeon HD 7870 GHz Edition have the potential to be AMD’s most compelling next-gen offerings yet. No, really. It’s true the flagship Radeon HD 7900 series offerings are the fastest single-GPU cards on the planet right now, but they’re also awfully expensive, with prices starting at $449 and ranging all the way up to $600. The Radeon HD 7700 series is less expensive, but overpriced considering the level of performance the cards deliver—so much so that we recommended previous-generation cards in our latest system guide, instead.
With these new arrivals, AMD may finally have next-gen cards that are both affordable and competitive. If the Radeon HD 7850 in particular can prove its mettle against enthusiast classics like the GeForce GTX 560 Ti and Radeon HD 6950, it could be an instant hit.
We were speaking quite literally when we said the Radeon HD 7800 series completes the Southern Islands trifecta. These puppies are driven by a new chip called Pitcairn, the third addition to the 28-nm Southern Islands GPU family. With a die size of 212 mm², Pitcairn is smaller than Tahiti (365 mm²) but a fair bit larger than Cape Verde (123 mm²).
Just like those other two chips, though, Pitcairn is fabbed on a 28-nm process and based on AMD’s Graphics Core Next architecture. It also has the same supplemental goodies, like PCI Express 3.0 support, ZeroCore Power, and a hardware video encoding block called VCE.
Scott covered those features, as well as Graphics Core Next, in some depth in his review of the Radeon HD 7970. If you haven’t read it already, I’d recommend doing that now. It’s okay; I’ll wait.
Here’s an abstracted overview of Pitcairn’s various components:
Being smaller than Tahiti, Pitcairn isn’t quite as well-furnished. AMD has cut the number of compute units from 32 to 20, leaving Pitcairn with 1280 stream processors and 80 texture units. (For the record, each compute unit has four texture units and four 16-wide vector units, also known as ALUs or stream processors.)
Two of the 64-bit memory controllers have been lopped off, as well, bringing the total down from six to four. Pitcairn’s path to memory is therefore 256 bits wide. However, AMD has endowed Pitcairn with the exact same number of ROP partitions and geometry engines as its larger sibling. Both chips can churn out 32 pixels and rasterize two triangles with each clock cycle.
Here’s how they compare, at a glance:
|Cape Verde||16||40/20||640||1||128||1500||123||28 nm|
Pitcairn is about 17% smaller than Barts, the chip that powers the Radeon HD 6800 series, and 46% smaller than Cayman, the core of the the 6900 series, but it has more transistors than either one. Its per-clock texture filtering and shader resources are somewhere in between, but as we’ll see on the next page, its higher clock speeds give it an advantage. Also, keep in mind that improved shader efficiency is one of the hallmarks of AMD’s Graphics Core Next architecture.
Comparing Pitcairn to Nvidia’s GF114 and GF110 chips (which drive the GeForce GTX 560 Ti and GeForce GTX 570, respectively) is a little trickier, since the GeForces are based on a completely different architecture. Still, Pitcairn looks well-equipped to face them—especially the GF114.
Before we get into our benchmarks, let’s take a look at the two cards Pitcairn powers: the Radeon HD 7870 GHz Edition and Radeon HD 7850. (Or, you know, you could just skip ahead to the performance-per-dollar scatter plots on the last page, if you’re more comfortable with an incomplete recapitulation of our hard work. Totally up to you.)
The faster of the two newcomers, the $349 Radeon HD 7870 GHz Edition, uses the Pitcairn GPU with all of its bits and pieces enabled. AMD clocks it at 1,000MHz, hence the GHz Edition suffix, and accompanies it with 2GB of GDDR5 memory set to operate at 1200MHz (for an effective peak transfer rate of 4800 MT/s). The card has a 190W power envelope, but AMD says “typical” power consumption is 175W.
The card is about 9.7″ (247 mm) long and requires a couple of six-pin PCIe power connectors. Our sample came with a copper-and-aluminum heatsink covered by one of AMD’s trademark Batmobile-inspired shrouds. A blower draws air from inside the case, through the heatsink fins, and outside the case via a vent in the port shield.
Speaking of connectors, the 7870 has the exact same port arrangement as the Radeon HD 7900 series: two mini DisplayPort outputs, one HDMI output, and one dual-link DVI port. AMD says the card supports up to six displays, but that’s only possible with a DisplayPort hub—and those are hard to come by, if not unavailable entirely.
The $249 Radeon HD 7850 is also Pitcairn-based, but it’s had a mild lobotomy to keep it from nipping at its sibling’s heels too much. Four of its 20 compute units have been disabled, leaving it with 1024 stream processors and 64 texture units. The core clock speed has been lowered to 860MHz, as well. Happily, other resources have gone unharmed. The 7850 also enjoys the exact same memory configuration as the 7870.
You may have noticed that our 7850 and 7870 samples look an awful lot alike, save for the “1GHz Edition” sticker on the latter. That’s because the two cards have the same circuit board and the same cooler. However, AMD tells us retail Radeon HD 7850s will be different. They’ll have shorter, stubbier circuit boards and matching third-party heatsinks and fans. Here’s one, built by Sapphire:
Another key difference is the 7850 only requires a single six-pin power connector. That’s true even for our sample:
According to AMD, the 7850 has a 150W TDP and typical power draw of around 130W. That’s 40-45W less than the 7870.
Ready to move on to the benchmarks? Not so fast. We have a big, meaty table full of peak theoretical numbers for you to pore over first:
|GeForce GTX 560 Ti||26||53||53||1.3||1644||128|
|GeForce GTX 560 Ti 448||29||41||41||1.3||2928||152|
|GeForce GTX 570||29||44||44||1.4||2928||152|
|GeForce GTX 580||37||49||49||1.6||3088||192|
|Radeon HD 6870||29||50||25||2.0||900||134|
|Radeon HD 6950||26||70||35||2.3||1600||160|
|Radeon HD 6970||28||84||42||2.7||1760||176|
|Radeon HD 7850||28||55||28||1.8||1720||154|
|Radeon HD 7870 GHz Edition||32||80||40||2.6||2000||154|
|Radeon HD 7950||26||90||45||2.9||1600||240|
|Radeon HD 7970||30||118||59||3.8||1850||264|
Thanks to its 1000MHz clock speed, the Radeon HD 7870 has higher pixel fill and rasterization rates than even the Radeon HD 7970. That’s because the two cards have the same number of ROPs, and they can both process two triangles per clock, but the 7970 is only clocked at 925MHz. The 7870 does have more modest texture filtering capabilities, though, not to mention substantially lower shader throughput and memory bandwidth.
Compared to the old Radeon HD 6970, which carried the same $349 price tag before it started mysteriously disappearing from Newegg’s stock, the 7870 looks rather good. Some of its theoretical peaks are slightly higher, and some are slightly lower, but keep in mind Pitcairn should be more efficient than the 6970’s Cayman chip.
Getting a feel for the contest between the 7870 and the GeForce GTX 570 is a little harder, but as we said on the previous page, the new Radeons are not ill-equipped. The same can be said about the matchup between the 7850 and the GTX 560 Ti—in that matchup, the 7850 is better outfitted in all but peak FP16 texture filtering. The GeForce’s ability to do FP16 filtering at the same speed as it handles integer formats speed gives the GTX 560 Ti a sizeable advantage there over the Radeon HD 7850.
Will our benchmark results confirm our expectations? Let’s find out.
Our testing methods
Before we go on, we should note a couple of things about our test setup.
First, sharp-eyed readers (and Nvidia fanboys) may notice that our GeForce GTX 560 Ti is an Asus model clocked at 830MHz. While that card is priced at $244.99, just $5 south of the 7850, GTX 560 Ti variants clocked as high as 900MHz can be had for $249.99. In short, one could accuse us of under-representing the 7850’s chief rival somewhat.
Rest assured that wasn’t our intention. AMD quoted a price range of $200-299 for the Radeon HD 7800 series when it briefed us last Tuesday, so when we got started on this review, we expected the 7850 might cost as little as $200. By the time AMD finally divulged its final pricing on Thursday afternoon, we’d already benchmarked the GTX 560 Ti and had no time to test another model. Just keep in mind that quicker GTX 560 Ti variants do exist as you’re reading the benchmarks over the next few pages.
Also, our test system includes a Core i5-750—a 45-nm quad-core processor that’s starting to grow a little long in the tooth. Someone in the market for a $349 card might be reasonably expected to own a faster CPU, perhaps of the Sandy Bridge variety. Once again, we would have loved to test with a faster product, but time constraints prevented us from doing so. This setup isn’t a deal-breaker, though. None of the games and applications we tested are terribly CPU-bound, and as you’ll see on the next few pages, faster GPUs had no trouble demarcating themselves from slower offerings. There’s no indication that the Core i5-750 acted as a significant bottleneck.
As ever, we did our best to deliver clean benchmark numbers. Tests were run at least three times, and we reported the median results. Our test systems were configured like so:
|Processor||Intel Core i5-750|
|North bridge||Intel P55 Express|
|Memory size||4GB (2 DIMMs)|
|Memory type||Kingston HyperX KHX2133C9AD3X2K2/4GX
DDR3 SDRAM at 1333MHz
|Memory timings||9-9-9-24 1T|
|Chipset drivers||INF update 220.127.116.115
Rapid Storage Technology 10.1.0.1008
|Audio||Integrated Via VT1828S
with 18.104.22.16800 drivers
|Hard drive||Western Digital Caviar Black 1TB
Samsung Spinpoint F1 HD103UJ 1TB SATA
|Power supply||Corsair HX750W 750W|
|OS||Windows 7 Ultimate x64 Edition
Service Pack 1
|Driver revision||GPU core
|Asus GeForce GTX 560 Ti DirectCU II||GeForce 295.73||830||1000||1024|
|Asus GeForce GTX 570 DirectCU II||GeForce 295.73||742||950||1280|
|Asus Radeon HD 6870 DirectCU||Catalyst 8.95.5-120224a||915||1050||1024|
|XFX Radeon HD 6950||Catalyst 8.95.5-120224a||830||1300||1024|
|Asus Radeon HD 6970 DirectCU II||Catalyst 8.95.5-120224a||890||1375||2048|
|Radeon HD 7850||Catalyst 8.95.5-120224a||860||1200||2048|
|Radeon HD 7870 GHz Edition||Catalyst 8.95.5-120224a||1000||1200||2048|
Thanks to Asus, Corsair, Kingston, Intel, Samsung, and Western Digital for helping to outfit our test rigs with some of the finest hardware available. AMD, Nvidia, and the makers of the various products supplied the graphics cards for testing, as well.
Unless otherwise specified, image quality settings for the graphics cards were left at the control panel defaults. Vertical refresh sync (vsync) was disabled for all tests.
We used the following test applications:
- Batman: Arkham City
- Battlefield 3
- Crysis 2
- The Elder Scrolls V: Skyrim
- LuxMark 1.0 Win64
- ShaderToyMark 0.2.0
- TessMark 0.3.0
- Fraps 3.4.7
- GPU-Z 0.5.8
Some further notes on our methods:
- We used the Fraps utility to record frame rates while playing a 90-second sequence from the game. Although capturing frame rates while playing isn’t precisely repeatable, we tried to make each run as similar as possible to all of the others. We tested each Fraps sequence five times per video card in order to counteract any variability. We’ve included frame-by-frame results from Fraps for each game, and in those plots, you’re seeing the results from a single, representative pass through the test sequence.
We measured total system power consumption at the wall socket using a P3 Kill A Watt digital power meter. The monitor was plugged into a separate outlet, so its power draw was not part of our measurement. The cards were plugged into a motherboard on an open test bench.
The idle measurements were taken at the Windows desktop with the Aero theme enabled. The cards were tested under load running Skyrim at its Ultra quality preset with FXAA enabled.
We measured noise levels on our test system, sitting on an open test bench, using a TES-52 digital sound level meter. The meter was held approximately 8″ from the test system at a height even with the top of the video card.
You can think of these noise level measurements much like our system power consumption tests, because the entire systems’ noise levels were measured. Of course, noise levels will vary greatly in the real world along with the acoustic properties of the PC enclosure used, whether the enclosure provides adequate cooling to avoid a card’s highest fan speeds, placement of the enclosure in the room, and a whole range of other variables. These results should give a reasonably good picture of comparative fan noise, though.
- We used GPU-Z to log GPU temperatures during our load testing.
The tests and methods we employ are generally publicly available and reproducible. If you have questions about our methods, hit our forums to talk with us about them.
|GeForce GTX 560 Ti (Asus)||53||53||128|
|GeForce GTX 570 (Asus)||45||45||152|
|Radeon HD 6870 (Asus)||51||26||134|
|Radeon HD 6950 (XFX)||73||37||166|
|Radeon HD 6970 (Asus)||85||43||176|
|Radeon HD 7850||55||28||154|
|Radeon HD 7870 GHz Edition||80||40||154|
The results of this synthetic test track fairly closely with the theoretical peak numbers we calculated. We’d have expected the 7850 to perform a little better, though, all things considered.
|GeForce GTX 560 Ti (Asus)||1660||128|
|GeForce GTX 570 (Asus)||2968||152|
|Radeon HD 6870 (Asus)||915||134|
|Radeon HD 6950 (XFX)||1660||166|
|Radeon HD 6970 (Asus)||1780||176|
|Radeon HD 7850||1720||154|
|Radeon HD 7870 GHz Edition||2000||154|
For what it’s worth, the Radeon HD 7970’s poor showing in TessMark back in December was due to a software glitch, as we subsequently pointed out in our 7950 review. The new Radeons fly here with AMD’s latest drivers, handily outpacing not just their predecessors, but also their direct rivals from the Nvidia camp.
In fact, the 7870 outdoes the GeForce GTX 570 despite the GeForce’s considerably higher peak theoretical rasterization rate. Why is that? We’d wager it has something to do with the underlying architecture behind those two cards. Pitcairn, just like Tahiti, has two geometry engines sitting entirely separate from the shader cluster. Nvidia’s Fermi architecture, meanwhile, has a geometry engine in each of its shader multiprocessors (SMs). There are 16 SMs in the GF110 and 15 in the GTX 570. Our guess is that Nvidia’s distributed approach incurs more synchronization overhead than AMD’s architecture, which lumps fewer, more powerful geometry engines together on the GPU.
|GeForce GTX 560 Ti (Asus)||1.3||128|
|GeForce GTX 570 (Asus)||1.4||152|
|Radeon HD 6870 (Asus)||2.0||134|
|Radeon HD 6950 (XFX)||2.3||166|
|Radeon HD 6970 (Asus)||2.7||176|
|Radeon HD 7850||1.8||154|
|Radeon HD 7870 GHz Edition||2.6||154|
And here, folks, is where the improved efficiency of AMD’s Graphics Core Next architecture can be witnessed. Despite its slightly lower theoretical peak, the Radeon HD 7870 zooms well ahead of the less efficient Radeon HD 6970. The new Radeons make short work of the competing GeForces, too.
Ditto for this general-purpose computing test, where the GPUs put their shaders to work running a ray-tracing renderer written in OpenCL. Graphics Core Next is supposed to reap sizable benefits in general-purpose applications, and the evidence shows that it does.
Batman: Arkham City
To warm up this latest batch of Radeons, we grappled and glided our way around Gotham, occasionally touching down to mingle with the inhabitants.
We tested at a 1080p resolution with DirectX 11 effects enabled, normal tessellation, high FXAA antialiasing, and everything else cranked up. Those settings seemed to yield the best compromise of image quality and smoothness on the 7870.
We should preface the results with a little primer on our testing methodology. Along with measuring average frames per second, we delve inside the second to look at frame rendering times. Studying the time taken to render each frame gives us a better sense of playability, because it highlights issues like skipping, stuttering, and microstuttering that can all occur—and be felt by the player—within the span of one second. Charting frame times shows these issues clear as day, while charting average frames per second obscures them.
For example, imagine one hypothetical second of gameplay. Almost all frames in that second are rendered in 16.7 ms, but the game briefly hangs, taking a disproportionate 100 ms to produce one frame and then catching up by cranking out the next frame in 5 ms—not an uncommon scenario. You’re going to feel the game hitch, but the FPS counter will only report a dip from 60 to 56 FPS, which would suggest a negligible, imperceptible change. Looking inside the second helps us detect such skips, as well as other issues that conventional frame rate data measured in FPS tends to obscure.
We’re going to start by charting frame times over the totality of a representative run for each card. That should give us an at-a-glance impression of overall playability, warts and all. (Note that, since we’re looking at frame latencies, plots sitting lower on the Y axis indicate quicker cards.)
For some reason, Batman: Arkham City exhibits high frame latency spikes rather frequently. We recorded similar frame time spikes on faster cards tested with a faster CPU and quicker storage, as well as on slower cards running the game with its DirectX 11 mode disabled. We can conclude with reasonable certainty that the skipping is inherent to this game.
A cursory overview of the graphs above tells us the Radeon HD 7870 and 7850 are close, and perhaps a little faster overall, than their competitors and predecessors. It would also appear the GeForce GTX 560 Ti exhibits more latency spikes than its rivals from the AMD camp.
We can slice and dice our raw frame-time data in three ways, as you’re about to see. By the way, we should caution that none of the graphs below can be construed as self-contained scoreboards; instead, they each show a different facet of the cards’ performance, and they should be viewed as a whole along with the raw frame-time plots above.
Our first graph shows average frames per second. Though this metric doesn’t account for irregularities in frame latencies, it does give us some sense of typical performance.
In our second graph, we’re demarcating the threshold below which 99% of frames are rendered. The lower the threshold, the more fluid the game. This metric offers a sense of overall frame latency, but it filters out fringe cases.
Our last graph tells us how long each card worked on frames that took longer than 50 ms to render. Ideally, the result should be “0” for every card. That’s because the illusion of smooth motion is likely to begin breaking down once frame latencies rise into that territory. (For reference, 50 ms frame times would work out to a 20 FPS average if they were sustained for a whole second.) In a nutshell, this metric tells us how badly each card skips during gameplay, if it does at all.
Now, let’s put it all together.
Clearly, the Radeon HD 7870 GHz Edition is the fastest card of the pack. It has the highest average frame rates, the lowest 99th-percentile threshold, and the least dire frame latency spikes. The Radeon HD 7850 is also positioned rather favorably compared to our GeForce GTX 560 Ti, though it’s no better than its most direct predecessor, the Radeon HD 6950, when we look at frame latency thresholds and worst-case scenarios. The difference between the 7850 and the 6950 is difficult to notice in this test.
We tested Battlefield 3 by playing through the start of the Kaffarov mission, right after the player lands. Our 90-second runs involved walking through the woods and getting into a firefight with a group of hostiles, who fired and lobbed grenades at us.
The game was run at its highest detail preset, Ultra, which couples MSAA and FXAA antialiasing as well as snazzy DX11 effects and tessellation.
Just like in Arkham City, wanton frame latency spikes are a fact of life in this game—except only on Nvidia cards. It’s strange, but we noticed a similar affliction in other levels and with other Nvidia GPUs and prior graphics driver revisions. The effect is noticeable in-game as an uneven, skipping motion when the camera should be panning smoothly.
With that quirk accounted for, it seems the new Radeons are, once again, outpacing their rivals and forebears.
Yep. In BF3, even the 7850 beats both the GTX 560 Ti and the 6950 handily. The 7800-series cards also take less than 50 ms to render all of their frames, which is what you want.
I’ve made no secret of my appreciation for Bulletstorm‘s cathartic gameplay and gorgeous environments, so it seems like a fitting addition to our test suite. Here, we played through the first 90 seconds of the “Hideout” echo.
The game’s graphical settings were maxed out. We cranked detail settings all the way up and enabled 8X multisampled AA.
Strange. While the Radeons seem to achieve lower overall frame times than their Nvidia counterparts, they all experience a huge latency spike (of around 200 ms) about half-way through the run.
The average FPS chart shows no indication of those latency spikes, of course; it merely shows the Radeons trouncing the GeForces. The same goes for the 99th-percentile rankings, which, by design, don’t account for rare or unique latency spikes. Our index of time spend beyond 50 ms, however, reveals the true extent of the problem.
In a scenario like this one, frame times below 16.7 ms, equivalent to frame rates above 60 FPS, are usually of no great value, because they render more quickly than the monitor can display them. (Typical LCD monitors are limited to a 60 Hz refresh rate.) So, the Radeons’ higher average frame rates are somewhat moot if they come hand-in-hand with higher latency spikes that make the game hitch or pause. But then again, as we can see in our 99th-percentile graph, the GeForces’ latency thresholds are a wee bit higher.
I suppose it all comes down to what you prefer: high fluidity marred by rare skips, or somewhat lower fluidity that’s better maintained throughout the game. The Radeons will deliver the former; the GeForces the latter.
We tested Crytek’s latest shooter by running and gunned through the game’s rendition of Battery Park, sticking to the same path through the level to avoid drastic differences between samples.
The game was set to run at a 1920×1080 resolution with the “Extreme” detail preset. Both the DX11 “ultra upgrade” and high-res texture pack were installed, and both were enabled.
All of the cards exhibit a fair amount of inconsistency in frame times, with latencies rising and falling in quick succession throughout each run. The GeForce GTX 560 Ti seems to maintain somewhat greater consistency than the Radeon HD 7850 and 6950, but the Radeons’ plots lie a little lower on the Y axis. Between the ~$350 cards, it looks like a toss-up.
Right. So, the new Radeons come out ahead of their predecessors and competitors overall, but the Radeon HD 7850 spends slightly more time above 50 ms than the GTX 560 Ti. All things considered, though, 12 ms in a 90-second run is small potatoes. We’ll give the win to the 7800 series here.
The Elder Scrolls V: Skyrim
Our Skyrim run involved running around the town of Whiterun, starting from the city gates, all the way up to Dragonsreach, and then back down again.
Like BF3 and Bulletstorm, Skyrim‘s detail settings were maxed out. We selected the Ultra preset, which includes 8X MSAA, and then we enabled FXAA in the advanced options for good measure.
Things look pretty close here overall, but the GTX 560 Ti does trail the 7850 and 6950 by a small margin.
Hmm. While the 7800-series cards pull off the highest average frame rates yet again, they also have the highest 99th-percentile thresholds. If you look at our plot, you might see why: the new Radeons venture up a little higher than they should between frames 4000 and 4500. That sequence corresponds, I believe, to the point in the run where we overlook the town from the castle—a slightly heavier workload than the rest. To be fair, though, the 99th-percentile results are all pretty close.
Oh, and don’t let that long, red bar in the over-50-ms graph worry you. The 7870 suffered an unfortunate but small hiccup, which can be seen in our frame-time plot. 5 ms spent above 50 ms out of a 90-second run is pretty much insignificant.
Thanks to AMD’s ZeroCore Power scheme, 7000-series Radeon GPUs switch to an ultra-low-power state when the display is off. The company says power draw for the whole card drops under 3W when that happens. Our findings show idle power consumption does drop dramatically when the display goes to sleep. (Keep in mind that our power consumption measurements apply to the entire system except for the monitor, not just the graphics card.)
No surprises here. The Radeon HD 7870 and 7850 are the most power-efficient of the bunch, but why wouldn’t they be? They both have small and nimble 28-nm GPUs, while our other cards must make do with larger, more power-hungry 40-nm chips.
Noise levels and GPU temperatures
None of these cards are terribly loud except for the Radeon HD 7870, whose blower-style cooler spins itself into a frenzy under load. That cooler design does have the benefit of exhausting hot air directly outside the case, which can be particularly helpful in multi-GPU configs. Still, there’s something to be said for the large, conventional fans AMD’s partners typically put on their cards.
Yeah, so, no need to worry about your shiny new 7800-series Radeon overheating. That kind of thermal headroom should be helpful to overclockers, as well. Speaking of which…
When we were briefed about the 7800 series last week, one AMD representative told us he pushed the Radeon HD 7870 from its stock 1000MHz clock speed to 1200MHz without so much as a voltage increase. We were curious to put that claim to the test, so we did.
We used MSI’s excellent Afterburner tool to control clock speeds and GPU voltage, AMD’s Catalyst Control Center to max out the PowerTune setting, and MSI’s Kombustor application to test stability and get a quick sense of performance. Rather than relate my findings in prose, I’m going to paste in my notes. They should be fairly self-explanatory.
By the way, we found that the card’s memory silently reverted to its default speed (1200MHz) when overclocked too high. The only way to tell was to run Kombustor’s built-in benchmark at each setting and see where performance peaked.
1100MHz — OK after 5 min burn-in
1200MHz — OK after 5 min burn-in
1300MHz — crash
1275MHz — crash
1250MHz — crash
1225MHz — crash
1200MHz — 6414 kombustor — stock
1300MHz — 6894 kombustor — OK after 5-min burn-in
1400MHz — 6423 kombustor — (resets to stock)
1350MHz — 7186 kombustor — OK after 5-min burn-in
1375MHz – 7312 kombustor — OK after 5-min burn-in
1300/1375MHz — 1.250v — crash
1300/1375MHz — 1.275v — crash
1300/1375MHz — 1.300v — crash
1275/1375MHz — 1.300v — OK after 5-min burn-in
Without overvolting, we succeeded in running the Radeon HD 7870 at 1200MHz with its memory chugging along at 1375MHz. Score one for AMD. We didn’t stop there, of course. Once we raised the GPU voltage from 1218mV to 1300mV, we were able to squeeze an extra 75MHz out of the GPU. That left us with core and memory speeds 28% and 15% above stock, respectively. How did that translate in terms of gaming performance?
Yow! We’re looking at a 20% increase in average frame rates, which is pretty spectacular. Did overclocking raise power consumption through the roof?
Apparently not. We recorded a 17W increase under load, which still puts the 7870 below even the old Radeon HD 6870.
Something tells me AMD’s partners are going to be rushing to offer Radeon HD 7870 variants with higher-than-stock clock speeds. If vanilla retail cards are as overclockable as our sample, though, paying a premium for a superclocked card may not be necessary.
Let’s round things out with a couple of our famous scatter plots. We’re laying average performance (based on the results from the games we tested) along the Y axis and prices along the X axis. The sweet spot will be the card closest to the top left of the plot, while the worst will be closer to the bottom right. We fetched prices for the new Radeons from AMD, and other prices were gleaned from Newegg.
By the way, we’ve excluded Bulletstorm from our averages. The game skews things heavily in favor of the AMD cards, and considering the latency spike issue we encountered, we didn’t think that was fair. The numbers below account for average performance across our other test cases: Arkham City, Battlefield 3, Crysis 2, and Skyrim.
We can also compile a value scatter plot out of our 99th percentile frame time data. For consistency’s sake, we’ve converted the frame times to frame rates, so desirable offerings are still at the top left.
What can we say? The Radeon HD 7870 GHz Edition and Radeon HD 7850 are plainly more desirable than the old Radeon HD 6970 and 6950. They’re not only faster for the money, as the plots above show. They also have substantially lower power consumption and, if our experience with the 7870 is any indication, obscene amounts of overclocking headroom.
I’d say the two new Radeons are also better options than the comparable GeForces. It’s true the GeForce GTX 560 Ti we tested isn’t one of the highest-clocked models, but the Radeon HD 7850 is so much faster, I doubt a clock speed increase for the GeForce would bridge the gap. The Nvidia parts also had more trouble maintaining consistently low frame times in the games we tested, a fact that’s reflected in our 99th percentile FPS per dollar plot.
And, again, the Radeons are way more power-efficient.
Those are all remarkable achievements, but they’re diminished by AMD’s somewhat conservative pricing. The key thing to note is that Pitcairn is considerably smaller than the GPUs inside the Radeon HD 6900 and GeForce GTX 560 series. In fact, it’s even smaller than Barts, a chip that powers Radeon HD 6800-series cards priced as low as $140. It seems like a given that the Radeon HD 7850 will find its way south of the $200 mark eventually, and that is a truly exciting prospect. Heck, we may even be treated to a price war once Nvidia’s 28-nm Kepler GPUs come out. If that happens, AMD clearly has plenty of ammunition.
As excellent as these new Radeons are, I’m a little bit disappointed by AMD’s choice of cooler for the Radeon HD 7870 GHz Edition. (We can’t really comment on the 7850, since our sample wasn’t representative of retail offerings.) Considering the Pitcairn GPU’s modest power requirements, I’d have liked AMD to tune its reference cooler for lower noise levels. That said, since few retail cards use reference coolers these days, that point may be moot.