Nvidia’s GeForce GTX 550 Ti is finally settling onto store shelves, having made peace with the fact that it’s slower and considerably less exciting than some of its pricier elders. Already, the latest GeForce faces another challenge—or rather, a challenger that hails from the Radeon camp and wants to throw down with the Ti on its home turf.
AMD’s Radeon HD 6790 is being introduced today, and it bears the exact same $149 price tag as the new kid from Nvidia. According to AMD, the 6790 should “comfortably” outpace its GeForce counterpart by as much as 30%, letting you enjoy games at a 1080p resolution. We found the GTX 550 Ti wasn’t always up to the task of cranking out smooth frame rates at 1680×1050 with antialiasing enabled, so that sounds like an attractive premise—if the Radeon delivers, of course.
Beside being a competitor to the GTX 550 Ti, the 6790 also provides a middle ground between AMD’s Radeon HD 5770, which sells for as little as $120 at Newegg these days, and the quicker Radeon HD 6850, which starts closer to the $165 mark. Before today’s release, Radeon fans had to deal with the truly unbearable dilemma of having to choose between one of those two cards. No longer! That’s called progress, folks.
Now, how did AMD conjure up an answer to Nvidia’s new $149 GPU so quickly? Well, it kind of didn’t. The Radeon HD 6790 is based on the Barts graphics processor already known for its starring roles in the Radeon HD 6850 and 6870. AMD has simply disabled a few bits and pieces to keep the 6790 from nipping at the heels of real 6800-series offerings. That said, you’d never know it from looking at the Radeon HD 6790 engineering sample AMD sent us for review. The card has the same imposing length, cooler, display output arrangement, dual six-pin power connectors as the Radeon HD 6870:
To keep things from getting confusing here in our labs, we thought it appropriate to slap a sticker on the 6790. Ahh, much better.
Of course, retail Radeon HD 6790 variants won’t have quite the same garage-sale look. AMD tells us boards designs “will vary greatly from what you’re seeing on . . . sample boards, including PCB, power connectors, cooler design etc.” Certain retail 6790 cards won’t require dual PCIe power connectors, which is probably a good thing—folks shopping for a $149 graphics card are probably lucky if their power supply has one PCIe power plug.
Here’s Sapphire’s take on the Radeon HD 6790, just for the sake of illustration:
Note the shorter circuit board and the snazzier-looking cooler. If the side were visible, you’d see a pair of DVI ports, one HDMI port, and a DisplayPort output. Sapphire’s card will apparently have dual power connectors, but AMD tells us a PowerColor offering with only one PCIe plug will hit stores some time after today.
Clearly, the new Radeon has a lot in common with the 6800 series. Why didn’t AMD simply call it the Radeon HD 6830? That kind of nomenclature wouldn’t exactly upset tradition, after all, and it’d undoubtedly be more fitting from an architectural point of view.
The answer is simple: despite featuring a larger and more capable GPU, the Radeon HD 6790’s specifications are strikingly similar to those of the Radeon HD 5770, which AMD now sells in pre-built PCs as the Radeon HD 6770.
|Juniper (Radeon HD 5770)||16||40||800||128||1040||166||40 nm|
|Barts (Radeon HD 6790)||16||40||800||256||1700||255||40 nm|
|Barts (Radeon HD 6850)||32||48||960||256||1700||255||40 nm|
|Barts (Radeon HD 6870)||32||56||1120||256||1700||255||40 nm|
From a bird’s eye view, the 6790’s only notable holdover from the 6800 series is the 256-bit memory interface, which the 5770’s Juniper GPU is physically incapable of matching. AMD couldn’t put Juniper on stilts, so the quick-and-easy alternative was to pay Barts a visit and shatter its tibias with a baseball bat. The fractures incapacitated four of Barts’ SIMD arrays, leaving it with 800 ALUs and the ability to filter only 40 textures per clock. This latest example of GPU hobbling also destroyed half of Barts’ rasterization capabilities, limiting it to pushing only 16 pixels/clock. As a result, the 6790 and the 5770 have identical ALU counts, and they can filter the same number of pixels and textures per clock. Make sense?
|GeForce GTS 450||12.5||25.1||25.1||601||783||57.7|
|GeForce GTS 450 AMP!||14.0||28.0||28.0||672||875||64.0|
|GeForce GTX 550 Ti||21.6||28.8||28.8||691||900||98.5|
|GeForce GTX 550 Ti Cyclone||22.8||30.4||30.4||730||950||103|
|GeForce GTX 550 Ti AMP!||24.0||32.0||32.0||768||1000||106|
|GeForce GTX 460 768MB||16.2||37.8||37.8||907||1350||86.4|
|GeForce GTX 460 1GB||21.6||37.8||37.8||907||1350||115|
|GeForce GTX 560 Ti||26.3||52.6||52.6||1263||1644||128|
|Radeon HD 5770||13.6||34.0||17.0||1360||850||76.8|
|Radeon HD 5770 SOC||14.4||36.0||18.0||1440||900||76.8|
|Radeon HD 6790||13.4||33.6||16.8||1344||840||134.4|
|Radeon HD 6850||24.8||37.2||18.6||1488||775||128|
|Radeon HD 6870||28.8||50.4||25.2||2016||900||134|
|Radeon HD 6950||25.6||70.4||35.2||2253||1600||160|
Studying maximum theoretical performance numbers sheds further light on the subject. The Radeon HD 6790 has slightly weaker number-crunching capabilities than the 5770 but considerably more memory bandwidth—even more than the Radeon HD 6870, in fact.
One traditional downside of hobbling an upmarket GPU to compete at the low end is die area. High-end GPUs tend to be larger, making them costlier to produce. Those higher costs in turn reduce margins, giving resulting products less wiggle room when the time comes to slide down the pricing scale. The Radeon HD 6790 isn’t in too bad a position, though. While its Barts GPU is indeed quite a bit larger than Juniper, at 255 mm² vs. 166², it’s not that much portlier than the GeForce GTX 550 Ti’s GF116 chip, which I measured at about 225 mm². Nvidia might have a cost-efficiency edge, but I doubt it’s a terribly great one. It’s also worth noting that Nvidia needs a fully capable GF116 to make a GeForce GTX 550 Ti, but AMD can slip Barts chips that don’t make the cut for the 6870 and 6850 into the 6790.
Now, let’s try to see if the Radeon HD 6790 has the right mix of performance and power efficiency to give the GeForce a run for its money. Time for benchmarks!
Our testing methods
Both AMD and Nvidia have released new graphics drivers since we published our GeForce GTX 550 Ti review, so we tested the new Radeon and re-tested the cards from our GeForce GTX 550 Ti review using Nvidia’s GeForce 270.51 beta release and AMD’s Catalyst 8.84.2_RC2 drivers. We still configured our Radeons’ Catalyst Control Panel like so, leaving optional AMD optimizations for tessellation and texture filtering disabled.
As ever, we did our best to deliver clean benchmark numbers. Tests were run at least three times, and we’ve reported the median result.
Our test system was configured as follows:
|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 18.104.22.1685
Rapid Storage Technology 10.1.0.1008
with Realtek R2.57 drivers
|Graphics||Gigabyte Radeon HD 5770 Super OC 1GB
with Catalyst 8.84.2_RC2 drivers
|AMD Radeon HD 6790
with Catalyst 8.84.2_RC2 drivers
|XFX Radeon HD 6850 1GB
with Catalyst 8.84.2_RC2 drivers
|Zotac GeForce GTS 450 1GB AMP! Edition
with GeForce 270.51 beta drivers
|MSI GeForce GTX 550 Ti Cyclone II 1GB
with GeForce 270.51 beta drivers
|Zotac GeForce GTX 550 Ti AMP! Edition 1GB
with GeForce 270.51 beta drivers
|Zotac GeForce GTX 460 1GB
with GeForce 270.51 beta drivers
|Hard drive||Samsung SpinPoint F1 HD103UJ 1TB SATA|
|Power supply||Corsair HX750W 750W|
|OS||Windows 7 Ultimate x64 Edition
Service Pack 1
Thanks to Intel, Kingston, Samsung, MSI, and Corsair 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:
- Aliens vs. Predator benchmark
- Sid Meier’s Civilization V
- Just Cause 2
- Metro 2033
- Fraps 3.3.2
- GPU-Z 0.5.1
Some further notes on our methods:
- Many of our performance tests are scripted and repeatable, but for Bulletstorm, 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 raised our sample size, testing each Fraps sequence five times per video card, in order to counteract any variability. We’ve included second-by-second frame rate results from Fraps for those games, and in that case, 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 Bulletstorm at a 1920×1200 resolution with 4X AA and 16X anisotropic filtering.
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 on all cards but the Radeon HD 6790, which wasn’t supported by GPU-Z’s temperature probing component. With that card, we ran the load test in a window and jotted down the temperature reported by AMD’s Catalyst Control Center.
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.
I’ve made no secret of my appreciation for Bulletstorm‘s cathartic gameplay and gorgeous environments, so it seems like a fitting start to our round of benchmarking. This shooter was tested at 1680×1050 with 2X antialiasing, medium post-processing quality, and other detail settings cranked up. Since this game has no built-in benchmarking mode, I played through the first 90 seconds of the “Hideout” echo five times per card, reporting the median of average and low frame rates obtained.
The Radeon HD 6790 is off to a nice start, although right off the bat, we can see it’s closer to our slightly souped-up Radeon HD 5770 than to the 6850. The Radeons’ strong overall showing here is likely due to optimizations in AMD’s latest round of drivers, which have boosted performance in games that use deferred shading, like Bulletstorm.
Civ V has several interesting tests, including a built-in compute shader benchmark that measures the GPU’s ability to decompress textures used for the graphically detailed leader characters depicted in the game. The decompression routine is based on a DirectX 11 compute shader. The benchmark reports individual results for a long list of leaders; we’ve averaged those scores to give you the results you see below.
As we noted in our review of the GeForce GTX 550 Ti, pixel-pushing capabilities may play a part in this test. That would explain why the 6790 falls behind despite its much higher peak GFLOPS—the GTX 550 Ti cards still have higher peak pixel fill rates.
In addition to the compute shader test, Civ V has several other built-in benchmarks, including two we think are useful for testing video cards. One of them concentrates on the world leaders presented in the game, which is interesting because the game’s developers have spent quite a bit of effort on generating very high quality images in those scenes, complete with some rather convincing material shaders to accent the hair, clothes, and skin of the characters. This benchmark isn’t necessarily representative of Civ V‘s core gameplay, but it does measure performance in one of the most graphically striking parts of the game. As with the earlier compute shader test, we chose to average the results from the individual leaders.
In this more conventional and shader-intensive rendering test, the Radeon HD 6790 ends up neck-and-neck with the fastest of the two GeForce GTX 550 Ti cards.
Another benchmark in Civ V focuses on the most taxing part of the core gameplay, when you’re deep into a map and have hundreds of units and structures populating the space. This is when an underpowered GPU can slow down and cause the game to run poorly. This test outputs a generic score that can be a little hard to interpret, so we’ve converted the results into frames per second to make them more readable.
The GeForces take the lead here. That said, even 29 FPS should feel relatively smooth in a game like Civilization V, where rapid camera movements aren’t as much of a staple as in fast-paced shooters.
Just Cause 2
Although it’s not the newest kid on the block, JC2 is a good example of a relatively resource-intensive game with flashy DirectX 10 visuals. It doesn’t hurt that the game has a huge, open world and addictively fun gameplay, either.
This title supports a couple of visual effects generated by Nvidia’s CUDA GPU-computing API, but we’ve left them disabled for our testing. The CUDA effects are only used sparingly in the game, anyhow, and we’d like to keep things even between the different GPU brands.
We tested performance with JC2‘s built-in benchmark, using the “Dark Tower” sequence.
Here’s another disappointing showing by the 6790, which barely edges out our souped-up Radeon HD 5770 and slips behind the GTX 550 Tis.
Sometimes, and especially with low-end GPUs like the GeForce GTX 550 Ti, treating yourself to a decent amount of fancy shader effects without killing frame rates means having to sacrifice antialiasing. We ran Metro 2033‘s built-in benchmark using the “High” graphical preset with 16X anisotropic filtering and no antialiasing. PhysX effects were left disabled to ensure a fair fight between all of our contestants.
Now, that’s better. The Radeon HD 6790 gets back in the lead in Metro, cranking out a comfortable 39 FPS.
Aliens vs. Predator
AvP uses several DirectX 11 features to improve image quality and performance, including tessellation, advanced shadow sampling, and DX11-enhanced multisampled anti-aliasing. Naturally, we were pleased when the game’s developers put together an easily scriptable benchmark tool. This benchmark cycles through a range of scenes in the game, including one spot where a horde of tessellated aliens comes crawling down the floor, ceiling, and walls of a corridor.
For these tests, we turned up all of the image quality options to their maximums, along with 2X antialiasing and 16X anisotropic filtering.
The Radeons take this one, no question about it.
AvP also gives us a feel for performance scaling as we ramp up the resolution. I’m not sure what to make of AMD’s promises of 1080p gaming with the Radeon HD 6790, though. Perhaps frame rates would be smoother if we’d disabled antialiasing, but we noticed that the 6790 ran AvP a little choppy at 1080p with 2X AA. It certainly looks like you’d want to cough up the extra dough for at least a Radeon HD 6850 (or a GeForce GTX 460 1GB) if you plan on spending a lot of time beyond 1680×1050.
Despite its portly Barts GPU, the Radeon HD 6790 isn’t much of a power hog. Both of our GeForce GTX 550 Tis draw more juice under load—especially the 1000MHz Zotac variant.
Noise levels and GPU temperatures
I wouldn’t pay too much attention to those noise and temperature numbers, for the simple reason that few (if any) retail Radeon HD 6790 graphics cards are likely to sport the same cooler as our engineering sample. Still, for what it’s worth, the quicker Radeon HD 6850 ran cooler and quieter than the 6790 under load.
The scatter plot below, which tracks overall performance on the Y axis and prices on the X axis, sums up the Radeon HD 6790’s showing quite well. (In case you’re wondering, we worked out overall performance by averaging frame rates across all of our game benchmarks at 1680×1050, Civilization V‘s leader and compute-shader tests excepted. Meanwhile, we grabbed prices corresponding to the cards we tested from Newegg and Amazon.)
Yes, the Radeon HD 6790 ends up slightly above the GeForce GTX 550 Ti cards overall, but look how close the Radeon HD 6850 lies on the price axis—and how much of a leap in performance it provides.
The unfortunate truth is that, for $149, the 6790 is kind of a raw deal. A graphics card without a cause, really. Head to Newegg right now, and you can grab a Radeon HD 6850 for $164.99—that’s before a $20 mail-in rebate that’ll bring the card down to just $144.99, as long as you eventually get that check in the mail. The 6850 produces much higher frame rates across the board, draws about as much power, and may actually emit less noise.
No matter how hard I try, I just can’t think of a situation where I’d recommend the 6790.
Perhaps if you have an irrational hatred for mail-in rebates and have a budget so tight that going over by $15 would force you to mortgage your house and take up residence under a bridge. If you’re that strapped for cash, though, perhaps a $149 graphics card isn’t a wise expense to begin with.