I don’t wish to alarm you, but two very powerful things are converging upon us simultaneously. First, we have the introduction of a new video card from Nvidia aimed smack dab at the soft, chewy center of the price-performance curve. Second, and even more consequentially, we’re on the cusp of a potentially huge upgrade cycle, prompted by one Miss Sandy Bridge, Intel’s sparkling new mid-range CPU. We recently polled you, our faithful readers, about your upgrade plans for Sandy Bridge, and more than a quarter of voters said they intend to upgrade either immediately or “soon.” Those are astonishing numbers, if you think about it.
Mash those two facts together, and you have an inevitable outcome: a number of interested parties would really like to sell you one of these new video cardsor something like itto go with your new system. AMD, Nvidia, and their partners are pumping up their performance, hacking away at prices, and doing everything else they can do grab your attention. Happily, that means some really nice choices should soon be available to you via your favorite online retailer.
One of those choices is our headliner today, the GeForce GTX 560 Ti graphics card. And yes, the name ends in “Ti”that’s not a typo. To decode it, look not toward the Texas-based chip company or the rapper. Instead, think periodic table. This is the GeForce GTX 560 “Titanium,” believe it or not, a name that hearkens waaaay back to 2001 and the GeForce3 Ti graphics card. (Yes, in a shocking example of career stagnation, I was reviewing graphics cards back then just as am I today.) We’ll explain the reasons behind this peculiar naming choice shortly, but first, let’s consider the revamped GPU that drives the GTX 560 Ti.
Please welcome the GF114
If you’ve been following the veritable truckload of new GPU releases over the past four months, you’ll know that Nvidia has been following up its famously late-to-market GeForce GTX 400-series graphics processors with a reworked GTX 500 series that’s arrived in more timely fashion. The somewhat shaky GeForce GTX 480 gave way to the world-beating GTX 580, based on a very similar chip with higher clock speeds, more units enabled, and lower power consumption.
The GTX 560 Ti’s release follows the same basic template. The card is based on the GF114 graphics processor, a reworked version of the GF104 graphics processor that lies under the heatsink of every GeForce GTX 460. That reworking has involved tuning the chip’s design to better fit TSMC’s 40-nm fabrication process. To improve performance and lower power consumption, Nvidia has used faster transistors in the speed-sensitive paths on the chip while deploying low-leakage transistors elsewhere. Beyond those tweaks, the GF114’s architecture is essentially the same as the GF104’s, with no other notable changes. (The GF100-to-GF110 transition included an upgrade to the texture filtering hardware to allow full-rate filtering of FP16 formats, but the GF104 already had that capability.)
At long last, Nvidia has relented from its policy of trying to keep die sizes obscured. The GF114’s die size is, officially, 360 mm². Chip size isn’t a terribly important metric for most folks to know, but it does give us a sense of what a GPU costs to manufacture. The GF114 looks to be just a little smaller than AMD’s Cayman chip but considerably larger than the Barts GPU used in the Radeon HD 6800 series.
So why the Ti?
If you know the history of the GeForce GTX 460, then Nvidia’s decision to bring back the Titanium designation just might make some sense to you. The GTX 460 started life at a rather modest 675MHz clock speed last July, but later versions crept up to well in excess of 800MHz last fall when Nvidia needed an answer to the Radeon HD 6870.
The fact that such a wide range of performance was available under a single product name caused some consternation in various quarters, a problem that was compounded by video card makers’ tendency to refer to these higher-clocked parts as “overclocked,” a word that does not truly apply. After all, the chips have been through a rigorous binning process, qualified for the speed used, and shipped in a boxed product with a full warranty. Those cards are approximately as much overclocked as I am a potential Chippendale’s dancer, which is to say not at all.
Anyhow, the tweaks to the GF114 have given Nvidia some additional performance headroom in several ways. All units on the chip are enabled, whereas one of the eight SM cores on the GF104 is disabled in the GTX 460. Thus, clock for clock, the GTX 560 Ti has more shader and texturing power and more polygon throughput than the GTX 460. Also, clock speeds are up. The GTX 560 Ti’s stock baseline frequency is 822MHz, and its gigabyte of GDDR5 memory runs at 4 GT/s, versus 675MHz and 3.6 GT/s in the original GTX 460.
| Peak shader
|GeForce GTX 460
|GeForce GTX 560
|GeForce GTX 570||29.3||43.9||1405||2928||152.0|
|GeForce GTX 580||37.1||49.4||1581||3088||192.0|
All in all, the new card is a higher class of product than the GTX 460, and Nvidia wanted to make that clear. But apparently, you know, not too clear. Rather than grabbing the obvious next number in the series, say GeForce GTX 565, to indicate higher performance, Nvidia somehow decided to reach deep into its bag of tricks and brush off the Titanium name.
Which is, after all, shiny.
|GeForce GTX 460 1GB||675||336||56||32||3.6||256||160W||$199.99|
|GeForce GTX 560 Ti||822||384||64||32||4.0||256||170W||$249.99|
|GeForce GTX 570||732||480||60||40||3.8||320||219W||$349.99|
|GeForce GTX 580||772||512||64||48||4.0||384||244W||$499.99|
Interestingly enough, the GTX 560 Ti doesn’t directly replace the GTX 460. The 560 Ti will list for $249.99 at online retailers, while the 460 will soldier on at a lower price point with lower performance. My sense is that eventually Nvidia will introduce a GTX 560 non-Ti based on the GF114 that properly replaces the 460.
Although GTX 560 Ti clocks start out much higher than the 460’s, Nvidia claims it’s not just eating up overclocking margin because this is a substantially re-engineered product. In keeping with its usual practice, the company says it has left ample room for board makers to produce higher-clocked variants of the 560 Tiand for enthusiasts to overclock their own cards, if they wish. Leaving such headroom has been part of Nvidia’s business model for some time now, and that tradition apparently continues.
Pictured above is a card based on Nvidia’s GeForce GTX 560 Ti reference design. Although it may look similar to the GTX 460, this is in fact a new, longer PCB (now 9″) attached to a larger, heavier cooler with a trio of heatpipes embedded. The output ports are standard Nvidia for this generation: two dual-link DVIs and one mini-HDMI connector. As you can see, the board requires two 6-pin aux power inputs. Max power draw is rated for 170W, and Nvidia recommends a 500W power supply unit.
We’d expect to see some cards based on this reference design selling at online retailers today for right around the $249.99 suggested price. As with the GTX 460, though, we can expect higher clocks and tremendous variety in board and cooling designs very soon. For instance….
We have three different examples of retail versions of the GeForce GTX 560 Ti for comparison, from the three major enthusiast motherboard manufacturers, all with custom coolers and board designs and all clocked higher than Nvidia’s baseline frequency.
The first such board to arrive in Damage Labs was Gigabyte’s GTX 560 Ti SOC, and it’s pretty remarkable thanks to a dizzying 1GHz GPU clock and 4580 MT/s memory. Gigabyte has left Nvidia’s base 822MHz and 4008 MT/s speeds in the dust, yet this card is slated to sell for only $269.99 when it hits online retailers next week. Those clock speeds mean the SOC will threaten the product sitting at the next rung up Nvidia’s stack, the GeForce GTX 570, in a number of key graphics throughput rates.
Like most of these card makers, Gigabyte cites several factors that purportedly contribute to its product’s superiority over the average reference design. Those include higher-quality components, rigorous testing of chips during their sorting into different speed grades, a custom board design (9.5″ long, or half an inch beyond the reference card), and a bitchin’ cooler with quad heatpipes and dual fans.
In fact, Gigabyte’s next-generation “Windforce” cooler has fans angled slightly outward in a way that the firm claims reduces turbulence (and thus noise.) We like the fact that the gap in between the two fans ought to leave some room for air intake, even with a card installed in the adjacent slot. Many of these fan-based coolers perform poorly in SLI (or in the company of a TV tuner or sound card, for that matter.)
The real secret to the SOC’s ridiculously high clock speeds, though, may be the component pictured above: a Proadlizer film capacitor from NEC/TOKIN. I don’t believe we’ve seen one of these on a graphics card before. It sits on the back side of the card right between the VRMs and the GPU and memory chips, purportedly providing “excellent noise absorption performance” and “high switching frequency.” The card’s six-phase power and other attributes may contribute to its ability to sustain higher frequencies, as well. Whatever the case, Gigabyte’s SOC is a pretty vivid illustration of a board maker taking the GTX 560 Ti to another level, and since it hit our labs first, we were able to run it through our full suite of performance tests.
Aesthetically, Asus’ GTX560 Ti Direct CU II TOP is my favorite GTX 560 Ti card so far. Whereas Gigabyte’s effort comes from the heavy-metal-flames school of industrial design, Asus opts for the more understated red-and-black 1985 VW GTI approach. The matte paint and racing stripes work pretty well, in my opinion.
Asus’ cooler has one less heatpipe, but all three of the pipes snake across the surface of the GPU, making direct contact, which purportedly results in better heat conduction. This card is unique among the group in sporting a metal brace running the length of the PCB to prevent warping. Asus is also proud of its selection of “super alloy” components for this board’s power delivery circuitry, which it claims provides superior performance and higher overclocking headroom than the reference design.
You may have to test that out for yourself, though, because the clock speeds on this card are a little more understated than the Gigabyte’s, with a 900MHz GPU core and 4.2 GT/s memory. Much like the Gigabyte, this card is priced at $269.99 and slated for availability at online retailers in 7-10 days. Asus also has a stock-clocked version of the GTX 560 Ti that should be available immediately for $249.99.
MSI’s choice of a brushed-metal-and-chrome approach rounds out our sampling of industrial design schools in PC hardware nicely. Nvidia’s reference card covers the black-and-neon theme, so I believe all we’re missing is a military/camo scheme to finish off all of the major schools.
Anyhow, MSI’s rendition of the GTX 560 Ti is adorned with the company’s familiar “Twin Frozr II” dual-fan cooler with quad heatpipes. This cooler performed well for us aboard MSI’s high-clocked (810MHz) version of the GTX 460, and we expect similarly good things here. Like the other guys, MSI touts its component selectionthese are claimed to be “military class”and custom cooler as sources of superiority to the poor, battered reference design.
At 880MHz and 4200 MT/s, this card is a little slower than the Asus and a lot slower than the Gigabyte, but MSI will only ask $259.99 for this offeringand it should be available immediately. Versions with 900 and 950MHz clock speeds, still at 4.2 GT/s memory, are planned, as well.
A pair of aces from AMD
You didn’t think AMD would let this momentous occasion pass without injecting a little excitement of its own into the conversation, did you? The Radeon guys have several responses to the GTX 560, all of which make things more interesting for us today.
The first response is, perhaps inevitably, a higher-clocked version of the Radeon HD 6870 with, you guessed it, a custom cooler that has dual fans and triple heat pipes. XFX’s Radeon HD 6870 Black Edition runs at 940MHz with 4.6 GT/s memory, up from the 6870’s stock speeds of 900MHz and 4.2 GT/s. Currently, it’s listed for $259.99 at Newegg, which is a little steep considering some other things we’re about to unload on you. There is a $30 mail-in rebate attached, for those who enjoy that peculiar form of abuse. (We’re not fans of mail-in rebates, whose perverse business model depends on keeping redemption rates low.) I believe this is the highest-clocked variant of the Radeon HD 6870 available, and it should provide a more suitable challenge for the GTX 560 Ti.
The next prong of AMD’s counterattack is a lower-priced variant of the Radeon HD 6950 whose memory size has been cut from 2GB to 1GB. This is perhaps the product that AMD should have introduced first, had it not been angling to press its advantage on multiple display support and Eyefinity gaming. One gigabyte of video RAM should be sufficient for the vast majority of users who have a single display with a resolution of 1920×1200 or less.
The card pictured above is just an engineering sample from AMD, but already, there’s a Sapphire-branded version of the Radeon HD 6950 1GB listed at Newegg for AMD’s suggested price of $259.99. If that price holds steady over time, the Radeon HD 6950 1GB should present formidable competition to higher-clocked versions of the GTX 560 Ti.
As if that weren’t enough, AMD is also dropping the suggested prices on a couple of other offerings. The stock Radeon HD 6870 is down to $219.99, and the Radeon HD 6950 2GB drops ten bucks to $289.99, with rebates adding potential savings beyond that. (Maybe.)
Along with the GTX 560 Ti’s introduction, these changes add up to a nice downward shift in overall pricing. As I said, GPU and video card makers want in on some of that sweet Sandy Bridge upgrade action. We’ll do our best to help you decide whether to upgrade by testing all of these cards against the incumbents in this price range, along with a couple of older GeForces that may be similar to a card you already own.
The first of those is an original GeForce 8800 GT 512MB. This card was a favorite of ours several years ago and remains in many enthusiasts’ systems today. We’ll let it be our proxy for cards of that era. If you own a GeForce 8800 GTS, GeForce 9800 GT, or a Radeon HD 3850 or 3870, you own roughly the same class of GPU.
The second older card we’ve tested is the “reloaded” version of the GeForce GTX 260, introduced over two years ago. The GTX 260 had an unusually long run as an attractive video card option due to shortages of DX11-class GPUs and the stagnation of GPU requirements in PC games. If you own a GeForce GTX 280 or a Radeon HD 4870 1GB, you have the same basic class of GPU and can probably consider your need to upgrade based on this card’s performance.
These older DirectX 10 cards wouldn’t run every single game at the settings we used, so we occasionally had to leave them out.
With its latest Catalyst 11.1a drivers, AMD has introduced some performance-tuning features to its user control panel, including an interesting new slider to limit tessellation levels used in DirectX 11 games. Since these options are becoming fairly complex, we’ll show you how we had the Radeons configured.
We believe these settings are the closest match to Nvidia’s defaults. We’ll have to play with the tessellation slider at a later date, when we have more time to devote to it. As the “AMD optimized” checkbox indicates, AMD intends to introduce profile-based tessellation reduction in future driver revisions (perhaps as a means of addressing things like the HAWX 2 controversy), but it tells us that no applications have been profiled yet.
Our testing methods
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 systems were configured like so:
|North bridge||X58 IOH|
|Memory size||12GB (6 DIMMs)|
|Memory type||Corsair Dominator CMD12GX3M6A1600C8
DDR3 SDRAMat 1600MHz
|Memory timings||8-8-8-24 2T|
|Chipset drivers||INF update 22.214.171.1245
Rapid Storage Technology 126.96.36.1994
with Realtek R2.51 drivers
|Asus Radeon HD 6850 1GB
with Catalyst 11.1a drivers
|Sapphire Radeon HD 6870 1GB
with Catalyst 11.1a drivers
|XFX Radeon HD 6870
Black Edition 1GB
with Catalyst 11.1a drivers
with Catalyst 11.1a drivers
with Catalyst 11.1a drivers
with Catalyst 11.1a drivers
|Asus GeForce 8800
with ForceWare 266.58 drivers
GTX 260 896MB
with ForceWare 266.58 drivers
| GeForce GTX 460
with ForceWare 266.58 drivers
|MSI Hawk Talon Attack GeForce GTX 460 1GB 810MHz
with ForceWare 266.58 drivers
560 Ti 1GB
with ForceWare 266.56 drivers
GeForce GTX 560 Ti 1GB SOC
with ForceWare 266.56 drivers
|Zotac GeForce GTX 570 1280MB
with ForceWare 266.58 drivers
|Hard drive||WD RE3 WD1002FBYS 1TB SATA|
|Power supply||PC Power & Cooling Silencer 750 Watt|
|OS||Windows 7 Ultimate x64 Edition
DirectX runtime update June 2010
Thanks to Intel, Corsair, Western Digital, Gigabyte, and PC Power & Cooling 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:
- Unigine Heaven 2.1
- ShaderToyMark 0.1.0
- 3DMark Vantage 1.0.2
- Aliens vs. Predator benchmark
- Battlefield: Bad Company 2
- F1 2010
- Sid Meier’s Civilization V
- Metro 2033
- StarCraft II
- Fraps 3.2.3
- GPU-Z 0.5.0
Some further notes on our methods:
- Many of our performance tests are scripted and repeatable, but for some of the games, including Battlefield: Bad Company 2, we used the Fraps utility to record frame rates while playing a 60-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 Yokogawa WT210 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 Battlefield: Bad Company 2 at a 1920×1080 resolution with 8X AA and 16X anisotropic filtering. We test power with BC2 because we think it’s a solidly representative peak gaming workload.
We measured noise levels on our test system, sitting on an open test bench, using an Extech 407738 digital sound level meter. The meter was mounted on a tripod approximately 10″ 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.
Key rates and performance in directed tests
| Peak bilinear
| Peak bilinear
| Peak shader
|GeForce GTX 460
|GeForce GTX 460 1GB 810MHz||25.9||47.6||47.6||1089||1620||124.8|
|GeForce GTX 560
|GeForce GTX 560
|GeForce GTX 570||29.3||43.9||43.9||1405||2928||152.0|
|GeForce GTX 580||37.1||49.4||49.4||1581||3088||192.0|
|Radeon HD 6850||25.3||37.9||19.0||1517||790||128.0|
|Radeon HD 6870||28.8||50.4||25.2||2016||900||134.4|
|Radeon HD 6870
|Radeon HD 6950||25.6||70.4||35.2||2253||1600||160.0|
|Radeon HD 6970||28.2||84.5||42.2||2703||1760||176.0|
The theoretical peak numbers in the table above will serve as a bit of a guide to what comes next. Different GPU architectures achieve more or less of their peak rates in real-world use, depending on many factors, but these numbers give us a sense of how the various video cards compare.
One match-up to watch: Gigabyte’s GeForce GTX 560 Ti SOC versus the much pricier GeForce GTX 570. The 560 Ti SOC is faster in all but two categories: rasterization rate and memory bandwidth. Could this version of the 560 Ti knock off its elder sibling?
This color fill rate test tends to be limited primarily by memory bandwidth rather than by ROP rates. That’s why the GTX 560 Ti SOC is a little slower than the GTX 570.
Despite a lower peak theoretical rate, the Radeon HD 6870 outperforms the GTX 560 Ti slightly in the integer filtering test. However, when we switch to FP16 texture formats, the GTX 560 Ti has a major advantage, both in theory and in delivered performance.
Notice the curious case of the Radeon HD 6970 here, which is slower than the 6950 in both tests. The cause of that anomaly is the PowerTune power capping feature in the Radeon HD 6900-series GPUs. The 6970, for whatever reason, is hitting its power cap and being limited more than the 6950. To demonstrate what happens without the power cap, I raised the PowerTune limit on the 6970 by 20% via AMD’s Overdrive utility. As you can see, that put the 6970 firmly out ahead of the 6950, as one would expect.
The first tool we can use to measure delivered pixel shader performance is ShaderToyMark, a pixel shader test based on six different effects taken from the nifty ShaderToy utility. The pixel shaders used are fascinating abstract effects created by demoscene participants, all of whom are credited on the ShaderToyMark homepage. Running all six of these pixel shaders simultaneously easily stresses today’s fastest GPUs, even at the benchmark’s relatively low 960×540 default resolution.
Up next is a compute shader benchmark built into Civilization V. This test 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.
Finally, we have the shader tests from 3DMark Vantage.
Clockwise from top left: Parallax occlusion mapping, Perlin noise,
GPU cloth, and GPU particles
Clearly, the question of which GPU architecture’s shader performance is better depends heavily on the sort of workload involved. To oversimplify a bit, we know that the more vertex-shader-intensive tests like 3DMark’s cloth and particles tests are generally dominated by the GeForces, while the Radeons perform better in the more pixel-shader-intensive tests.
Cross-brand comparisons are difficult for that reason, but interestingly enough, we have a couple of sibling rivalries worth watching. The aforementioned battle between the GTX 560 Ti SOC and the GTX 570 remains fairly tight, but the GTX 570 has a clear edge in most cases. Meanwhile, the Radeon HD 6870 Black Edition somehow outperforms the theoretically much superior Radeon HD 6950 in couple of tests. We’d chalk that up to the PowerTune cap on the 6950, which the 6870 lacksand which is much more likely to be an issue in these synthetic tests than when running a typical gaming workload.
We can measure geometry processing speeds pretty straightforwardly with the Unigine Heaven demo. This demo doesn’t really make good use of additional polygons to increase image quality at its highest tessellation levels, but it does push enough polys to serve as a decent synthetic benchmark.
Although the theoretical rasterization rates for the Radeon HD 6950 and the GeForce GTX 560 Ti are very similar, good tessellation performance involves much more than just rasterizationand Nvidia has an undeniable architectural advantage in this generation of GPU in terms of overall geometry processing throughput. These results reflect that.
F1 2010 steps in and replaces CodeMasters’ previous effort, DiRT 2, as our racing game of choice. F1 2010 uses DirectX 11 to enhance image quality in a few, select ways. A higher quality FP16 render target improves the game’s high-dynamic-range lighting in DX11. A DX11 pixel shader is used to produce soft shadow edges, and a DX11 Compute Shader is used for higher-quality Gaussian blurs in HDR bloom, lens flares, and the like.
We used this game’s built-in benchmarking facility to script tests at multiple resolutions, always using the “Ultra” quality preset and 4X multisampled antialiasing.
In formulating the testing regimen for this review, I vowed to focus on the wildly popular 1920×1080 resolution as much as possible. In this case, the GTX 560 Ti is generally a little slower than the stock Radeon HD 6870 at that resolution, as it is across the board. At the two lower resolutions, the GTX 560 Ti SOC looks lethal, outperforming the 6950 1GB, but that changes at 2560×1600.
The thing is, this is largely bickering over comparative numbers, which is helpful as a part of the overall performance picture, but consider this fact: even the slowest card in the test, the original GTX 460 1GB, delivers an average of 42 frames per second and a minimum of 39 FPS. In other words, any of these cards will run this game just fine, even at the “Ultra” quality preset we’re using.
In addition to the compute shader test we’ve already covered, 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.
The high-quality pixel shaders used in these leader scenes appear to map well to AMD’s shader architectures, both the older Barts vec5 and the newer Cayman vec4, than they do to Nvidia’s Fermi shaders.
Another benchmark in Civ V focuses, rightly, 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.
When it comes to core gameplay, the GeForce cards tend to be a little faster than the Radeons in Civ V. For whatever reason, there’s very little separation between the 6870, 6950, and 6970, while the GeForces tend to scale up in performance as expected from one model to the next. Perhaps the large number of detailed models on screen is causing the Radeons to hit a geometry throughput bottleneck? That might also explain why the Radeons aren’t much faster here, at 1920×1080, than they were in our past tests of the same game at 2560×1600.
We’ve included the GeForce GTX 260 and 8800 GT in these results even though the game has to fall back to DirectX 10 in order to run on them. Obviously, even the GTX 260 is overmatched at these quality settings.
Up next is a little game you may have heard of called StarCraft II. We tested SC2 by playing back an epic eight-player match using the game’s replay feature. This particular match was about 30 minutes in duration, and we captured frame rates over that time using the Fraps utility. Thanks to the relatively long time window involved, we decided not to repeat this test multiple times, like we usually do when testing games with Fraps. After we’d captured all of the data, we decided to focus our sample period on the last 500 seconds of the game, when the action was most intense. We’ve focused our frame-by-frame graphs on an even smaller 200-second portion of that period in order to keep them readable.
We tested at the settings shown above, with the notable exception that we also enabled 4X antialiasing via these cards’ respective driver control panels. SC2 doesn’t support AA natively, but we think this class of card can produce playable frame rates with AA enabled—and the game looks better that way.
Notice that we broke our rule and tested at 2560×1600 in this game. We did so because StarCraft II isn’t really GPU-bound with any of these newer graphics cards at 1920×1080. Only at 2560×1600 could we really stress these GPUs, although, sadly, going to this resolution caused the GeForce 8800 GT to hit the limits of its 512MB frame buffer.
The Radeons have a pretty pronounced advantage in this game, as they have every time we’ve tested SC2. Any of the newer cards would be a nice upgrade from the Georce GTX 260, though.
Battlefield: Bad Company 2
BC2 uses DirectX 11, but according to this interview, DX11 is mainly used to speed up soft shadow filtering. The DirectX 10 rendering path produces the same images.
We turned up nearly all of the image quality settings in the game. Our test sessions took place in the first 60 seconds of the “Heart of Darkness” level.
The GTX 560 Ti and the 6870 Black are in a dead heat here, while the GTX 560 Ti SOC is looking like a giant-killer by placing in between the Radeon HD 6970 and the GTX 570, both much more expensive cards.
We decided to test Metro 2033 at multiple image quality levels rather than multiple resolutions, because there’s quite a bit of opportunity to burden these GPUs simply using this game’s more complex shader effects. We used three different quality presets built into the game’s benchmark utility, with the performance-destroying advanced depth-of-field shader disabled and tessellation enabled in each case.
We’ve included the two older, DX10-only cards here, even though they can’t handle DX11 tessellation, simply for comparison. Although they’re doing less work, they don’t gain any major advantage over the newer DX11 cards in terms of frame rates, as you’ll see.
In our past testing with this game, the trend has been that the Radeons grow relatively stronger as the quality level rises, and that trend appears to hold once more, though not in terribly pronounced fashion. Generally, the GTX 560 Ti matches the Radeon HD 6870 Black, while the GTX 560 Ti SOC closely shadows the 6950 1GB.
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 the max, with two exceptions. We held the line at 2X antialiasing and 8X anisotropic filtering simply to keep frame rates in a playable range with most of these graphics cards.
The mini-trend of near parity between two pairs of competitors, the 6870 Black-GTX 560 Ti and the 6950 1GB-GTX 560 Ti SOC, continues here. The relative standings are almost identical, regardless of the display resolution.
Now for some power and noise testing. We’ve included all three of the GTX 560 Ti cards with custom coolers and clock speeds for comparison.
By the way, we’ve changed our workload for the “load” tests. This time around, we’re using Battlefield: Bad Company 2 at 1920×1080 to load up the system. We’ve found that our test system draws more power while running this game than most others, making it a solid choice for the job, but these results aren’t directly comparable to our past articles where we used Left 4 Dead 2.
With the exception of the much higher clocked Gigabyte SOC card, all of the systems with GTX 560 Ti cards draw less power at idle than those with Radeons installed. The situation is reversed when running a game, as the GTX 560 Ti-based systems require more power than the competition. XFX’s 6870 Black Edition is particularly efficient.
One slight puzzle here is the case of the Radeon HD 6950 1GB, which is hungrier for power than its 2GB variant. AMD tells us the 6950 1GB may use older, less dense DRAM chips built on a larger fabrication process, which could explain some of the higher power draw. Factor in things like VRM and PSU inefficiencies, and that may go a long way toward explaining the 18W difference in total system power draw. However, it’s possible our early review sample from AMD, which isn’t a finished, shipping product, has other issues.
Another interesting item: notice how the GTX 560 Ti SOC and the GTX 570 have very similar power draw to go along with their fairly closely matched performance. Funny how two cards based on chips of different sizes can arrive at the same basic power-performance balance, isn’t it? Once you push past a certain point with higher clock speeds on a smaller chip, the exponential increases in power draw required to push further begin to cause problems. Gigabyte’s 560 Ti SOC isn’t at a bad place at all, but the GF114 looks to be near its practical limits here for a consumer product.
Noise levels and GPU temperatures
The combination of noise levels and GPU temperatures offered by the GTX 560 Ti reference card looks nearly ideal, and the XFX 6870 Black Edition performs so much like it, we had to do a double-take. Both are tuned for low noise levels, obviously, yet they don’t allow GPU temperatures to rise very high at all, as these things go.
Clearly, the three big motherboard makers are tuning their thermal solutions with an eye toward keeping GPU temperatures low, even if that means higher noise levels under load. As we’ve said before, we think that’s unfortunate. We’d prefer a little tolerance for higher temperatures if it means the fans will be quieter under load.
Of those three, the most impressive cooling performance comes from the Gigabyte SOC card. Despite the fact that its 1GHz GPU has substantially more heat to dissipate (as its readings on the watt meter indicate), the Gigabyte cooler keeps the chip at 63° C while producing less noise than the stock GTX 570 and 6970 coolers, both respectably quiet solutions.
MSI’s twin-fan cooler also looks to be very potent without producing too much noise. We just wish we could trade an increase of 15° C or so in GPU temperature for whatever that would give us on the decibel meter. This could well be the quietest card of the group with different tuning. It’s still quite good as it is, though.
Asus’ DirectCU II cooler is the loudest in the entire bunch, despite the fact that several other products have more heat to dissipate. Rather than hiss innocuously like most cooling fans, the fans on the Asus cooler emit a mid-pitched whine that registers strongly on our decibel meter. That’s unfortunate, because we really liked Asus’ first-generation DirectCU cooler. This one evidently needs some work in order to be competitive.
As you’ve no doubt gathered, this is a ridiculously close contest between some formidable competitors. Sorting out which one is best isn’t going to be easy, but I have a few definite thoughts on these matters.
Let’s begin by going to our price-performance scatter plot, generated by averaging the frame rates across all six of our real-world gaming tests. We’ve used the results from the 1920×1080 resolution where possible, since that is our focus for this review. Our prices come from two sources. For any new products or cards with revised pricing, we’ve taken AMD and Nvidia at their word. For the rest, we’ve used the lowest prevailing pricethat is, the lowest price with multiple listingsat Newegg. The one exception: cards like the 810MHz version of the GTX 460 1GB are unique, so we just used the exact listed price for such products. As always, mail-in rebates were not factored into the mix.
Generally, the best deals tend to gravitate toward the top left corner of the plot, while the worst deals will be closer to the lower right corner.
Now we have a nice visual on the price-performance equation, and I can distract you by talking about other things, as well. I’d like to start explaining a practical reality for those folks using a single monitor at 1920×1200, 1920×1080, or less. For the most part, with today’s games, any of the products in the scatter plot above will serve you well. For our tests, we’ve intentionally chosen games with particularly graphics-intensive workloads, and we’ve tested them at peak quality levels with high degrees of antialiasing and the like. You saw how we had to bump StarCraft II up to 2560×1600 just to stress the GPUs. You should also know that we ruled out testing some really popular games, including the shooters Medal of Honor and Call of Duty: Black Ops, simply because they ran constantly at 90 FPS on most of these video cards.
Although the Radeon HD 6850 and GTX 460 1GB have frame-rate averages in the mid-30s in the scatter plot above, they’re going to be more than competent for most contemporary games at 1920×1080 with just a few minor compromises in image quality, such as dropping down from 8X to 4X antialiasing. Either of them would be a massive upgrade from a GeForce 8800 GT or something like it, as our tests made abundantly clear. They’re quite a bit more capable than a GeForce GTX 260, as well. You can spend less than $200 and do quite well for yourself in today’s market.
The Radeon HD 6870 is a great deal at $220, too, and kind of ruins the stock-clocked GTX 560 Ti’s coming out party, given how close the two are in overall performance.
There’s something to be said for spending a little more to get the best option available, though, and in this broad price class, the best options are at $259 to $269, in my view. The higher-clocked variants of the GeForce GTX 560 Ti and the 1GB version of the Radeon HD 6950 stand out.
On the AMD side, stepping up from Barts to Cayman gets you higher geometry throughput and the new EQAA antialiasing modes. These additions have taken a bit of the shine off of the 6870, since they’re features that every DirectX 11 GeForce has included from the beginning. (EQAA mirrors Nvidia’s CSAA by producing higher quality edge antialiasing without much of a performance hit.) The 6950 1GB is a steal at $259. Unless you’re planning on driving a four-megapixel display or multiple monitors via Eyefinity, you’re not likely to miss the second gigabyte of video RAM in this version of the card, either.
For ten bucks more, Gigabyte’s SOC version of the GTX 560 Ti looks to be a singularly good deal, with even higher performance than the 6950 1GB in our overall index. The GTX 560 Ti SOC is a more polished product than the our early 6950 1GB sample, with measurably lower noise levels. Going with a GeForce will get you even higher geometry throughput than Cayman, too, a feature that may have contributed to the 560 Ti’s performance advantage in Civilization V and could matter more in future games. Unfortunately, you’ll have to wait a week before this particular card becomes available, and if it proves too popular, it might be hard to find.
Had we fully tested the MSI version of the GTX 560 Ti at 880MHz and $259, it would probably sit right on top of the 6950 1GB in our price-performance scatter plotand the MSI card is slated to become available today. All told, the MSI GTX 560 Ti “OC” and the Radeon HD 6950 1GB are incredibly close competitors, and you could probably do well simply by taking your pick between the two.
Yes, all of this hoopla has again come down to small slivers of difference between the best options from AMD and Nvidia, amazingly enough. These two companies and their partners have a very good handle on the competitive situation, and they’ve positioned their products accordingly. We’re left to sort things out, and sometimes, the best answer is “take your pick.”
One thing we do know: these new entries at $259-269 are where you want to be. The GeForce GTX 570, Radeon HD 6950 2GB, and Radeon HD 6970 don’t deliver enough additional performance to justify their pricesat least, not at the 1080p display resolution on which we’ve focused today.