Sign in with
Sign up | Sign in
GeForce GTX 650 Ti Review: Nvidia's Last Graphics Card For 2012
By , Igor Wallossek,
1. GeForce GTX 650 Ti: The Last Kepler-Based Card For 2012

We've seen Nvidia launch three graphics cards in the last two months, all based on its Kepler architecture. Today, we bring you the fourth, and the last desktop-oriented introduction, Nvidia says, for 2012. Say hello to the GeForce GTX 650 Ti.

There's a fairly substantial price gap between the new $120 GeForce GTX 650 and $230 GeForce GTX 660, so this board's arrival is really not a surprise. After all, Nvidia needs something to do battle with AMD's Radeon HD 7770 and 7850, along with replacing its own GeForce GTX 560. The Radeon HD 6850 and 6870 are still floating around, too, and those cards remain attractive options in the same price range. The GeForce GTX 650 faces a formidable battle if it wants to prove its worth at $150.

Nvidia arms its new value-oriented contender with the same GK106 GPU we saw in Nvidia GeForce GTX 650 And 660 Review: Kepler At $110 And $230. Only, this time around, parts of the chip are disabled in order to create a performance profile that fits between the GeForce GTX 650 and 660.

GK106's third (and curiously asymmetric) GPC is disabled, as is one of the chip's ROP partitions and one of its 64-bit memory interfaces. The resulting configuration yields four streaming multiprocessors totaling 768 CUDA cores and 64 texture units. The remaining two ROP clusters can output 16 full-color raster operations per clock, and the aggregate memory interface is 128-bits wide.


GeForce GTX 650GeForce GTX 650 Ti
GeForce GTX 660GeForce GTX 560GeForce GTX 460
Shader Cores
384
768960
336336
Texture Units
32
64
80
5656
Color ROPs
16
1624
3232
Fabrication process
28 nm
28 nm28 nm40 nm40 nm
Core/Boost Clock
1058 MHz
925 MHz980/1033 MHz
810 MHz675 MHz
Memory Clock
1250 MHz GDDR5
1350 MHz GDDR51502 MHz GDDR5
1002 MHz GDDR5900 MHz GDDR5
Memory Bus
128-bit
128-bit192-bit
256-bit256-bit
Memory Bandwidth
80 GB/s
86.4 GB/s144.2 GB/s128.2 GB/s
115.2 GB/s
Graphics RAM
1 or 2 GB GDDR5
 1 or 2 GB GDDR5
2 GB GDDR5
2 GB GDDR51 GB GDDR5
Power Connectors
1 x 6-pin1 x 6-pin
1 x 6-pin2 x 6-pin2 x 6-pin
Maximum TDP
64 W
110 W
140 W
160 W
160 W
Price
$117 - $150 (Newegg)
~$150 (MSRP)$230 - $250 (Newegg)
$167 - $200 (Newegg)
End Of Life


Most of the GeForce GTX 650 Ti's specifications end up somewhere between the recently-introduced GeForce GTX 650 and GTX 660, except for the memory subsystem. A similar 128-bit interface and 16 raster ops per clock are virtually identical to the GeForce GTX 650. A small bump in data rate accounts for the tiny bandwidth increase from 80 GB/s to 86.4 GB/s. Unfortunately, that's a long way off from last generation's GeForce GTX 460 with the 256-bit bus. Even AMD's older Radeon HD 6850 and 6870 cards offer 128 and 134 GB/s of throughput.

If Nvidia's new budget-oriented board has one weakness for folks looking to game at higher resolutions with anti-aliasing applied, based on what we know about the GeForce GTX 650 and 660, memory bandwidth is going to be it. This is an area where overclocking stands to really help the reference configuration's performance. 

Nvidia's GeForce GTX 650 Ti

The GeForce GTX 650 Ti's diminutive 5.75" x 4" PCB reminds us of the low-end GeForce GT 430. It's a surprisingly fast card, however, despite its compact dimensions.

Nvidia's reference design is equipped with 1 GB of GDDR5 memory. The back side of the PCB is bare, though you'll find memory modules there on more expensive 2 GB versions of the card. By default, memory operates at 1350 MHz, yielding an effective 5400 MT/s data rate. The GK106 GPU itself runs at 925 MHz, and there is no GPU Boost functionality available, similar to what we already saw from the GeForce GTX 660.

The sample we received only has one mini-HDMI and two dual-link DVI outputs, though we know the Kepler-based GPUs we've reviewed so far actually support up to four screens at a time. Add-in board vendors have the option to expose additional display connectivity if they choose, and we'll be checking out a couple of partner boards on the following pages.

A 110 W TDP requires a single six-pin auxiliary power connector, since a PCI Express x16 slot is only able to supply 75 W.

There is no SLI connector on the GeForce GTX 650 Ti. Although some lower-end boards support multi-card configurations over the PCI Express, Nvidia deliberately handicaps this model by preventing cost-conscious gamers from buying one card today and adding a second one down the road. AMD, in comparison, allows for CrossFire operation on its Radeon HD 7770 and 7750 boards.

The reference cooling solution employs one 75 mm axial fan, and, according to our measurements, it does a satisfactory job of keeping Nvidia's 28 nm cool without generating much noise.

2. Zotac GeForce GTX 650 Ti AMP! Edition

Zotac's AMP! Edition cards are notorious for their aggressive factory overclocks, and the GeForce GTX 650 Ti in our possession is no exception. It sports a 1033 MHz core frequency, 108 MHz over Nvidia's reference spec, and memory operating at 1550 MHz, which is 200 MHz beyond Nvidia's design. The GDDR5 overclock, in particular, is expected to yield performance improvements in our graphics benchmarks.

This board also boasts 2 GB of memory, which is twice the capacity of Nvidia's stock config.

All of those tweaks make the GeForce GTX 650 Ti AMP! Edition Zotac's flagship model. It commands a $30 premium over the recommended price Nvidia gave us, though you should be able to find Zotac's standard 2 GB model for $170 and a reference-clocked 1 GB card for $155, if the extra memory isn't important to you.

As promised on the previous page, adding an extra gigabyte of memory to the GeForce GTX 650 Ti means populating the back of its PCB with modules. Although Zotac's PCB is the same size as our sample from Nvidia, its layout is slightly different.

Zotac exposes two full-sized HDMI and two dual-link DVI outputs on its AMP! Edition board, enabling four independent displays that you can arrange into one three-screen Surround array with a fourth desktop monitor. Unfortunately, we don't expect any vendor's GeForce GTX 650 Ti to deliver suitable three-screen gaming performance, so any multi-monitor arrangement is going to be for productivity purposes only.

One auxiliary six-pin power connector is found on the top of Zotac's card, similar to the reference model.

Zotac uses an 85 mm axial fan for its cooling solution, which is 10 mm larger than Nvidia's own design. Both boards feature a similarly-sized heat sink, though.

3. Gigabyte GeForce GTX 650 Ti (GV-N65TOC-2GI)

Compared to Nvidia's reference design, Gigabyte's GV-N65TOC-2GI appears huge as a result of its massive cooler extending almost four inches beyond its PCB.

A 1033 MHz core clock is 108 MHz faster than the stock spec, but GDDR5 memory operating at 1350 MHz matches Nvidia's reference specification.

The circuit board that Gigabyte uses is the same size as Nvidia's reference model, though the component layout is clearly customized. The fact that we find memory around back indicates that we're dealing with 2 GB of capacity. Gigabyte tells us its card will sell for $170, or $175 with the Assassins Creed 3 game bundle. This is a deal that Nvidia offers with the GeForce GTX 650 Ti through select retailers (excluding China and Japan).

Gigabyte exposes a unique combination of HDMI, VGA, and two dual-link DVI connectors to support four simultaneous displays. VGA is clearly out of vogue amongst enthusiasts, but there are still plenty of screens equipped with the analog input.

Most GeForce GTX 650 Tis will undoubtedly feature auxiliary power connectors toward the back of the card. Gigabyte's implementation, however, uses an extra-long heat sink. So, the company put its six-pin plug in the middle.

Gigabyte's cooling solution employs two 95 mm fans, each 20 mm in diameter larger than the reference model's single 75 mm fan. Instead of leaning on a solid heat sink, this board features two heat pipes to purportedly transfer heat more quickly through the sink. This is easily overkill on a GeForce GTX 650 Ti, as we'll see in the thermal benchmarks.

4. Test Setup And Benchmarks

Nvidia's GeForce GTX 650 Ti can't be expected to deliver ample performance beyond 1920x1080, and cards in this class aren't fast enough to accelerate the highest detail settings with anti-aliasing cranked up. So, we're running today's tests at high quality settings without anti-aliasing, and then again again with the feature turned on. In many cases, we're forced to drop the quality preset in order to realize playable frame rates with anti-aliasing turned on.

We're including Nvidia's GeForce GTX 650 and 660 to evaluate the performance of Kepler-based cards on either side of the GTX 650 Ti in the company's line-up. We're also testing the GeForce GTX 460 192-bit, factoring in the capabilities of a previous-generation card. As a point of reference, our GeForce GTX 460 is very similar to a GeForce GTX 560 SE, in case you have one of those sitting around. A GeForce GTX 560 is in there, too, to represent the $170 card you could have purchased before Kepler-based boards started invading the same space.

AMD's Radeon HD 7770 stands in as the company's low-end contender. The Radeon HD 6850 and 6870 are previous-gen boards selling between $150 and $170. They're still available, and therefore still relevant to our analysis of which card to buy.

Finally, AMD recently introduced a 1 GB Radeon HD 7850, which sells for about $180. We've seen the card as low as $165 without rebates, though, so we're inclined to include it as potent competition in the same price range.

Test System
CPU
Intel Core i7-3960X (Sandy Bridge-E), 3.3 GHz, Six Cores, LGA 2011, 15 MB Shared L3 Cache, Hyper-Threading enabled, Overclocked to 4.2 GHz
Motherboard
ASRock X79 Extreme9 (LGA 2011) Chipset: Intel X79 Express
Networking
On-Board Gigabit LAN controller
Memory
Corsair Vengeance LP PC3-16000, 4 x 4 GB, 1600 MT/s, CL 8-8-8-24-2T
Graphics
Zotac GeForce GTX 650 Ti AMP! Edition
1033 MHz GPU, 2 GB GDDR5 at 1550 MHz

Gigabyte GeForce GTX 650 Ti OC
1033 MHz GPU, 2 GB GDDR5 at 1350 MHz

Reference GeForce GTX 650 Ti
925 MHz GPU, 1 GB GDDR5 at 1350 MHz

Reference GeForce GTX 660
980/1033 MHz Base/Boost GPU, 2 GB GDDR5 at 1502 MHz

MSI GeForce GTX 560
810 MHz GPU, 2 GB GDDR5 at 1002 MHz

Zotac GeForce GTX 460 192-bit
675 MHz GPU, 768 MB GDDR5 at 900 MHz

Gigabyte GeForce GTX 650
1111 MHz GPU, 2 GB GDDR5 at 1250 MHz

EVGA R7850 Core Edition, Radeon HD 7850 1GB
860 MHz GPU, 1 GB GDDR5 at 1200 MHz

Reference Radeon HD 6870
900 MHz GPU, 1 GB GDDR5 at 1050 MHz

Reference Radeon HD 6850
775 MHz GPU, 1 GB GDDR5 at 1000 MHz

Reference Radeon HD 7770
1000 MHz GPU, 1 GB GDDR5 at 1125 MHz

All overclocked cards reduced to reference specification for testing
Hard Drive
Samsung 470-Series 256 GB (SSD)
Power
ePower EP-1200E10-T2 1200 W
ATX12V, EPS12V
Software and Drivers
Operating System
Microsoft Windows 7 x64, Service Pack 1
DirectX
DirectX 11
Graphics Drivers
Nvidia: 306.38 beta
AMD: Catalyst 12.9 Beta
Benchmarks
Battlefield 3
Campaign Mode, "Operation Swordfish" 60-second Fraps
Crysis 2
DirectX 11, 60 second Fraps
Batman: Arkham City
Version 1.0.0.0, Built-in DirectX 11 Benchmark
Metro 2033Full Game, Built-In Benchmark, "Frontline" Scene
DiRT Showdown
Version 1.0.0.0, DirectX 11 Benchmark
The Elder Scrolls V: Skyrim
Update 1.4.27, THG Benchmarks save, 25-second Fraps
Max Payne 3
Version 1.0.0.4.7, Chapter 3, save 16, 65-second Fraps
World Of Warcraft
DX 11, Elwynn Forest area, 30-second Fraps
5. Benchmark Results: Batman: Arkham City

The low minimum frame rates we see in Batman: Arkham City are a result of drops in between benchmark scenes; they don't reflect real-world gameplay, as you can see from the frame rate-over-time chart. We run our first test without MSAA, but with FXAA enabled to smooth out some of the jagged lines and edges.

The GeForce GTX 650 Ti does very well, beating out all of its competition except for the Radeon HD 7850 and GeForce GTX 660 Ti.

Switching out FXAA for MSAA applies more of a load to the memory subsystem, allowing AMD's Radeon HD 6870 to pass the GeForce GTX 650 Ti. The rest of the field trails Nvidia's new value-oriented board, though the GeForce GTX 560 and Radeon HD 6850 aren't far behind.

6. Benchmark Results: Battlefield 3

The single-player campaign of Battlefield 3 is decidedly more GPU-limited than the multi-player component. So, we'll start with the High detail preset with no MSAA applies, and the Post AA filter set to High.

The GeForce GTX 650 Ti competes with AMD's Radeon HD 6870 and Nvidia's own GeForce GTX 560. but the Radeon HD 7850 1 GB and GeForce GTX 660 lead the pack.

In order to achieve playable performance with MSAA active, we had to lower the quality preset to Medium and disable the FXAA-based filter, motion blur, and ambient occlusion options. This specific combination of settings caused Nvidia's cards to demonstrate intermittent black textures, which we'd be most inclined to blame on the company's beta graphics driver.

Aside from the observed quality issues, the GeForce GTX 650 Ti rides the boundary of playability with a minimum 29 FPS frame rate. The 192-bit GeForce GTX 460 generates a similar average frame rate, but encounters strange performance dips that slow performance at inopportune times.

All of the Radeon cards trail behind, except for the Radeon HD 7850 1 GB, which finishes second only to Nvidia's GeForce GTX 660.

7. Benchmark Results: Crysis 2

We're running through today's Crysis 2 benchmark manually, as we're starting to discover that the automated tool we were using previously returns much higher performance numbers.

With DirectX 11 enabled, the GeForce GTX 650 Ti handles the High detail preset with a 31 frame-per-second minimum and 51 FPS average, which is slightly faster than AMD's Radeon HD 6850. The Radeon HD 6870, 7850, and GeForce GTX 560/660 beat Nvidia's new GK106-based board, though.

Crysis 2 doesn't benefit from edge anti-aliasing unless the Very High detail preset is used. And, in order for us to get playable performance on the GeForce GTX 650 Ti, we have to revert to Crytek's DirectX 9 code path. At this setting, overall performance is similar to the previous benchmark run.

8. Benchmark Results: DiRT Showdown

We disable the global illumination option in DiRT Showdown to get smooth performance from our mid-range cards. We also recently changed our benchmarking sequence in this title to achieve more consistent performance.

The GeForce GTX 650 performs almost identically as the GeForce GTX 560, though AMD's Radeon HD 7770 keeps up to both of them.

We know that this title, part of AMD's Gaming Evolved program, favors that company's cards as a result of its lighting model. Perhaps that's why the Radeon HD 7850 1 GB performs on par with the GeForce GTX 660.

Enabling 4x MSAA doesn't do much to change our finishing order. The GeForce GTX 650 Ti never drops below 40 FPS.

9. Benchmark Results: Max Payne 3

Max Payne 3 is fairly demanding at its highest detail settings. So, we're starting with the Normal quality preset and no MSAA (though FXAA is turned on).

The GeForce GTX 650 Ti gives us good performance levels, which are high enough to beat AMD's Radeon HD 6850. Nvidia's new card does lag behind the GeForce GTX 560 and Radeon HD 6870 a bit.

In order to get playable performance with 4x MSAA enabled, we had to drop from DirectX 11 to DirectX 10.1, turn off FXAA, and drop anisotropic filtering to 8x. Even in light of our adjustments, we experienced notable dips below 30 FPS on Nvidia's GeForce GTX 650 Ti. So, we'd recommend keeping MSAA turned off in this title unless you have a more powerful graphics card.

10. Benchmark Results: Metro 2033

Long one of the most demanding titles in our suite, Metro 2033 continues to bring even flagship GPUs to their knees. We're dropping the detail preset to Medium and turning on adaptive anti-aliasing in order to realize playable performance in this title.

Although Nvidia's GeForce GTX 650 Ti does well, achieving playable performance, it's beaten by the GeForce GTX 560 and Radeon HD 6870.

With 4x MSAA enabled, the GeForce GTX 650 Ti's limited 128-bit memory bus becomes a severe bottleneck, as its finishing position falls under AMD's Radeon HD 7770.

11. Benchmark Results: The Elder Scrolls V: Skyrim

We begin with Skyrim's Ultra detail preset, but turn off MSAA in favor of FXAA.

The GeForce GTX 650 Ti manages great performance, never dropping below 46 FPS. Once again, though, Nvidia's older GeForce GTX 560 and AMD's Radeon HD 6870 do slightly better than the new card.

We turn on 4x MSAA, but yet the GeForce GTX 650 Ti maintains very playable frame rates. Moreover, the finishing order we saw from the previous graph doesn't change.

12. Benchmark Results: World Of Warcraft

Historically, Blizzard's World of Warcraft favors Nvidia's GeForce boards. Our first test gauges how well the GeForce GTX 650 Ti handles the Ultra preset in DirectX 11 mode, sans anti-aliasing.

We aren't presented any surprises; the new board follows Nvidia's GeForce GTX 560 and AMD's Radeon HD 6870.

The finishing order doesn't change with 4x MSAA applied, though the frame rates naturally dip.

13. Overclocking

We pushed Gigabyte's GeForce GTX 650 as far as it'd go and achieved a maximum core overclock of 1125 MHz, with the GDDR5 memory operating at 1600. All it took was a 1.15 V GPU voltage.

In addition, we're including benchmark results for Zotac's factory-overclocked GeForce GTX 650 Ti AMP! Edition and Gigabyte's GV-N65TOC-2GI at their stock settings.

Pushed to its highest overclock, the GeForce GTX 650 Ti edges out AMD's Radeon HD 6870 and Nvidia's GeForce GTX 560 in all but two of our eight high-detail benchmarks. Zotac's AMP! card isn't far behind, making a case for its $180 price tag compared to the GeForce GTX 560.

The biggest problem we see with an overclocked GeForce GTX 650 Ti is AMD's Radeon HD 7850 1 GB, which manages to stay ahead of it in all of our benchmarks.

14. OpenCL: GPGPU Benchmarks

Unlike the proprietary CUDA and APP APIs, OpenCL is platform-agnostic, presenting developers with a common base for creating applications optimized for parallelized architectures. Specifically, GPGPU is becoming more and more popular, since programmers now no longer need to code for a specific piece of hardware, but rather for the API, with the driver handling the rest. Gaming prowess aside, not all cards are created equal in terms of GPGPU performance, and so we’ll take a closer look at how the GeForce GTX 650 Ti stacks up against the competition, both old and new.

Of course, the most interesting comparison comes from comparing Nvidia's latest to its direct competitor, AMD’s Radeon HD 7850 with 1 or 2 GB of memory. Additionally, we are also evaluating how the various factory-overclocked GeForce GTX 650 Ti models behave at their respective clock frequencies. We are purposely not including the Radeon HD 7770 and GeForce GTX 650, since the GTX 650 Ti is meant to round out Nvidia’s mid-range family, while setting itself apart from these smaller models. Let’s see whether it succeeds.

As expected, AMD’s Radeons lead the pack when it comes calculating SHA-256 hashes. If you’re into bitmining, then none of Nvidia’s current- or past-generation cards are a good fit, and that applies to the GeForce GTX 650 Ti as well, even if you factor in its lower power consumption.

As comparisons with the factory-overclocked GeForce GTX 650 Tis show, performance scales practically linearly with clock frequency.

Once again, we're using LuxMark 2.0's slightly simpler scene for testing. Otherwise, the bars representing the GeForce GTX 650 Ti would have been nearly imperceptible. We see performance scale based on clock rate once again, and AMD's Radeon HD 7850 remains far out of reach.

PostFX combines OpenGL and OpenCL into one demanding workload. Although this benchmark comes directly from Nvidia’s own SDK, it also serves to demonstrate how quickly performance can shift with the introduction of a new architecture, such as AMD’s GCN.

The NQueens problem is the sole ray of light for the Nvidia cards here, and the GeForce GTX 650 Ti is finally able to beat out AMD's Radeon HD 7850.

15. OpenCL: GPGPU Benchmarks (Basemark CL)

Fluid Operations

The performance delta between the GeForce GTX 650 Ti and the GTX 660 is very slim indeed. Stepping up the GTX 660 Ti and the 670 yields a much bigger boost. Indeed, the lower-end model wins out over its bigger sibling in this discipline thanks to a higher clock rate.

Interestingly, the factory-overclocked GeForce GTX 650 Ti shows no such improvement. Only the pricier cards with more shader processors appear to benefit from higher frequencies.

Wave Simulation

Wave simulations also aren't the GeForce GTX 650 Ti’s forte, regardless of clock rate. Actually, the extra frequency does surprisingly little to help performance.

Julia Rendering

Julia fractal rendering allows the Kepler architecture shine a bit, and all three GeForce GTX 650 Ti models are able to squeeze in between the Fermi-based GeForce GTX 580 and 570.

Mandelbulb Rendering

Whereas the Julia fractal test looked pretty bleak for Nvidia, the situation flips when it comes to Mandelbulb rendering. Within the GeForce portfolio, Nvidia’s newest release is not only able to beat the Fermi-based cards, but the factory-overclocked variants also take on the lower-clocked OEM version of the GeForce GTX 660. Meanwhile, AMD can’t come anywhere near the GeForces cards' performance

16. OpenCL: Image Processing (Basemark CL)

Image Processing

Next, we turn to image editing using a selection of filters. Since AMD’s entire family of desktop cards soundly trounces anything Nvidia has to offer, we’ll focus on the GeForce boards and their results.

When it comes to image editing using OpenCL, the GeForce GTX 650 Ti always places toward the back of the pack, posing a threat only to  older Fermi-based cards.

17. OpenCL: Video Processing (Basemark CL)

Video Processing

Although the filters used in this suite are quite similar to those employed in the previous page's image processing tests, the outcome isn’t as consistent when it comes to video processing.

Overall, AMD's cards don't perform as well (with one exception) due to their weaker video hardware. We believe there is still some room for improvement on the driver side, since we see a portion of the Radeons' theoretical performance getting lost somewhere along the way.

Among the GeForce cards, we see the same outcome we saw on the previous page, with Kepler-based cards beating those with Fermi-derived GPUs and the new mainstream boards outpacing the previous high-end. Within the Kepler family, the GeForce GTX 650 Ti places exactly where we would have expected it, if not a little lower.

The GeForce GTX 650 Ti only gets passing grades when it comes to compute-intensive workloads. Although it manages a surprise win here and there, it is obvious that its main competition, AMD’s Radeon HD 7850 1 GB, remains comfortably out of reach overall. Ironically, the GeForce GTX 650 Ti may end up plugging the rather pronounced performance gap between the Radeon HD 7770 and the HD 7850.

18. Temperature And Noise

Let's consider the GeForce GTX 650 Ti's general usability characteristics, starting with power draw:

Despite its higher TDP, the GeForce GTX 650 Ti's power consumption under load is identical to the vanilla GeForce GTX 650. This is an excellent result, and about 25 W less than the Radeon HD 6850.

The temperatures of our GeForce GTX 650 Ti samples are quite low. Nvidia's reference board tops out at 60 degrees Celsius at maximum load. Gigabyte's cooling solution, with its two heat pipes and dual fans, facilitates peak load temperatures of 44 degrees, the lowest in our field of contenders.

All of our GeForce GTX 650 Ti samples demonstrate very quiet acoustics.

With noise, power, and heat all taken into consideration, Nvidia's GeForce GTX 650 Ti should be very easy to live with compared to previous-gen GeForce and Radeon cards.

19. GeForce GTX 650 Ti: A Good Value At $150

With Nvidia's GeForce GTX 650 set at 100%, here's what the chart looks like comparing all of the cards in today's review:

We get an interesting outcome from the new GeForce GTX 650 Ti, which returns the largest gap between its high- and low-detail results. We can blame a relatively narrow 128-bit memory bus for this, coupled with a fairly potent GPU.

At fairly conservative settings, the GK106-equipped GTX 650 Ti nearly overtakes Nvidia's GeForce GTX 560 and Radeon HD 6870. But when MSAA is applied at higher quality levels, the board takes a disproportionately large hit, ending up a few percentage points faster than the Radeon HD 6850, on average. As we saw in the benchmarks, that disparity can be mitigated somewhat through overclocking. In addition, most newer titles support alternative anti-aliasing modes, such as FXAA, that are easier on bandwidth-starved board designs.

Assuming it surfaces at the $150 target Nvidia tells us to expect, the GeForce GTX 650 Ti offers the best performance at its price point, with average frame rates that outclass the Radeon HD 6850. The Radeon HD 7770 is beaten even more soundly. Conservative power, heat, and noise measurements make the card easy to live with and build an entry-level gaming system around, too.

Are there any reasons to shy away from a $150 GeForce GTX 650 Ti? Only that there's a 1 GB Radeon HD 7850 selling for $30 more. We think that gamers with the extra money would do well to invest it in the notably-faster AMD card, even though it uses about 30 W more power under load. We've seen a Pitcairn-based board selling for as little as $165 already, which could be a short-lived gimmick on AMD's part to give us a moment of pause. If it's not, then that's a real shot across Nvidia's bow.

It's impossible for us to know where prices will settle over the next few weeks. What is clear, however, is that Nvidia's GeForce GTX 650 Ti is fast enough and affordable enough to serve as a compelling option for mainstream gamers, particularly now that the Radeon HD 6850 and GeForce GTX 560 are at the end of their effective lives and disappearing from store shelves.