Angina

Gtx 650 ti technical specifications. Test configuration, tools and testing methodology

As intended by its creators, the video card GeForce GTX The 650 Ti is a direct competitor to the AMD Radeon HD 7770. When comparing them, one cannot fail to note the progressive design of the Kepler architecture. This is especially true for the area and number of transistors involved in the graphics core. An important point is that the GK106 has 64 TMUs, while the HD 7770 has only 40. And the memory frequency of the NVIDIA model is almost 1 GHz higher. All of these factors should have a positive impact on productivity. And given the absence of limitations of GPU Boost technology, the video card is easy to overclock.

But there are also negative aspects. Graphics accelerators of a class below the GTX 650 Ti do not have the Sli connector. And as practice shows, it’s in vain! It is inexpensive solutions that attract many users to study multi-graphic configurations. Of course, for a fairly small amount you can relatively easily increase productivity, and at the same time get acquainted with all the delights of Sli. In this case, one, albeit relative, advantage remains on the side of AMD.

Specifications

Name	HD 7770	GTX 650 Ti
Codename	Cape Verde XT	GK106
Technical process, nm	28	28
Size of core/cores, mm 2	123	221
Number of transistors, million	1500	2540
Core frequency, MHz	1000	925
Number of shaders (PS), pcs.	640	768
Number of rasterization blocks (ROP), pcs.	16	16
Number of texture units (TMU), pcs.	40	64
Maximum fill rate, Gpix/s	16	14.8
Maximum texture sampling speed, Gtex/s	40	59.2
Pixel/vertex shader version	5.0 / 5.0	5.0 / 5.0
Memory type	GDDR5	GDDR5
Effective memory frequency, MHz	4500	5400
Memory capacity, MB	1024	1024
Memory bus, bit	128	128
Memory bandwidth, GB/s	72	86.4
Power consumption (2D/3D), W	3 / 80	nd/110
Crossfire/Sli	Yes	No
Recommended price, $	140	150

Test stand

Motherboard: MSI Z77A-GD65 (Intel Z77, LGA 1155);
CPU: Intel Core i7-3770K 4700 MHz (100 MHz x 47, 1.35 V);
Cooling system: water cooling system;
Thermal interface: Arctic Cooling MX-2;
RAM: Corsair DDR3 1600 MHz, 4 modules x 4 GB, (7-8-7-20-1T, 1.65 V);
Hard drive: Corsair Force GT, 128 GB;
Power supply: Enermax REVOMAX 1250 Watt;
Audio card: ASUS Xonar HDAV 1.3;
Operating system: Microsoft Windows 8 x64;
Driver version: NVIDIA – 310.xx., AMD – 12.11 beta.

List of used instrumentation and instruments

Sound level meter: Center 320;
Multimeter: Fluke 289;
Electricity tarifficator: E305EMG;
Microphone: Philips SBC ME570.

Electricity consumption level

The test involves video cards based on the reference design. Measurements are carried out in the Unigine Heaven Benchmark v2.5 program with maximum quality settings in a resolution of 2560x1440. The highest energy consumption is recorded.

Power consumption of video cards *

Power consumption of video cards * taking into account overclocking

Please enable JavaScript to see graphs

* The entire test bench, excluding the monitor.

Testing tools and methodology

In some games, where possible, built-in performance measurement tools were used:

3Dmark 2011 – 1920x1080;
Unigine Heaven Benchmark v2.5;
Total War Shogun II;
Colin McRae Dirt III;
Batman: Arkham City;
F1 2011;
Hard Reset.

For the following games, performance was measured using the FRAPS v3.5.9 utility:

Metro 2033;
Aliens vs Predator 3;
Battlefield III;
TESV Skyrim;
The Witcher 2;
Deus Ex - Human Revolution;
Sleeping Dogs;
Sniper Elite V2.

VSync was disabled during tests. To avoid errors in measurement error, all tests were performed three times. When calculating the average FPS, the arithmetic mean of the results of all playthroughs was taken as the final result.

Since video cards with a memory capacity of less than 2 GB impose a limitation on the maximum resolution, testing had to be carried out only in the most popular resolution 1920x1080. But don’t expect any concessions; the level of quality in the games remains the same, the same as that of older solutions. That is, in essence, I only propose to take measurements in Full HD and draw your own conclusions based on the results. In any case, you can always lower the graphics quality and get an adequate increase in speed.

Test results

The table and graphs use abbreviated names of video card models (for example, NVIDIA GeForce GTX 680 - GTX 680, AMD Radeon HD 7970 - HD 7970, and so on).

List of video cards and their frequencies:

Name	Rated frequencies GPU/Mem, MHz	Overclocking GPU/Mem, MHz
GTX 680	1000/1500	1250/1800
GTX 660 Ti	915/1500	1250/1650
GTX 660	980/1500	1150/1675
GTX 650 Ti	933/1350	1200/1650
GTX 650	1058/1250	1300/1600
HD 7970 GHz	1050/1500
HD 7970	925/1375	1200/1700
HD 7870	1000/1200	1200/1400
HD 7850	860/1200	1200/1400
HD 7770	1000/1125	1200/1300
HD 7750	800/1125	1150/1300

The performance of the NVIDIA GeForce GTX 650 Ti was taken as the starting point.

3DMark 2011

Settings:

Extreme profile;
DirectX 11;
Resolution: 1920x1080;
Four graphics tests;
Results: number of GPU Score points.

GPU Score

1920x1080

Please enable JavaScript to see graphs

Model	Average 1920%
GTX 680	129
GTX 660 Ti	71
GTX 660	48
GTX 650 Ti	0
GTX 650	-47
HD 7970 GHz	108
HD 7970	86
HD 7870	37
HD 7850	9
HD 7770	-20
HD 7750	-43
GTX 680 OC	165
GTX 660 Ti OC	100
GTX 660 OC	65
GTX 650 Ti OC	22
GTX 650 OC	-35
HD 7970 OC	137
HD 7870 OC	58
HD 7850 OC	41
HD 7770 OC	-8
HD 7750 OC	-37

Unigine Heaven

Benchmark version 2.5.
Settings:

DirectX 11;
Resolution 1920x1080;
Stereo 3D – off;
Shaders – high;
Tesselation – normal;
Anisotropic filtering (Anisotropy) – 16x;
Anti-aliasing – 8x.

Frames/sec

Min.|Avg.
1920x1080

Please enable JavaScript to see graphs

Model	Average 1920%
GTX 680	139.3
GTX 660 Ti	68.9
GTX 660	64.4
GTX 650 Ti	0.0
GTX 650	-47.5
HD 7970 GHz	151.1
HD 7970	130.1
HD 7870	68.0
HD 7850	47.9
HD 7770	-2.7
HD 7750	-25.6
GTX 680 OC	179.9
GTX 660 Ti OC	87.2
GTX 660 OC	80.8
GTX 650 Ti OC	23.7
GTX 650 OC	-34.2
HD 7970 OC	186.3
HD 7870 OC	89.5
HD 7850 OC	82.6
HD 7770 OC	12.3
HD 7750 OC	-16.0

Frame rate was measured using Fraps. The mission “Rock and a hard place” has been selected. After the first checkpoint we go down the hillside into the valley to the enemy fortifications. The test scene included, in addition to the descent, the beginning of a firefight during the assault on the first row of fortifications. The fire was fired at two points through a sight from behind the stones to the right and left of the road. Taking into account a simple descent, the general procedure is easily repeatable, and the final results are almost independent of random factors. Plus, the frame contains not only a large area with detailed textures, but also the lighting effects of shots and explosions. This helps to create a completely adequate picture of real-life performance both in complex, intense scenes of the single-player campaign and in multiplayer battles. All detail and quality parameters are on Ultra under DirectX 11, MSAA 2x anti-aliasing mode is enabled. A video of our test is available.

Testing was carried out in the village at the first Southern Shelf location. They ran from the village gate down the road to the bandits' checkpoint. Three split grenades were thrown there, two clips were shot from an acid pistol and three from a shotgun. Maximum graphics quality, anti-aliasing enabled. Medium quality PhysX effects activated. Thanks to explosions and shooting, all these effects are manifested in to the fullest. You can see how the test was carried out in the game.

Testing was carried out on the second Celerium mission immediately after the flight at the moment the assault on the enemy base began. This detailed level with atmospheric effects and bright spotlights has been specially selected to be one of the most beautiful in the game. There was a short run with shooting. A video of the test scene is available. Four reps. Maximum graphics settings for DirectX 11, anti-aliasing MSAA 4x.

Two five-time runs of the Ambush demo from the Crysis Warhead Benchmarking Tool. Maximum graphics settings (Enthusiast) under DirectX 10 with AA 2x anti-aliasing at a resolution of 1920x1080.

Testing was carried out in “manual” mode. We took a walk through the territory of the first village and the surrounding thickets. The abundance of garlands and other lighting sources creates a significant load on the video card, which is why this place was chosen for testing. A video of the test walk is available. Three runs. Maximum quality settings for DirectX 11 with SSAO active, anti-aliasing disabled. The viewing angle is set to the default value - 73.15°.

Hitman: Absolution

Three runs of a standard gaming benchmark test. All graphics settings in DirectX 11 are at maximum quality, only SSAO is set to “normal quality”. MSAA and FXAA anti-aliasing were not activated.

Medal of Honor: Warfighter

Testing was carried out in the first introductory mission, full of explosions. After the second checkpoint, when it was necessary to blow up the car, we made a short run around the territory, then the explosion was carried out and we followed along the falling and burning structures to a closed container. A video recording of the test moment can be viewed. Four runs, all graphics settings at maximum, AA 2x anti-aliasing.

Three runs of the built-in performance test. Maximum quality settings. The Anti-Aliasing Level setting, which controls FXAA anti-aliasing quality and shading quality, is set to High. Of course, the main impact on performance is not FXAA, but “shading”, which in this case implies the use of a more advanced global illumination system.

XCOM: Enemy Unknown

Maximum graphics quality settings. Launched new game and performance was measured during the introductory scene and during the first few seconds of the game level demo. Testing time is limited to 52 seconds, the video recording can be viewed. To reduce the error, two repetitions were performed.

3DMak 11

Testing in Extreme mode at 1920x1080 resolution.

Energy consumption

For a comprehensive assessment of energy consumption, measurements were taken in several games:

Hitman: Absolution;

The measurements were made with a Cost Control 3000 device. The highest readings during the tests were taken into account, based on which the average value was calculated. Test results

Assassin's Creed 3

In the new Assassin's Creed at nominal value, the GeForce GTX 650 Ti shows identical results to the GeForce GTX 560, being inferior to the Radeon HD 7850 by only 5%. The gap between the Radeon HD 6850 and the Radeon HD 7770 is very significant, although the latter compensates for it with overclocking. With increasing frequencies, the difference between GeForce GTX 650 Ti and GeForce GTX 560 are small, but some advantage still remains with the old video adapter.

Minor difference between GeForce GTX 650 Ti and GeForce GTX 560. Only when overclocking does a representative of the previous generation achieve an advantage over a newcomer. At nominal value, the Radeon HD 7770 is 18-22% weaker than ASUS, and when overclocked, the difference between them decreases to 10-12%. Overclocking increases the performance of the video adapter in question by 14-16%, which is noticeably lower than the increase in GPU frequency - due to the low overclocking of GDDR5 memory.

He gave us a surprise. If with high quality PhysX effects, NVIDIA solutions in this game have a noticeable advantage, then with average quality PhysX even one processor copes quite well, giving the Radeon HD 7850 a good advantage over all other participants. But this processor is an overclocked Core i7-3930K. Weak CPUs will no longer be able to provide comfortable frame rates without a GeForce video adapter. Among NVIDIA representatives, the GeForce GTX 560 demonstrates the highest fps, but it also manages to overtake the Radeon HD 7850 only when overclocked. But the GeForce GTX 650 Ti performed poorly in this test, demonstrating results approximately on par with the GeForce GTX 460 and Radeon HD 6850.

The ASUS video adapter in Black Ops 2 confidently outperforms the Radeon HD 7770 and GeForce GTX 460. Compared to the Radeon HD 6850, the GeForce GTX 650 Ti has an advantage in average fps of 6%, but the minimum is 3% worse. When comparing overclocked video adapters, we have a minimal advantage over the accelerated GeForce GTX 460, a gap of 3/17% (minimum/average fps) from the Radeon HD 7770, and again an ambiguous situation in comparison with the Radeon HD 6850. Overclocking itself gives an increase of 15-17%.

Confident victory over GeForce GTX 460 and Radeon HD 7770 in Warhead. The Radeon HD 6850 is worse than the GeForce GTX 650 Ti in terms of the minimum nominal value, but when overclocked it achieves a full advantage.

A minimal advantage over the GeForce GTX 560 in nominal terms and equal results when overclocked. Radeon HD 7770, Radeon HD 6850 and Radeon HD 460 are weaker in both modes.

Hitman: Absolution

The GeForce GTX 650 Ti is hot on the heels of the GeForce GTX 560 in Medal of Honor: Warfighter, outperforming the GeForce GTX 460 by 10-15%, the Radeon HD 7770 by 19-21%, and the Radeon HD 6850 by 27%. Overclocking gives the usual increase of 15%.

The minimum advantage of ASUS over the Radeon HD 7770 is 1-2%. This time the Radeon HD 6850 turns out to be more productive than the GeForce GTX 650 Ti. The old GeForce GTX 460 performed very poorly in this game, taking the place of an outsider. Even the GeForce GTX 560 has an advantage over the GeForce GTX 650 Ti only in terms of average fps. In this game, overclocking allows our video adapter to increase its results by 17%. But even with such an increase, it is inferior to the forced Radeon HD 7770.

Energy consumption

A pleasant surprise was the same power consumption indicators for systems with Radeon HD 7770 and GeForce GTX 650 Ti. In our previous tests, the AMD representative still turned out to be more economical. The relationship between the other participants is quite familiar. Perhaps last time we came across a more successful, economical copy of the Radeon HD 7770 or a too power-hungry GeForce GTX 650 Ti. In any case, based on the experience of this and the previous comparison, we can conclude that there is a small difference in power consumption between the Radeon HD 7770 and the GeForce GTX 650 Ti.

Conclusions

Let's start with general comments on the GeForce GTX 650 Ti. The new budget video adapter demonstrates performance at the level of middle-class representatives of the old generation. It is faster than the GeForce GTX 460, in most tests showing results close to the GeForce GTX 560. Of the 11 test applications, the GeForce GTX 650 Ti is better than the GeForce GTX 460 in eight, in two of them it has parity and in only one game is the newcomer inferior to the veteran. The GeForce GTX 560 loses to the GeForce GTX 650 Ti in only two games; in most other cases, the difference between them is a few percent, which is easily compensated by the increase in frequencies of the newcomer. When overclocked, it is no longer possible to compete with the GeForce GTX 560, but the advantage over the GeForce GTX 460 remains, despite the low performance increase in the GTX 650 Ti reviewed this time and the impressive overclocking potential of the old graphics adapter. The Radeon HD 7770 is inferior to the GeForce GTX 650 Ti in absolutely all applications. But in most cases, the AMD accelerator can compensate for this by overclocking. The situation is approximately the same when compared with the Radeon HD 6850, with some exceptions. This video adapter has one victory over its opponent in the nominal mode (Sleeping Dogs), and when overclocked in five applications it is on par with the accelerated GeForce GTX 650 Ti or even slightly faster than it.

The GeForce GTX 650 Ti combines good performance with low power consumption. We can confidently say that this video adapter is the best in terms of the combination of these characteristics among 128-bit solutions. But against the backdrop of the appearance of inexpensive Radeon HD 7850 with video memory reduced to 1 GB, the purchase of top versions of this GeForce may already be in question, because the price difference between them is only a couple of dollars. But simpler and cheaper versions do not lose their attractiveness.

The reviewed ASUS GTX650TI-1GD5 falls into the category of such popular GeForce GTX 650 Ti. At the same time, this video adapter is noticeably better in technical design than reference models - a higher-quality printed circuit board with durable components, plus a good and quiet cooling system. However, this does not provide significant advantages over competing models. In terms of noise and temperature characteristics, this ASUS corresponds to some models with a single-fan cooler (Inno3D GTX 650 Ti - Sparkle Caliber X460G video card;

Syntex - Seasonic SS-750KM power supply;

Thermalright - Thermalright Venomous X.

Since the release of the GeForce GTX 680, NVIDIA has been introducing new cards based on the Kepler architecture almost every month. And so, we barely had time to test the GTX 650, the previous new product, when the next addition arrived - the GTX 650 Ti.

Judging by the name, the GTX 650 Ti is formally positioned as a more powerful version of the “simple” GTX 650. But this picture is somewhat at odds with the real state of affairs. After all, the GTX 650, in fact, is a close relative of the GT 640, since both cards are built on the basis of the youngest GPU in the Kepler line - GK107. By the way, it is used for the production of GT 650M laptop video adapters. The GT 640 and GTX 650 differ in the type of video memory chips: GDDR3 for the first and GDDR5 for the second. As you can see from our tests of the GT 640, outdated memory greatly burdens the GK107 chip, but based on the specifications, the GPU looks quite powerful for its category. So the GTX 650 has every chance to become a worthy rival to the Radeon HD 7750, which we will check when the sample finally reaches our test laboratory. For now, let's return to the 650 Ti.

Reference GTX card The 650 Ti looks similar to the GTX 650, which we will see once again when we get to the hardware itself, but the graphics processor makes it similar to the GTX 660. After it, the GTX 650 Ti became the second model to use the GK106 GPU. In general, it would be strange to expect that NVIDIA will be satisfied by releasing a single card based on the GK106. When producing any more or less complex GPUs, you always end up with a certain percentage of crystals with partially functional computing units or simply a lower clock frequency reserve, which would be a sin not to use on a video adapter of a lower class by disabling the defective areas. So the GTX 650 Ti is equipped with chips cut down in this way.

The GK106 processor is interesting with its fun asymmetrical architecture. In the Kepler family, NVIDIA combines an array of CUDA cores and texture units into units called Streaming Multiprocessors (SMX), and the two SMXs, along with the Raster Engine geometry unit, make up a larger unit - the Graphics Processing Cluster. More details about the design and operating principle of these units are written in our review of the GeForce GTX 680. So, GK106 consists of three GPCs, two of which, as expected, contain two SMXs, and the third - only one.

In the GK106 chips used for the production of GeForce GTX 650 Ti, one of the five SMXs is disabled, depending on which part of the circuit the defect occurs in. Therefore, there are two possible options for “trimming” computational blocks. If the defective SMX is located in a "small" GPC, then that GPC will drop out completely. In another case, SMX is disabled in one of the large GPCs. In theory, the latter option provides higher geometric performance at the cost of a single "free" Raster Engine block. In practice, there is no need to talk about any kind of lottery, because the resulting performance of the GTX 650 Ti will still be limited by the fill rate. Now it will become clear why.

Four SMXs give us 768 CUDA cores and 64 texture units: twice as many as the GTX 650. At the same time, in the GTX 650 Ti, the GK106 has one memory controller disabled, a piece of the L2 cache associated with it, and several ROPs. The result is the same back-end configuration as the GTX 650: 16 ROPs plus a 128-bit memory bus.

The similarity with the GTX 650 is also enhanced by the standard amount of video memory in the reference version: 1024 MB. Today this is already a clear sign budget models. The minimum amount of memory that protects against a sharp loss of performance in heavy graphics modes can be called 1.5 GB. In the most memory-hungry benchmarks (for example, Metro 2033) at a resolution of 2560x1440 with antialiasing, it is noticeable that, for example, the GeForce GTX 580, which has 1.5 GB on board, still slips through the narrow gates without a critical drop in FPS, but The GTX 570 with its 1.28 GB is already running a slideshow. That is, depending on the amount of memory, even with a more or less fast GPU, there may be a sharp line between “you can play” and “you can’t play.” Will the GTX 650 Ti find itself in a situation where the GPU is still capable, but performance is limited by the amount of memory, we will check in benchmarks.

The GTX 650 Ti departed from the GTX 660 in clock speeds as well. The processor runs at 925 MHz, and this is a fixed value. GPU Boost technology is not used here, but there is a possibility that overclocking the card will give a noticeable increase in FPS. More powerful Kepler models, due to GPU Boost, largely select the frequency reserve in normal mode. Even with serious overclocking, the effect is sometimes negligible, which is why there is talk of “corn megahertz”. The memory clock speed of the GTX 650 Ti is also lower than the standard 6008 MHz for all higher-end cards.

In terms of power consumption (TDP 110 W), the GTX 650 Ti turned out to be not as lightweight a card as the GT 640 and GTX 650 (64-65 W). The Radeon HD 7770 is also noticeably more energy-efficient (80 W), but the HD 7850 is already more power-hungry (130 W).

	GT 640	GTX 650	GTX 650 Ti	GTX 660	GTX 660 Ti	GTX 670	GTX 680
Main Components
GPU	GK107	GK107	GK106	GK106	GK104	GK104	GK104
Number of transistors, million	1300	1300	2 540	2 540	3 540	3 540	3 540
Technical process, nm	28	28	28	28	28	28	28
GPU clock frequency, MHz: Base Clock / Boost Clock	900/ND	1058/ND	925/ND	980/1033	915/980	915/980	1006/1058
Stream processors	384	384	768	960	1344	1344	1536
Texture blocks	32	32	64	80	112	112	128
ROPs	16	16	16	24	24	32	32
Video memory: type, volume, MB	GDDR3, 2048	GDDR5, 1024/2048	GDDR5, 1024	GDDR5, 2048	GDDR5, 2048	GDDR5, 2048	GDDR5, 2048
Memory clock frequency: real (effective), MHz	891 (1782)	1250 (5000)	1350 (5400)	1502 (6008)	1502 (6008)	1502 (6008)	1502 (6008)
Memory bus width, bits	128	128	128	192	192	256	256
Interface	PCI-Express 3.0 x16	PCI-Express 3.0 x16	PCI-Express 3.0 x16	PCI-Express 3.0 x16	PCI-Express 3.0 x16	PCI-Express 3.0 x16	PCI-Express 3.0 x16
Image output
Interfaces	1 x DL DVI-I, 1 x DL DVI-D, 1 x Mini HDMI 1.4a	1 x DL DVI-I, 1 x DL DVI-D, 1 x Mini HDMI 1.4a	1 x DL DVI-I, 1 x DL DVI-D, 1 x Mini HDMI 1.4a	1 x DL DVI-I, 1 x DL DVI-D, 1 x HDMI 1.4a, 1 x DisplayPort 1.2	1 x DL DVI-I, 1 x DL DVI-D, 1 x HDMI 1.4a, 1 x DisplayPort 1.2	1 x DL DVI-I, 1 x DL DVI-D, 1 x HDMI 1.4a, 1 x DisplayPort 1.2	1 x DL DVI-I, 1 x DL DVI-D, 1 x HDMI 1.4a, 1 x DisplayPort 1.2
Max. permission		VGA: 2048x1536, DVI: 2560x1600, HDMI: 4096x2160	VGA: 2048x1536, DVI: 2560x1600, HDMI: 4096x2160		VGA: 2048x1536, DVI: 2560x1600, HDMI: 4096x2160, DP: 4096x3112	VGA: 2048x1536, DVI: 2560x1600, HDMI: 4096x2160, DP: 4096x3112	VGA: 2048x1536, DVI: 2560x1600, HDMI: 4096x2160, DP: 4096x3112
Max. power consumption, W	65	64	110	140	150	170	195
Average retail price, rub.	No data	No data	No data	No data	No data	No data	No data

The GTX 650 Ti is positioned as a card for tight-fisted gamers who have previously played at 1280x1024 and 1680x1050 resolutions. Now they have the opportunity to upgrade to 1920x1080 for little money. The recommended global retail price for the GTX 650 Ti is the same as the GTX 550 Ti - $149. In Russia, this amount is transformed into 4,999 rubles.

It's nice to see that the GTX 650 Ti from NVIDIA partners has already appeared in Moscow online stores, and at a price close to the recommended one. What else can you buy for comparable money? Among cards of the previous generation, the Radeon HD 6850 costs about the same on average, and the GeForce GTX 560 costs a little less.

More interesting is how the new product compares with the modern model range AMD. In the Radeon HD 7000 line, there is a significant gap between the HD 7770 and HD 7850, both in price and performance, and the GeForce GTX 650 Ti falls right in the middle. If the HD 7770 can be purchased for an amount of 3,500 rubles, then prices for the HD 7850 start at 6,000 rubles.

Note: there are varieties of HD 7850 with 1 GB of memory, which cost about 500 rubles. less than full-fledged 2 GB options. We did not include such a card in the set of rivals for the GTX 650 Ti for the reason that we do not have official evidence to consider it a reference model and not a private initiative of individual manufacturers. On the AMD website, for example, the characteristics of the HD 7850 indicate 2 GB of memory, and nothing else. But after the tests we will return to this topic.

⇡ Design

After partner retail options, on the example of which we tested the GeForce GTX 660 and GTX 660 Ti, we again pick up a reference sample from NVIDIA. As is usually the case with budget models, reference cards are not sent to partners, and the GTX 650 Ti will look different at retail. From the photographs of already announced models from several manufacturers, it is clear that no one is attracted by the standard design of the cooler, but the design of the printed circuit board, most likely, in most cases will coincide with the standard one.

The dimensions of the board are the same as that of the GT 640, but, of course, with more complex wiring and an additional power connector.

NVIDIA chose a design with a single-slot mounting frame, since the DisplayPort video output is routed on the board. But the height of the cooler still does not allow the card to remain within the dimensions of one slot.

The radiator of the reference cooling system resembles the low-profile coolers that are now included in many Intel processors. Overall the card looks light and modest. One of the competitors, the Radeon HD 7770, even with a significantly lower TDP in the reference version, has a dual-slot frame and a large closed-type cooler.

⇡ Fee

Despite the fact that the memory in the GTX 650 Ti operates at a lower frequency, on the board we see the same Hynix H5GQ2H24AFR-R0C chips with a standard effective frequency of 6 GHz as on the older Kepler cards. There are only four chips on the front side of the PCB, and four more contact pads are empty on the back surface. Some OEMs will likely take advantage of this to make a GTX 650 Ti with double the brains without having to come up with an original board design.

The power supply system is minimalistic: two phases for the GPU and one for memory. VDDC (CPU voltage) is controlled by the uPI Semiconductor uP1728P chip.

GeForce GTX 650 Ti | Last Kepler-based map in 2012

Over the past two months, Nvidia has released three graphics cards, all of them based on the Kepler architecture. Today we will tell you about the fourth and, according to Nvidia, the last desktop video card for 2012. So say hello GeForce GTX 650 Ti.

We know that this game is part of the AMD Gaming Evolved program, and thanks to the special lighting model, AMD cards take advantage. Perhaps that's why Radeon HD 7850 1 GB beats the GeForce GTX 660.

Activating 4x MSAA did not change the rankings much. Minimum indicator GeForce GTX 650 Ti at the same time remained above the 40 frames per second mark.

Max Payne 3

Max Payne 3 is a fairly demanding game, especially at maximum graphics settings. So we started with the Normal preset without MSAA, but with FXAA active.

The Elder Scrolls V: Skyrim

We start this test with Ultra settings, but turn off MSAA in favor of FXAA.

But no matter how we accelerate GeForce GTX 650 Ti, she couldn't outrun AMD Radeon HD 7850 1 GB in none of the tests.

GeForce GTX 650 Ti | OpenCL: GPGPU tests

Unlike CUDA and APP's native APIs, OpenCL is cross-platform neutral and provides developers with a common framework for building applications optimized for parallel architectures. GPGPU is especially quickly gaining popularity because programmers can write code not for specific hardware, but for APIs, and the driver will deal with the rest. When it comes to GPGPU performance, not all cards are created equal, so we want to look at competitiveness GeForce GTX 650 Ti in this area in more detail.

Of course, the most interesting thing is to compare new map Nvidia with its direct competitor: AMD Radeon HD 7850 with 1 or 2 GB memory. In addition, we will evaluate the behavior of different versions GeForce GTX 650 Ti with factory overclocking. We specifically do not include Radeon HD 7770 And GeForce GTX 650, because GeForce GTX 650 Ti completes the family of Nvidia mid-range video cards, separating itself from the lower models. Let's look at her progress.

As we expected, AMD Radeon cards lead the way in SHA-256 hashing. Nvidia cards are not suitable for bitmining, regardless of generation, this also applies to GeForce GTX 650 Ti, even despite low power consumption.

Performance GeForce GTX 650 Ti increases depending on the degree of factory overclocking and clock frequency.

For tests we again use a simplified LuxMark 2.0 scene. Otherwise the columns GeForce GTX 650 Ti would be indistinguishable. Performance scales with clock speed, and AMD Radeon HD 7850 leads by a huge margin.

To create an intensive workload, the PostFX test combines work with OpenCL and OpenGL. Although this test is taken from Nvidia's SDK, it demonstrates how quickly the performance landscape can change in the face of AMD's new GCN architecture.

It's interesting that GeForce GTX 650 Ti with factory overclocking it does not show such acceleration. It seems that only the more expensive general-purpose processors benefit from higher frequencies.

Wave emulation

Regardless of the clock speed, wave emulation is not the most strong point GeForce GTX 650 Ti. In fact, the extra frequency does little to improve performance.

Julia Rendering

Julia's fractal rendering allows the Kepler architecture to improve its results slightly, and all three models GeForce GTX 650 Ti were able to squeeze between the Fermi-based GeForce GTX 580 and 570.

Mandelbulb Rendering

In the Julia test, Nvidia cards performed poorly, however, in the Mandelbulb test (the 3D version of Mandelbrot) the opposite was true. The new Nvidia cards outperformed models based on the Fermi architecture, and the overclocked versions even outperformed the OEM GeForce GTX 660 with a reduced frequency. Meanwhile, AMD graphics adapters can't even come close to the performance of Nvidia cards.

GeForce GTX 650 Ti | OpenCL: Image Processing (Basemark CL)

Image processing

Next we will edit the images using a set of filters. AMD graphics cards give no chance to Nvidia graphics cards, so we focus only on the results of GeForce cards.

When editing with OpenCL, GeForce GTX 650 Ti consistently ranks last, sometimes tied with older Fermi architecture cards.

GeForce GTX 650 Ti | OpenCL: video processing (Basemark CL)

Video processing

Although we use almost the same filters as in the image processing tests, the results are noticeably different in the video tests.

In general, due to lower video resources, AMD cards performed worse (there is one exception). We believe that the drivers need to be further improved, since theoretically some of the performance of Radeon cards is lost somewhere.

The results among GeForce graphics cards are similar to those in the image processing tests. Kepler beats Fermi, and the new mid-range cards outperform the previous generation's high-end models. New within the Kepler family GeForce GTX 650 Ti takes a completely predictable place.

GeForce GTX 650 Ti is quite low. Nvidia reference card temperature at maximum load reaches 60 C. The Gigabyte cooler with two fans and heat pipes reduces this figure to 44 degrees - this is the lowest result among all participants.

All samples GeForce GTX 650 Ti exhibit very low noise levels.

If we take into account the noise level, energy consumption and heat dissipation, then GeForce GTX 650 Ti seems like a viable option, especially compared to previous generations of GeForce and Radeon graphics adapters.

GeForce GTX 650 Ti | Good solution for $150

Indicator GeForce GTX 650 taken as 100%. The comparison chart for all video cards looks like this:

Result GeForce GTX 650 Ti It turned out interesting. You can observe a significant gap between the performance at low and high detail. We think the relatively narrow 128-bit memory interface coupled with a fairly powerful GPU can be to blame for this.

At moderate settings the video card GeForce GTX 650 Ti, equipped with GK106 GPU, practically beats Nvidia GeForce GTX 560 and Radeon HD 6870. But when MSAA is enabled on high settings performance drops sharply, resulting in this model being only a few percent faster Radeon HD 6850 according to the average. In the tests we saw that the result can be improved with overclocking. In addition, most modern games support various anti-aliasing modes, such as FXAA, which is better processed on low-bandwidth cards.

Considering Nvidia's recommended price of $150, GeForce GTX 650 Ti offers best level performance in this price range, and its frame rate is higher than Radeon HD 6850 . Radeon HD 7770 it was completely destroyed. Thanks to low power consumption, temperature and noise, the card is easy to use and easier to assemble an entry-level system with.

Are there any reasons to refuse GeForce GTX 650 Ti for $150? Only if you add another $30 and buy Radeon HD 7850 1 GB. We think gamers will choose the more expensive, but also more quick map AMD, even though it consumes 30 watts more power under load. We've already seen a Pitcairn-based card for $165, which is likely a temporary move on AMD's part to make buyers think twice before purchasing. If not, then this is a serious criticism of Nvidia.

It is impossible to predict what will happen to prices in a few weeks. But it's obvious that it's new GeForce GTX 650 Ti fast and affordable enough to be a viable option for mid-range systems, especially when Radeon HD 6850 And GeForce GTX 560 gradually disappearing from store shelves.

Company NVIDIA continues to build its line of graphics adapters based on 28-nanometer GPUs with Kepler architecture. The most demanding enthusiasts, as well as thrifty gamers, already have plenty to choose from, but now Californians are throwing into the battle a solution for those who take a balanced approach to choosing a video card, without considering it necessary to spend a significant amount on “gaming” hardware. It is for such users that NVIDIA offers GeForce GTX 650 Ti. Let's see what a new product with a price tag of $150 is capable of?

GeForce GTX 650 Ti

For GeForce GTX 650 Ti GPU is used GK106, the same as in the case of the older model GeForce GTX 660. The nominal operating clock frequency of the chip is 925 MHz. In this case, GPU Boost technology is not supported, therefore the graphics core is not overclocked above a specified value during operation.

Let us recall that GK106 uses the Kepler architecture, is manufactured using a 28-nanometer process technology and contains 2.54 billion transistors. In the case of the GeForce GTX 650 Ti, the chip has some simplifications. First of all, we note a decrease in the number of SMXs modules from 5 to 4. Each of them carries 192 computational units, which means that their total number has decreased from 960 to 768. The number of texture units has also decreased proportionally - from 80 to 64, but the units rasterization is now 16, instead of 24.

For GeForce GTX 650 Ti a 128-bit memory bus is used, which for an adapter of this level can already be considered a certain limiter. However, to compensate for the possible lack of memory bandwidth (BRAM), the clock frequency of GDDR5 chips has a fairly high recommended value of 5400 MHz, which provides a memory bandwidth of 86.4 GB/s.

The GeForce GTX 650 Ti uses a PCB with the same design as the GeForce GTX 650. This means that the reference video cards are very compact. The cooling system is two-slot, although its design is very simple. Obviously, the cooler was designed taking into account the low power consumption of the video card. The declared TDP level is 110 W. The adapter requires additional power; to connect it, a six-pin connector is located on the PCB. For systems with quad-core processors and GeForce GTX 650 Ti, the manufacturer recommends using a power supply with a power of 400 W or more.

Like other members of the Kepler family, the GeForce GTX 650 Ti video card supports PhysX, Adaptive VSync, NVIDIA Surround technologies and TXAA, FXAA anti-aliasing modes, however, it does not allow SLI mode to be used. The loss is small, but adapters of this level do not often play in doubles. But, nevertheless, we note for ourselves that in the previous generation lines, lower-speed video cards also had the ability to use two accelerators.

Technical characteristics of video cards:

	GeForce GTX 660	GeForce GTX 650 Ti	GeForce GTX 650	GeForce GTX 560	GeForce GTX 550 Ti
Crystal name	GK106	GK106	GK107	GF114	GF116
Manufacturing process, nm	28	28	28	40	40
Crystal area, mm²	221	221	118	367	238
Number of transistors, billion	2,54	2,54	1,3	1,95	1,17
GPU clock frequency, MHz	980/1033	925	1058	810	900
Number of stream processors	960	768	384	336	192
Number of texture blocks	80	64	32	56	32
Number of rasterization blocks	24	16	16	32	24
Memory (type, volume), MB	GDDR5, 2048	GDDR5, 1024/2048	GDDR5, 1024	GDDR5, 1024	GDDR5, 1024
Memory bus, bit	192	128	128	256	192
Memory frequency, MHz	6008	5400	5000	4008	4100
Memory bandwidth, GB/s	144	86,4	80	128	98,4
Energy consumption, W	140	110	64	150	116
Recommended price	$229	$149	$109	$170-190*	$110-125*

ZOTAC GTX650TI 2GB AMP! Edition (ZT-61103-10M)

Already traditionally, modifications of AMP video cards! Edition from ZOTAC have an improved frequency formula. In this case, factory tuning is on a serious scale. The chip frequency is increased from 925 to 1033 MHz (+11.7%), while the memory runs at 6200 MHz (+14.8%), instead of the recommended 5400 MHz.

The video card is very compact. The cooler is covered with a black plastic cover on top. The cooling system is based on an aluminum radiator of simple design, which is blown by a 90 mm axial-type fan. The blades of the latter are made of translucent plastic and are orange in color. In combination with a black printed circuit board and a protective casing of the same color, it looks quite impressive.

The length of the printed circuit board is 145 mm, the total length of the adapter, taking into account the slightly protruding top cover, is 155 mm. The adapter barely exceeds the dimensions of the PCI-Express x16 slot, so there will be no problems installing a video card, even in a compact case.

The PCB is very similar to the reference one, but has its own characteristics. The stabilizer is made according to a three-phase circuit (2+1). Modification AMP! Edition, in addition to having increased clock speeds, carries 2 GB of memory on board. At the same time, 4 chips are located on top and to the right near the GPU, and the same number of chips is on the back side of the board.

The reference adapter is equipped with three video outputs - a pair of DVI, as well as mini-HDMI, while the video card from ZOTAC has four interface connectors on the mounting panel: two DVI, and the same number of full-length HDMI.

The package package of the video card does not differ in anything extraordinary: manual, brochure quick installation, a disk with drivers, an adapter from a pair of Molex to a six-pin connector for connecting additional power, and an adapter from DVI to VGA.

Price

Recommended price of GeForce GTX 650 Ti for the American market – $149 for the model with 1 GB of memory, reference PCB and clock speeds. For Ukraine the declared cost is 1399 UAH. Price of a boosted video card with 2 GB of memory from ZOTAC at initial stage marked at level $189 . IN the latter case The difference from the original is significant, however, in our opinion, the retail price will still be determined by the real demand in the market.

In progress

From the point of view of cooling efficiency, the cooler used has proven itself quite well. Despite its modest design, its capabilities were sufficient to remove heat from the GK106 and do so without much strain.

Note that the fan is connected using two contacts, which means that its rotation speed is regulated not using PWM, but by changing the voltage. However, it is very difficult to evaluate the algorithm of its operation. The fact is that under almost any load the fan rotated stably at 23% of maximum value, and only after the merciless Furmark warmed up the GPU to 66 degrees, the rotation speed increased to 25%. It is difficult to talk about quantitative indicators, since the actual revolutions in this case are not monitored, however, such a moderate increase in the rotation speed did not affect the overall tone of the CO operation. The cooler operates quite quietly regardless of the conditions. In general, there are no serious complaints about the cooling system, but, of course, a more flexible control algorithm would be very appropriate here. In rest mode, it would obviously be possible to get by with more modest speeds

Test results

Unfortunately, at the time of testing, the available utilities for adjusting video card parameters did not allow us to correctly control the GPU clock frequency. For this reason, the diagrams show the results of only the overclocked modification from ZOTAC in its standard mode (1033/6200 MHz). Indicators GeForce GTX 650 Ti with reference values, frequencies will be on average 10–15% lower than that of a forced video card. This should definitely be taken into account during the overall assessment of the new device based on the GK106 chip. At the same time, let us not forget that in retail sales GeForce GTX 650 Ti from many vendors will arrive in a tuned form, with increased clock frequencies, while the reference formula will be used only for the most affordable modifications.

For the above reason, we can only preliminary evaluate the performance of the GeForce GTX 650 Ti. However, even the data obtained still gives an idea of the potential of the new solution. So, after additional overclocking, the GeForce GTX 650 Ti manages to compete almost equally with the GeForce GTX 560 and Radeon HD 6870. Even if we omit the very optimistic results in synthetic 3DMark 11 and Lost Planet 2, the new product on GK106 shows very decent results on average. It is noticeably ahead of the Radeon HD 7770, however, given the latter’s price at $120, this is not surprising.

It is also interesting to compare the capabilities of the GTX 650 Ti and its formal predecessor, the GTX 550 Ti. The predecessor is not only inferior to the new GK106. If there is not a gap between them, then there is a very noticeable difference. Owners of GF116 video cards who want to change their video card should look towards faster solutions in order to fully experience the contrast and get satisfaction from the upgrade. At the same time, the obtained figures are evidence that technology does not stand still.

If we compare GeForce GTX 650 Ti video cards with the younger model without the “Ti” index, it is obvious that these are adapters different levels. Twice more computing processors and texture units allows the new product to break away from the younger model in terms of performance. Moreover, the difference is so great that one might wonder whether the device on the GK107 chip as a whole should have been classified as part of the GTX family? However, the difference in the cost of adapters is also significant, so parity in terms of price/performance is generally maintained; it is only important to remember that in the case of the GTX 650, the “Ti” index has significant weight.

Results

GeForce GTX 650 Ti– a compact, cost-effective solution that can provide acceptable speeds in Full HD mode in modes with medium or even high graphics quality in any modern games. The performance of the forced video card is very decent compared to the offers on the market in the same price category. With a retail price of ~$150–160, the new product will certainly be very popular.

Taking a short historical excursion, we note that the GeForce GTX 550 Ti at the time of its announcement (March 2011) was also offered for $150. A year and a half later, for the same money, we get a device with a significantly increased level of performance, which has also become noticeably more economical. Of course, the period of “console stagnation” affects the pace of development of graphics adapters for PCs, but the slogan “more functional, faster, more energy efficient” is still relevant.

Introducing GeForce GTX 650 Ti NVIDIA has occupied an operational space in which its main competitor currently still does not have alternative proposals, based on modern GPUs. AMD will certainly try to respond to such an attack. It is likely that partners who are ending sales of video cards based on Barts chips (HD 6850/6870) and cannot now offer interesting options in the most popular price category will help her take this step. One of the stages will be to reduce the cost of the Radeon HD 7850, which in the version with 1 GB of memory will possibly fall in price to $169. However, this is a higher-level solution, while for direct competition with the GeForce GTX 650 Ti the AMD line lacks more available model on the Pitcairn chip. Is it time for the announcement of the Radeon HD 7830 and a new battle between irreconcilable competitors?