Top-end video cards are always interested in the most demanding players. Seriously enthusiastic enthusiasts are often willing to spend a lot of money to get the best possible performance and picture quality in their favorite projects. For NVIDIA, these are valuable customers and the company makes a lot of effort so that they do not look around when choosing the next new thing for their ultimate PCs. With the launch of the new graphic line GeForce RTX based on chips with the Turing architecture, the top position in the gaming adapter series is occupied by the GeForce RTX 2080 Ti. Let's look at what the owner of such a video card can expect.
GeForce RTX 2080 Ti
For the GeForce RTX 2080 Ti, the TU102 graphics processor is used, which has a very large (by the standards of gaming GPU) crystal area – 754 mm² and contains 18.6 billion transistors. The chips are manufactured by Taiwanese company TSMC at 12 nm standards.
|GeForce RTX 2080||GeForce GTX 1080||GeForce RTX 2080 Ti||GeForce GTX 1080 Ti|
|Crystal Name [1965901010")19659012] TU102||GP102|
|Manufacturing Process||12 nm  151616] 15 nm  15] 19] ] 16 nm|
|Crystal area||545 mm²||314 mm²||754 mm²||471 mm²|
|The number of transistors,||13.6 billion||7.2 billion .||18.6 billion.||11.8 billion.|
|. The clock frequency of the GPU||is 1515/1710 MHz||1607/1733 MHz||1350/1545 MHz||1 481/1582 MHz|
|Textural blocks||184||160||272 [19)rasterization||64||64||88||88|
|Tensor cores||368||] –||544||–|
|Memory (type, capacity)||GDDR6, 8 GB||GDDR5X, 8 GB||GDDR6, 11 GB||GDDR5X, 11 GB ||GDDR6, 11 GB||GDDR5X, 11 GB||GDDR6; Memory bus||256 bits||256 bits||352 bits||352 bits|
|Effective memory clock frequency||14 000 MHz||10 008 MHz||14 000 MHz||11,008 MHz|
|Memory bandwidth||448 GB / s||320 GB / c  616 GB / s||484 GB / s|
|Power consumption, W||215/225 W||180 W||250 W||250 W|
What consumes such a transistor budget? If we talk about the traditional graphic part, the TU102 includes 4352 CUDA-computers, 272 texture modules and 88 rasterizers. Even if we consider quantitative indicators, without taking into account architectural changes, the number of shader units and TMUs increased by 21.4% compared to what the GP102 used for GTX 1080 Ti has.
As in the case with its predecessor, for GeForce The RTX 2080 Ti uses a 352-bit memory bus. However, instead of GDDR5X chips (11,008 MHz), GDDR6 chips with an effective frequency of 14,000 MHz are used for the new product, because the bandwidth has been increased from 484.4 GB / s to 616 GB / s (+ 27.2%). As you can see, even for the top-end gaming graphics card NVIDIA decided not to use multilayer memory HBM2. The graphics processor is already very cumbersome and expensive to manufacture, and it didn’t make sense to further increase the complexity of the GPU and the cost of its production.
Formally, the GeForce RTX 2080 Ti remains within the thermal package of 250 W, however, despite identical stated values, the power consumption is sure The top model will be higher than the RTX 2080. Recall that the GeForce GTX 1080 Ti also has a TDP of 250 W, so it’s curious to evaluate the actual values.
The company developed an original design for GeForce RTX 2080 Ti. Externally, the reference version of the adapter NVIDIA is very similar to the younger model – GeForce RTX 2080. Subjectively, this is a very stylish and effective device with a metal casing and a pair of axial-type fans. Externally, the GeForce RTX 2080 Ti Founders Edition is very similar to the stock RTX 2080, we reviewed earlier. However, for the top model, another hardware platform is used with an enhanced power stabilization subsystem. Versions of the Founders Edition adapter still remain exclusive to NVIDIA, while Californian partners are developing their own devices. Often, reference boards are used for them, whereas cooling systems are developed by video card manufacturers. But for top versions of GeForce RTX 2080 Ti, companies prefer to carry out a full redesign of adapters. We will take a closer look at one of these models.
ASUS originally proposed several modifications of the GeForce 2080 Ti. The top version at the start was ASUS ROG-STRIX-RTX2080TI-O11G-GAMING.
Externally, the model is very similar to the devices of the previous generation ROG Strix series, but the already recognizable design of the external protective casing of graphite color creates this similarity, while the differences in content are more than enough.
For ASUS ROG-STRIX-RTX2080TI-O11G-GAMING, a cooling system is used with a massive 2.7-slot radiator incorporating a cassette of aluminum plates pierced with six heat pipes.
Copper is used for the heat remover nickel-base with a polished surface for better contact with the GPU. In the manufacturer's interpretation, such processing is called MaxContact.
The video card received the original printed circuit board with a powerful 19-phase (16 + 3) power subsystem, in the circuit of which the DrMOS elements are used.
For blowing a massive radiator block Three 90mm fans with a curious impeller shape are used. The outer radius of the blades are connected by a restrictive ring. According to the developers, this design increases air pressure on the radiator areas blown in, and also reduces the noise level.
At the upper edge, a second-generation NVIDIA NVLink port is provided for data synchronization when several video cards are connected. In the case of GeForce RTX 2080 Ti, the bus has a bandwidth of up to 100 GB / s – twice as much as the RTX 2080.
The video card is equipped with a pair of BIOS chips. A toggle switch is located at the upper edge of the PCB to select the active firmware version. The staff mode is Performance Mode, and Quite Mode mode suggests reducing the noise level of the cooling system with some increase in operating temperatures.
Two 8-pin connectors are used to connect additional power. The adapter is equipped with LED indicators for connecting power cables. According to the manufacturer, for a system with ASUS ROG-STRIX-RTX2080TI-O11G-GAMING, a power supply unit with a power of 650 W is needed.
The video card has an additional backlight. The light guide encircles the protective casing, a large logo glows on the protective panel on the reverse side of the PCB. If it suddenly happens that the motherboard does not have a connector for connecting an RGB tape, this is now provided on the video card.
Of course, backlight synchronization is supported in this case, which is configured using the proprietary application AURA.
The manufacturer has provided an additional button on the panel to turn off the illumination. Although, if your system is assembled in a closed case, you can extinguish the lights in the already mentioned utility much faster.
Two HDMI are displayed on the interface panel, a pair of DisplayPort and USB Type-C outputs for connecting VR headsets next generation.
Alone, the video card turns off the fans while working in the passive mode.
The radiator unit has enough capacity to keep the chip temperature at 40–41С at an open stand. 
Under load in Performance mode, the graphics processor warmed up to 62C. The GPU-Z version that was relevant at the time of the tests did not allow monitoring the rotational speed of the fans, however, in general, the noise level was below average for video cards. The GPU accelerated to 1980 MHz at peak, although typical operating frequencies were 30–40 MHz lower.
When switching the firmware version to Quiet Mode, the noise level decreases further. Fan speed is reduced. In this mode, the GPU was already warming up to 75 ° C, but its performance remained almost unchanged. The peak frequency of the GPU is fixed at 1965 MHz. Minimum reduction, which practically does not affect the number of frames / c.
An efficient cooler with a huge radiator blown by three fans, allows you to vary the mode of the video card. If you want to be quieter – you can safely use the Quiet Mode. The temperature of the processor is kept at an acceptable level and the performance does not suffer.
In express mode, we managed to force the video card further. With an increase in the base frequency from 1350 to 1500 MHz, the mean acceleration value increased from 1650 to 1800 MHz.
However, this value is uninformative, where the peak 2130 MHz is more interesting during the game load. Yes, the actual GPU frequencies on long sessions fluctuate depending on the nature of the load, but after additional acceleration it was quite possible to observe stable 2060-2080 MHz. In normal mode, the temperature of the chip under load has increased to 68C, that is, there is enough reserve for frequency experiments with GPUs.
In passing, we note that the memory chips were also able to be accelerated from the effective 14,000 to 15,800 MHz (+ 12.9%).
|Processor||Intel Core i7-8700K (3.7 / 4.7 GHz)||Intel, www.intel.ua|
|Video card||GeForce RTX 2080 FE  NVIDIA, www.nvidia.com|
|ASUS ROG-POSEIDON-GTX1080TI-P11G-GAMING; ASUS ROG STRIX-GTX1080-O8G-GAMING||ASUS, www.asus.ua|
|Motherboard||ASUS ROG MAXIMUS X Hero (Intel Z370)||ASUS, www.asus.ua|
|Operational memory||G.Skill SniperX 2x8GB DDR4-3400 (F4-3400C16D-16GSXW)||AMD, www.amd.com|
|Drive||HyperX Savage 960 GB (SHSS37A / 960G); Kingston A1000 480 GB (SA1000M8 / 480G)||HyperX, www.hyperxgaming.com|
|Power Supply||Thermaltake Toughpower Grand TPG-1200M (1200 W)||Thermaltake, www.thermaltakeusa|
|Monitor||Acer Predator X27 (27 ″, 3840 × 2160, 144 Hz, G-Sync)||Acer, www.acer.ua|
We used GeForce RTX to evaluate the performance of the NVIDIA top-end graphics 2080 Founders Edition. Note that this is a slightly accelerated version of the RTX 2080. We also did not try to get the reference GeForce RTX 2080 Ti from the ASUS adapter with factory overclocking. The base frequency initially corresponds to the set one – 1350 MHz, but to guess the algorithm of the auto overclocking system operation is problematic, because it can differ for each video card model. In general, the concept of a reference adapter in this case is rather arbitrary, because the reference model initially has an accelerated frequency formula.
For comparison, the results of the nominal predecessor GeForce GTX 1080 Ti were also used, and to get out of the flagship devices, the diagrams are presented more affordable GeForce GTX 1080.
The first synthetic test Time Spy Extreme demonstrates a noticeable advantage of the GeForce RTX 2080 Ti over the other participants. Alas, 3DMark developers are still preparing a stage with measuring performance when using ray tracing.
In real games using 4K resolution and presets with maximum graphics quality, the balance of forces of all participants strongly depends on the engine used, but in general the situation is quite predictable. GeForce RTX 2080 Ti is out of reach for other participants.
In most cases, the top model on a GPU with Turing architecture allows you to already feel really comfortable in 4K mode. The average 60–80 frames / c, even in heavy projects, is quite impressive. However, there are exceptions. In the same Deus Ex: Mankind Divided, Tom Clancy's Ghost Recon Wildlands or Final Fantasy XV, even with a GeForce RTX 2080 Ti, you will have to play with the settings to reach at least 60 frames / sec.
With a resolution of 2560×1440, performance increases noticeably . The top model for Turing certainly retains its leading position, although the odds over the RTX 2080 and GTX 1080 Ti in some games are reduced to 20–30%. It can be assumed that for such conditions, the Core i7-8700K already requires frequency acceleration.
We did not evaluate the performance in the 1920 × 1080 mode; for games with such a resolution, it is unlikely that someone will chase the top-end video card. But it is easy to assume that the difference in the performance of the test participants is even less.
For GeForce RTX 2080 Ti and GeForce GTX 1080 Ti video cards, an identical thermal package is declared – 250 W, but the actual power consumption of the system with the new flagship was higher. 
However, it should be borne in mind that at higher fps the load on the central processor also increases, and therefore the CPU contributes to the overall increase in power consumption.
GeForce RTX 2080 FE turns out to be more economical, and the best this nomination remains GeForce GTX 1080 with its TDP of 180 watts. But even in these cases, you shouldn’t disregard the influence of the CPU.
What about DXR?
GeForce RTX video cards have hidden reserves in the form of hardware blocks for working with ray tracing and tensor cores for various matrix calculations. However, game developers are just beginning to use the capabilities of DXR in their projects, and on this way they will have to get their hands on it. Unfortunately, at the time of practical testing, the GeForce RTX 2080 Ti was not yet available Battlefield V is the first full-fledged game that received support for ray tracing. However, according to numerous results available on the network, it can be seen that when DXR is activated, the performance is at least halved.
In the case of GeForce RTX 2080 Ti, you can get 40–45 frames per second at 4K screen resolution and minimally available ray tracing settings. If you further improve the quality of DXR, performance decreases even more.
NVIDIA recommends that you use the Low profile for tracing temporarily, asserting that problems with performance degradation in other modes will be solved in conjunction with the DICE developers.
It is curious that some of the DXR features will be used not only for new games, but also for projects that have long been on the market. So for Wolfenstein II: The New Colossus with the next patch received the support of Adaptive Shading technology, which allows to reduce the number of calculations over objects that are in the shadow or in a static position. Judging by the first reviews, this optimization allows you to get an additional 2-7% increase in performance. Controversial achievements, especially when it comes to a project a year ago, which can no longer be found in the popularity charts.
Obviously, NVIDIA needs more successful examples of using DXR, which would really justify the purchase of expensive graphics cards. GeForce RTX 2080 Ti is undoubtedly good in any conditions, but the difference in price with the devices of the previous generation should be compensated by impressive graphic effects. And with this just not everything is smooth.
According to the test results, it is obvious that the GeForce RTX 2080 Ti is now the most productive gaming graphics card on the market. The top model with a processor on the NVIDIA Turing architecture has a significant advantage over the flagship of the previous generation – GeForce GTX 1080 Ti, as well as the GeForce RTX 2080. The additional 30–40% cannot be compensated for by any overclocking. At the same time, the hardware of the RTX series supports ray tracing and has tensor cores for various matrix computations.
Top-end video cards have never been distinguished by the best price-performance ratio. Demanding players are always ready to pay for the maximum possible fps. However, this time the developers have set an unprecedented bar for the game segment. The recommended price for the base versions of the GeForce RTX 2080 Ti is set at $ 1,000, while the accelerated version of the Founders Edition is priced at $ 1,200 in the US market. At the same time, the original forced models are even more expensive. In Ukraine, the price of GeForce RTX 2080 Ti starts from about UAH 39,000. ($ 1400), and versions of factory overclocking and higher frequencies can cost more than 45 000 UAH. ($ 1600). This is almost twice as expensive as asking for GeForce GTX 1080 Ti video cards. There is something to think about.