Review of the GeForce RTX 2080 Founders Edition: an artist sees this

By the time the next-generation discrete graphics are launched, NVIDIA often prepares its own reference video card models. Developed by Californians, devices later become the basis for adapters that fall on the shelves under well-known brands. For the presentation of new products with Turing architecture, the company also developed original versions of video cards. Today we have a NVIDIA GeForce RTX 2080 Founders Edition model on review.

NVIDIA GeForce RTX 2080 Founders Edition

The presented version of GeForce RTX 2080 Founders Edition is offered by the developer as a reference model. The original device was originally forced by the manufacturer.

The TU104 graphics processor with Turing architecture used for the GeForce RTX 2080 operates at sufficiently high frequencies. The baseline is set at 1515 MHz, but the value of dynamic acceleration can vary. In particular, for starter versions of adapters, this is 1715 MHz, whereas for Founders Edition modifications, this is 1800 MHz. Recall that this is a certain average value, which is significantly exceeded during a typical game load.

GDDR6 standard memory chips operate at an effective 14,000 MHz. The GeForce RTX 2080 uses a 256-bit bus, because the total bandwidth is 448 GB / s. For comparison, the GTX 1080 has 320 GB / s, while the top GTX 1080 Ti with a 352-bit bus has 484 GB.

For the reference video cards of the GeForce RTX line, the developers radically changed the design of the cooling system. For more than ten years, air coolers with one centrifugal compact fan have been used for basic NVIDIA devices. Over time, the shape of the radiator blocks, the size of the evaporation chambers and the design of the protective decorative casing changed, but the principle of operation remained the same. For GeForce RTX, engineers have prepared a new design of CO, which even outwardly no longer causes any associations with devices of previous generations.

The design of the cooling system still uses an evaporation chamber that is attached to a very massive radiator unit. Passive cooler, consisting of a large base and a whole set of plates, covers almost the entire area of ​​the printed circuit board.

The dimensions of the aluminum plates are small. Despite the obvious desire to increase the area of ​​dispersion, it was important for developers to remain within the framework of a two-slot design.

Two 90-mm axial-type fans are used to blow the radiator block. Each has 13 fairly large blades. This layout allows you to effectively cool all the elements of the video card. Coolers with centrifugal fans allow the heated air to be thrown out of the system unit, but in order to cope with powerful graphics chips, it is necessary to increase the fan speeds, increasing the level of noise produced.

The “vocal” video cards do not please many people and especially frustrate those who do not use headset for games. The adapter version with a pair of axial fans and a large radiator should improve cooling efficiency and reduce CO noise compared to previous generation devices.

The radiator cooler unit covers the aluminum decorative plate from above. The reverse side of the adapter is also covered with a metal plate. Both parts fit perfectly, so it seems like an inseparable contour. Externally, the video card resembles a bar with rounded corners.

The plate on the reverse side of the PCB is used not only to strengthen the structure, but also as a cooling element. Heat-conducting stickers are stuck in the VRM and memory chips area.
NVIDIA thoroughly assembled the adapter using dozens of mounting screws of various types and sizes. Of course, this will not stop the most inquisitive enthusiasts before disassembling the device. But for most users, this is not necessary.

In order not to disturb the factory thermal interface, we used GeForce RTX 2080 Founders Edition stock images to evaluate the layout of the board.

The reference printed circuit board will be used for a number of third-party RTX 2080 models . For a video card, a 10-phase (8 + 2) VRM scheme with DrMOS power elements and the controlling uP9512P controller is used. Micron GDDR6 memory chips are placed in a U-shaped way around the GPU. There is a lot of free space on the PCB. We draw attention to the unusually large crystal TU104-400A. Despite the 12-nanometer manufacturing process, the area of ​​the silicon wafer is 545 mm². Recall that the area of ​​the GP104 chip used for the GeForce GTX 1080 is 314 mm². Even the top GP102, on which GTX 1080 Ti is based, is more compact – 471 mm². Obviously, the cost of manufacturing the GeForce RTX has increased significantly, but NVIDIA consciously took this step, relying on the expansion of the functionality of the game graphics.

When developing the new family of video cards, engineers paid special attention to the power supply adapter. GeForce RTX 2080 Founders Edition dynamically adjusts the number of active phases of the GPU power supply depending on the current load on the GPU. In this way, it is possible to obtain the highest possible efficiency regardless of the nature of the load and the current operating currents.

To connect additional power to the video card, there are two connectors: 6 and 8 pins. In this configuration, the adapter can be brought up to 300 W, taking into account the capabilities of the PCI Express x16 slot.

Recall that for the standard versions of the GeForce RTX 2080 TDP is declared at the level of 215 W, while for the Founders Edition accelerated modification, the thermal packet is 10 W more – 225 watts. For systems with a GeForce RTX 2080 video card, the developer recommends using power supplies with a capacity of 650 W.

As we have already mentioned, the GeForce RTX 2080 Founders Edition has a dual-slot layout. The total length of the video card is 266.7 mm, width – 115.7 mm. These are quite compact dimensions for video cards of this class. Many manufacturers offer original adapters of the GeForce RTX series in a 2.5-3 slot design, as well as devices with extended lengths. In this case, the problems are practically excluded even when installing the video card in a relatively compact package.

Despite its modest size, the GeForce RTX 2080 Founders Edition weighs almost 1.3 kg. Yes, there is a lot of metal here. Aluminum vestments and a large radiator unit with an evaporation chamber are making themselves felt.

As for the exterior lighting decor, NVIDIA is traditionally kept to this end. No riot of colors with RGB illumination. Only on the side panel the large “GeForce RTX” is elegantly highlighted in green. If necessary, you can turn off the backlight, change the brightness or activate the pulsation mode.

GeForce RTX video cards can be combined into SLI bundles. In the case of older RTX 2080/2080 Ti models, the high-speed NVIDIA NVLink interface of the second generation is used to synchronize data between the adapters.

For the GeForce RTX 2080, one port with a bandwidth of up to 50 GB / s is used, while for the flagship it has two ports twice the data transfer rate.

The connector for the connection is located on the top edge of the video card, next to the mounting bracket. In the case of GeForce RTX 2080 Founders Edition, the port is elegantly hidden by a stub that follows the cooling system circuit.

The high NVLink data transfer speeds allow the GPU to use access to the local memory of neighboring adapters, as well as to organize various modes of interaction between video cards when building a scene, not limited to AFR .

For the GeForce RTX 2080/2080 Ti connection, a special adapter is needed – the GeForce RTX NVLink Bridge. Decorated in the style of reference adapters bridge from NVIDIA will cost $ 79. A considerable amount for the connector, but this is unlikely to be a problem for the owner of a pair of top adapters. Video card manufacturers are already offering their adapters at a similar price.
The video card's interface panel contains three DisplayPort 1.4a video outputs and one full-format HDMI 2.0b. DisplayPort ports allow you to connect screens with resolutions up to 8K at 60 Hz. HDMI can cope with 4K with the same refresh rate.

The digital DVI interface is not suitable for high-resolution monitors and refresh rates, because NVIDIA developers thought that this port no longer has a place on the GeForce RTX panel. However, the Founders Edition video card comes with an adapter with DisplayPort, which still allows you to use DVI to connect the screen (up to 2560 × 1600 @ 60 Hz).
The bracket also has a VirtualLink port, presented in USB Type-C format. This is a promising channel that allows you to connect VR-helmets with a single cable. In addition to the image, the port can also transmit power up to 35 watts. Obviously, in this case, the power consumption of the video card will increase.

Part of the panel is occupied by the grille to output heated air, although you should not count on it. The design with axial fans in principle does not imply directional exhaust, in this case, the air outlet through the interface panel is also not conducive to the orientation of the plates of the radiator unit.

For a system with GeForce RTX 2080 Founders Edition, it is desirable to use a blown body with forced air out of the system block.

The work

For the reference model GeForce RTX 2080 Founders Edition, there is no passive mode of operation of the cooling system. Obviously, the manufacturer fears that in compact cases the radiator with a relatively small dissipation area may not cope with heat dissipation, especially in configurations with a pair of video cards.

In idle mode, the GPU temperature was below 30C. Amazing figure, but to get such values ​​you need to blow the radiator unit. And without a load, the fans rotate at a speed of about 1500 rev / min. Yes, high-quality models with three-phase motors and good impeller configuration work very quietly for such values. For a system with a closed case, this will definitely not be a problem, but still the video card is obviously not silent.

In our opinion, the rotation speed of the fans could be safely reduced to 1000–1200 rpm. Thus, having further reduced the emitted noise, without losing much of the cooling efficiency.
Like many colleagues in the workshop, during the first review of the GeForce RTX 2080, we noted that the new generation of adapters in idle mode consume a few watts more power than predecessors. The developers immediately responded to such comments, promising to promptly rectify the situation.

Some changes are already visible in driver version 411.70. At minimum load, the working frequency of the memory chips is reduced from 200 MHz to 100 MHz, and the voltage on the GPU instead of 0.725 V stable in this mode is now periodically reduced to 0.7185 V.
Under the game load, the graphics processor warmed up to 72 degrees. At the same time, the rotation speed of the fans increased to 1,800 rpm. Overall great results. The noise produced by the video card increased, but the adapter worked incomparably quieter than reference models of previous generations with a single centrifugal fan. Here, the developers managed to kill two birds with one stone – to make the video card quieter and improve the cooling of the GPU. The proposed cooler copes with the GeForce RTX 2080 without any problems, leaving room for additional overclocking of the video card.

Under load, the graphics processor accelerated up to 1980 MHz. However, these are single bursts. In games, the clock frequency often fluctuated between 1920–1935 MHz. Curiously, these are slightly higher rates than those of the previously considered ASUS GeForce RTX 2080. The forced version of the Taiwanese manufacturer had an identical frequency formula with an average acceleration value of up to 1800 MHz, while the actual frequency under load was at the level of 1900 MHz.

Overclocking

Of course, it was interesting to check the overclocking potential of GeForce RTX 2080 Founders Edition. Recall that the developers have proposed the NVIDIA Scanner tool for new video cards, which allows one-click the mouse to determine the frequency potential and automatically overclock the video card. The process of diagnostics and self-testing takes about 15 minutes, after which the frequency values ​​are established at which the stability of the graphics processor is preserved. For evaluation, we used the EVGA Precision X1 utility.

This way of accelerating the video card is suitable for owners who do not have the time / desire or basic skills of self-acceleration. Nevertheless, the resulting increase in GPU frequencies is relatively small, moreover, this method does not imply the acceleration of memory chips.
In manual mode, we managed to achieve noticeably better results. Initially, increasing the Power Target parameter to 124%, the base operating frequency of the chip was shifted from 1515 MHz to 1655 MHz (+ 9.2%). The average acceleration value increased from 1,800 MHz to 1,940 MHz (+ 7.7%). This is all without increasing the supply voltage.

The memory chips also managed to accelerate noticeably, raising the resulting effective frequency from 14 000 to 15 800 MHz (+ 12.9%).
After additional overclocking, the temperature of the video card naturally changed. In games, the GPU was already warming up to 77C, and in order to maintain such a value, the rotational speed of the fans automatically increased to almost 2100 r / min. At an open stand, additional increment of revolutions could be seen from the increased noise level, but everything is within reason. Even in this state, the GeForce RTX Founders Edition worked noticeably quieter than the reference GTX 1080/1080 Ti. Note that the new cooling system copes with overclocking without much difficulty.

The obtained maximum 2115 MHz looks optimistic, but these are still single peak pulses. Often under load, the graphics processor ran at 2025–2075 MHz. The TU104 chip confidently passes the 2 GHz mark, although in general the frequency boundaries are very similar to those of the previous generation top GPUs.

Testbed Configuration

Processor Intel Core i7-8700K (3.7 / 4, 7 GHz) Intel, www.intel.ua
Video card 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
Storage 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.com
Monitor Acer Predator X27 (27 ″, 3840 × 2160, 144 Hz, G-Sync) Acer, www.acer.ua

Performance

We already had the opportunity to evaluate the performance of GeForce RTX 2080 in current games with the first acquaintance with a video card based on TU104. In most cases, an adapter based on a Turing graphics architecture is minimally ahead of the flagship line of the previous generation – GeForce GTX 1080 Ti, and turns out to be 20–40% more productive than the nominal predecessor GeForce GTX 1080.

The advantage of the new product is more obvious in 4K mode, while with a lower resolution, the CPU becomes the limiter, even when it comes to the Core i7-8700K.

As practice has shown, the GeForce RTX 2080 Founders Edition can sometimes have a minimal advantage over wounds due to GPU Boost its reviewed ASUS DUAL-RTX2080-O8G video card. Although the overall performance of both versions is very close.

In current projects, the flagship based on Pascal also looks very convincing, only in a number of games the RTX 2080 yields about 5-10%.

The diagrams also show the results of GeForce RTX 2080 Founders Edition after acceleration. Additional acceleration of the GPU and memory chips brings a very good performance boost in 4K mode. Perhaps, sometimes it is even higher than could be expected after a moderate forcing of a video card without increasing the supply voltages. In some games, the average number of frames / s increased by more than 10%, and minimum values ​​were tightened. This is a question about the hidden reserves of new items that can be obtained without much effort.

Of course, the GeForce GTX 1080 Ti can also be further overclocked and get a similar performance boost. Overclocking does not change the balance of power. Not afraid to repeat, we note that both video cards are very close in their capabilities in the available game projects.

When the graphic resolution is reduced, the difference becomes even less noticeable. In the 1440p mode, the GP102-based graphics card literally shadows the RTX 2080. Both video cards are perfect for those who have a monitor with a resolution of 2560 × 1440 and an update rate above 60 Hz. Although, in a number of projects, even at 1440p, setting the maximum graphics quality is undesirable due to the very high resource consumption.

However, the reserves of the GeForce GTX 1080 Ti are limited by possible overclocking, whereas the GeForce RTX 2080 potentially has a good opportunity to accelerate in other ways. Recall that the new RTX series adapters are equipped with tensor blocks for matrix computing and hardware RT-cores for processing the ray tracing. Раскрывать возможности этой стороны новинок на базе Turing, увы, пока приходится с помощью технологических демонстраций, но результаты интригуют.

Одна из доступных демо Infiltrator на базе движка Unreal Engine 4 уже сейчас позволяет оценить некоторые преимущества использования DLSS (Deep Learning Super Sampling). После предварительного «обучения» с подготовкой сцен, производительность видеокарт GeForce RTX значительно увеличивается.

Если при использовании сглаживания TAA адаптеры RTX 2080 и GTX 1080 Ti имеют близкие средние показатели кадров/c, то с DLSS видеокарта c чипом TU104 значительно прибавляет в производительности (+43%), тогда как адаптеру на Pascal такой режим сглаживания не доступен.

В прошлый раз мы наблюдали подобное магическое ускорение в оптимизированной демке Final Fantasy XV, хотя прирост при использовании DLSS там был несколько скромнее – 30%. Интересно было бы увидеть подобное отличие в реальных играх.

С нетерпением ожидаем и возможности на «живых» примерах оценить преимущества эффектов при использовании трассировки лучей. Уже до дыр затерта подготовленная демка Star Wars Reflections, очень эффектно выглядят Project Sol и показанный во время запуска GeForce RTX ролик из готовящейся Metro Exodus.

На подходе целый ряд интересных проектов с поддержкой Ray Tracing, но к моменту запуска GeForce RTX не оказалось уже готовых игр, с помощью которых NVIDIA можно было убедить в преимуществах графики нового поколения даже самых заядлых скептиков.

Буквально на днях компания Microsoft предложила крупное обновление Windows 10 October 2018 Update (1809), в котором среди прочих функций включ на поддержка DirectX Raytracing (DXR) для работы с трассировкой лучей. Возможно официальный релиз нового API позволит разработчикам активнее использовать новые возможности в своих проектах. Очевидно, что унификация упростит процесс для всех желающих применять гибридный метод построения картинки.

HDR и G-Sync

Во время тестирования новой видеокарты мы также проверили ее работу в режиме с расширенным динамическим диапазоном (HDR, High Dynamic Range). Для такого эксперимента использовался один из наиболее скоростных и функциональных мониторов на рынке – Acer Predator X27.
27-дюймовая модель имеет разрешение 3840×2160 с частотой обновления до 144 Гц, а также поддержкой G-Sync и HDR. Обе последние функции в теории не должны сказываться на производительности видеокарт, но тоновая коррекция все же требует определенных ресурсов.
Для оценки мы использовали несколько игр, нативно поддерживающих HDR, а также активировали технологию адаптивной синхронизации G-Sync.

Как видно по результатам, для видеокарты GeForce RTX 2080 Founders Edition использование HDR и G-Sync никаким образом не влияет на производительность, а вот видеокарта GeForce GTX 1080 Ti может в подобном режиме потерять 5–10% от общего количества кадров/с. Да, это все частные случаи. Игровые мониторы с HDR и G-Sync все слабо распространены и остаются дорогим удовольствием. Но, увлеченным энтузиастам, которые планируют обзавестись топовой видеокартой и ультимативным экраном, стоит взять на заметку.

Энергопотребление

Программным способом разработчикам удалось снизить энергопотребление GeForce RTX 2080 в режиме покоя, потому при отсутствии нагрузки система с новой видеокартой имеет такие же показатели, как и у ПК с адаптерами предыдущего поколения.
Как мы уже выяснили при первом знакомстве с RTX 2080, в штатном режиме новая видеокарта под нагрузкой потребляет больше энергии, чем GeForce GTX 1080, но оказывается на 20–25 Вт экономичнее GeForce GTX 1080 Ti. Модификация Founders Edition показала примерно такие же результаты, как и ранее протестированная версия от ASUS.

После дополнительного разгона видеокарты с повышением параметра Power Limit до 124%, энергопотребление системы с форсированной RTX 2080 увеличилось примерно на 30 Вт. Часть этих затрат можно списать на центральный процессор, у которого также добавилось работы из-за повышения количества кадров/c. Но в любом случае стоит учитывать, что для разгона видеокарты на TU104 понадобится эффективная система охлаждения. От фактической температуры также зависит прирост рабочей частоты GPU.

Цена

О высоких ценах на новые видеокарты не высказывался только ленивый. Действительно, в сравнении с предшественниками рекомендованная стоимость устройств поколения Turing заметно выросла.

Результаты тестов в текущих играх в полной мере не отражают всех возможностей новых видеокарт, а потому увеличившиеся ценники непросто обосновывать потенциальным покупателям. Последних зачастую мало волнуют возросшая себестоимость изготовления GPU и потенциальные возможности, которые пока можно ощутить лишь в технодемках.

Референсные модели доступны только на сайте производителя, потому вряд ли они появятся в свободной продаже на розничном рынке Украины. Желающие заполучить именно Founders Edition наверняка найдут способ, ну, а для тех, кого в принципе интересуют возможности GeForce RTX 2080, уже доступны разноплановые модели от партнеров.

от 27 699 грн

Предложений: 37

от 33 185 грн

Предложений: 5

от 27 599 грн

Предложений: 17

от 27 999 грн

Предложений: 4

от 26 216 грн

Предложений : 24

Стоимость устройств из первых поставок ожидаемо выше даже дерзких предположений – $1000+, но цены наверняка будут снижаться. Хотя бы до рекомендованных значений. Напомним, что для GeForce RTX 2080 Founders Edition это 26 990 грн (~$960), а стоимость базовых версий должна быть еще ниже.

Плюсы: Высокая производительность; поддержка технологий RTX; двухслотовая конструкция; заводской разгон GPU; эффективная система охлаждения; уровень шума под нагрузкой; хорошая энергоэффективность; стильный внешний дизайн; качество сборки и материалов

Минусы: Высокая цена; не предусмотрен пассивный режим работы кулера при минимальной нагрузке

Вывод: GeForce RTX 2080 Founders Edition – оригинальная видеокарта с GPU на базе графической архитектуры NVIDIA Turing. Версии адаптеров от разработчиков традиционно выглядит очень эффектно, а модели линейки GeForce RTX получили еще и эффективные кулеры, отлично охлаждающие все элементы видеокарты. Если говорить о производительности, то средняя модель нового семейства оказалась мощнее игрового флагмана предыдущего поколения, даже когда речь идет о классическом растровом рендеринге. Однако для GeForce RTX это лишь часть возможностей. Свой потенциал эти видеокарты будут раскрывать по мере использования в играх перспективных трассировки лучей, эффектов на основе алгоритмов глубокого обучения и целого ряда других графических технологий.

Технические характеристики

Обзор видеокарты GeForce RTX 2080 Founders Edition: художник так видит" style="float:left; margin-right:15px; NVIDIA GeForce RTX 2080 Founders Edition
Уведомить о появлении в продаже
Тип Видеокарта
GPU GeForce RTX 2080
Объем памяти, ГБ 8
Тип памяти GDDR6
Интерфейс PCI Express 3.0 x16
Система охлаждения активная
Дизайн системы охлаждения референсная
Бэкплейт +
Частоты работы GPU, МГц 1515 (Boost – 1800)
Частоты работы памяти, МГц 14000
Шина памяти, бит 256
Выходные разъемы 1xHDMI (2.0b), 3xDisplayPort (1.4), 1xUSB Type-C
Размеры, мм 266x115x2-Slot
Низкопрофильная
Дополн ительное питание 1x6pin, 1x8pin
Потребляемая мощность, Вт 225
Рекомендуемая мощность блока питания, Вт 650
Поддержка DirectX DirectX 12
Разное NVIDIA CUDA Cores: 2944; RTX-OPS: 60T

Leave a Reply

Subscribe To Our Newsletter

Join our mailing list to receive the latest news and updates from our team.

You have Successfully Subscribed!