After reviewing the Turing architecture and new NVIDIA technologies, we begin to get hands-on with the new graphics cards. In this article, we will look at two versions of the GeForce RTX 2080, conduct game testing and compare with the GeForce GTX 1080 Ti. First, let’s talk about the GeForce RTX 2080 Founders Edition, and then we’ll explore the capabilities of the equivalent ASUS Dual GeForce RTX 2080 graphics card.
GeForce RTX 2080 Founders Edition
According to the name, the GeForce RTX 2080 can be considered a successor to the GeForce GTX 1080, but NVIDIA is positioning the new product faster than the GeForce GTX 1080 Ti. And this is quite a standard phenomenon, as each new generation raises the performance bar. The GeForce RTX 2080 is also positioned as a solution capable of providing 60 fps in 4K mode. And this is quite real, if you do not chase the opportunity to unscrew all the settings sliders to the limit in new games.
At the heart of the GeForce RTX 2080 is the TU104 GPU with six GTC clusters of eight multiprocessor SM blocks each. In total, this GPU has 3072 CUDA stream processors, but due to two deactivated SMs, 2944 stream cores are active. At the same time, the potential of the GPU is represented by 368 tensor cores, 46 RT cores for ray tracing and 184 texture units.
TU104 is similar to GP104 but with more compute units. The analogy is strengthened by the presence of a 256-bit memory bus. But in the new generation, the SM structure has been redesigned, a number of optimizations have been carried out, and new features have been introduced in working with shaders. We talked about this in the last review, so we will briefly recall some key points.
Turing’s promising features include support for Mesh Shading, Variable Rate Shading, and Texture Space Shading, which can improve performance significantly, but require implementation by developers. The main technological breakthrough was hardware-accelerated ray tracing, which for the first time will allow this technique to be used in real time for rendering in games. But for now, this is only the first step in this direction, so NVIDIA is introducing hybrid rendering, where tracing capabilities will be used to implement some additional effects using NVIDIA RTX technology. The company is also working on integrating artificial intelligence technologies into the consumer segment. With the help of neural networks and deep learning, you can implement additional functions to improve the picture, and the acceleration of calculations is provided by special tensor cores. DLSS (Deep Learning Super-Sampling) anti-aliasing has already been introduced, which is an evolution of the TAA algorithm, but provides better image clarity with increased performance. And in this review, we will compare the performance of these anti-aliasing methods in one of the specialized tests.
At the moment, it is difficult to assess the real benefits of Turing in trace rendering or when using tensorboxes. This potential of the new GeForce RTX will be revealed over time. This is a topic for a separate review in a few months. If we talk about the computing potential for traditional rendering methods, then the GeForce RTX 2080 with its 2944 stream processors looks weaker than the GeForce GTX 1080 Ti with 3584 cores. The old flagship and other blocks have more, plus a 352-bit memory bus. But according to the results of the tests, you will see that the GeForce RTX 2080 is indeed faster than the GeForce GTX 1080 Ti.
|Video adapter||GeForce RTX 2080||GeForce GTX 1080 Ti||GeForce GTX 1080|
|Number of transistors, million pieces||13600||12000||7200|
|Process technology, nm||12||16||16|
|Core area, sq. mm||545||471||314|
|Number of CUDA Stream Processors||2944||3584||2560|
|Number of tensor cores||368||–||–|
|Number of RT cores||46||–||–|
|Number of texture blocks||184||224||160|
|Number of render units||64||88||64|
|Core frequency Base, MHz||1515||1480||1607|
|Boost core frequency, MHz||1710 (1800)||1582||1733|
|Memory bus, bit||256||352||256|
|Memory frequency, MHz||14000||11008||10010|
|Memory size, GB||8||11||8|
|Supported version of DirectX||12 (12_1)||12 (12_1)||12 (12_1)|
|Interface||PCI-E 3.0||PCI-E 3.0||PCI-E 3.0|
For the GeForce RTX 2080, the base core frequency of 1515 MHz is declared. The average Boost Clock for regular graphics cards is 1710 MHz, and for Founders Edition models it is 1800 MHz. Initially, there is a small frequency differentiation, and standard versions from NVIDIA should be faster than non-reference models from partners. Although many OC versions from different manufacturers have already been presented, where the Boost Clock is at 1800 MHz and even higher. All GeForce RTXs use the new GDDR6 memory, which is an evolution of GDDR5X. The frequencies of the new modules have increased, but the power consumption has remained at the old level. The effective communication frequency is equivalent to 14 GHz, while the memory capacity is 8 GB. This is the only nuance that can confuse against the backdrop of the GeForce GTX 1080 Ti, because the old top-end card has 11 GB.
The TU104 processor has more transistors than the GP102. And even when switching to a new process technology, 12nm FinFet has a larger die. At the same time, the final TDP for the GeForce RTX 2080 is declared at 215 (225 W for the Founders Edition). This is below the TDP of the old flagship GeForce GTX 1080 Ti, but higher than the consumption of the direct predecessor GeForce GTX 1080. And this is against the background of the fact that the GeForce RTX 2080 is the fastest video card in this trio.
NVIDIA has changed the design of cooling for the first time in a long time. Abandoning the usual “turbine”, the manufacturer equipped the Founders Edition video cards with a new cooler with two fans. Due to this, the thermal and noise characteristics of the GeForce RTX are noticeably better than those of previous models. The GeForce RTX 2080 is a few dB quieter than the Pascal reference models, and operating temperatures are below 75°C, which is really impressive.
If in the past generation standard NVIDIA video cards were a kind of compromise solution, now such models are devoid of any drawbacks. They operate at a higher Boost and initially have excellent thermal and noise characteristics. But the cost of the GeForce RTX Founders Edition is $100 higher than the standard recommended price.
Such video cards are supplied in a small box. The device itself is securely fixed inside in a special form.
The photo below shows the GeForce RTX 2080 Founders Edition packaging. The video card is also packed in a metallized package. Inside the box there are instructions and a DisplayPort-DVI adapter.
The length of the GeForce RTX 2080 is approximately 26.5 cm. Cooling occupies the standard two slots. From above, the video card is covered with a casing with a pair of fans, the impeller diameter of which is 85 mm. The shape of the blades resembles the fans of ASUS video cards.
The design uses a combination of black and silver elements. Body with rounded smooth edges.
The reverse side is covered with a metal plate, on which, as well as on the front side of the case between the fans, the name of the video card is indicated. In the corner there are two power connectors — 6 and 8 pins.
On the side face there is a large illuminated GeForce RTX inscription.
We had limited time to get acquainted with this video card. I had to do without a breakdown. Therefore, here are official photos to demonstrate the internal cooling device and the circuit board.
Under the fans is a large heatsink that occupies the entire available area. Contact with the graphics chip is made through the evaporation chamber. There is a massive metal base under the heatsink, which traditionally contacts other elements on the board, removing heat from transistors and memory chips.
Such cooling looks solid. New graphics cards even weigh more than many other models. A pleasant heaviness in the hand immediately indicates that the manufacturer did not save on metal elements and a radiator.
The printed circuit board resembles the solutions of the previous series with an increased number of phases — now there are 10 of them. The memory is made up of eight GDDR6 modules. On the left side of the board there is a connector for switching video cards in SLI mode via the NVLink interface.
The GeForce RTX 2080 and GeForce RTX 2080 Ti of the Founders Edition are identical.
The frequency characteristics of the GeForce RTX 2080 Founders Edition are standard. This is the base frequency of 1515 MHz with Boost Clock 1800 MHz, the effective memory frequency is 14000 MHz.
Practical acquaintance with the GeForce RTX 2080 Founders Edition will be limited to checking the temperature regime. At the time of testing, EVGA Precision X1 was the best monitoring solution. The load was carried out by the Final Fantasy XV Windows Edition benchmark in 4K with Ultra-graphics, which allows us to evaluate the cooling capabilities in games with heavy graphics conditions. Based on the results of the 14-minute test, the GPU did not warm up above 73 ° C. The average Boost frequency was close to the level of 1860 MHz, there were rare drawdowns up to 1830 MHz. Peak jumps Boost reached 1965-1980 MHz.
The fan speed reached a maximum value of 1880 rpm. Such parameters were demonstrated on an open stand in a room with a temperature of 22 °C. Inside the case, in more severe operating conditions, the heating may be higher. However, the graphics card clearly has the potential to hold initial temperatures while maintaining rotational speeds within 2000 rpm or so.
The GeForce RTX 2080 has separate monitoring and fan control. They can work both in synchronous mode and at different speeds set by the user. All this is available in EVGA Precision X1, as well as the ability to manually overclock the core and memory.
The utility has a fine-tuning of fans through curves, where the dependence of speed on temperatures is set at several points.
With the transition to GPU Boost 4.0, the Turing family has opened up new Boost settings. The user can set several temperature points at which the reduction of operating frequencies occurs.
In the presence of highly efficient cooling, all these fine adjustments are not so important. Operating temperatures are not high, and the main factor limiting frequencies is the power limit. Here, the standard + 20% to the initial level is available to the user, just like Pascal video cards. Theoretically, the Founders Edition has a slightly higher initial power limit, which gives more room for subsequent overclocking. However, ASUS, MSI, Gigabyte and other manufacturers will clearly offer top-end versions of the GeForce RTX 2080 and GeForce RTX 2080 Ti, where the TU104’s potential will be even higher.
As for overclocking, we experimented with it on the example of an ASUS video card, which is reviewed below.