NVIDIA recently introduced three updated GeForce RTX Super graphics cards. The novelties were a preemptive blow to the release of the Radeon RX 5700 XT and Radeon RX 5700. Having ready-made solutions on a progressive architecture, NVIDIA simply slightly adjusted the characteristics and frequencies, increasing performance for new models. As a result, both old models and new Super versions appeared on the market, which can confuse a potential buyer. We will understand all the nuances in a series of reviews where we compare different NVIDIA video cards. And let’s start our acquaintance with the GeForce RTX 2070 Super
First we need to talk about the characteristics. At the moment, the GeForce RTX 2060 Super, GeForce RTX 2070 Super and GeForce RTX 2080 Super are presented. All of them use previously released Turing architecture GPUs, but in a new configuration. All of them fully support all new NVIDIA technologies, including ray tracing, hardware-accelerated deep learning technologies, and more. That is, they allow you to activate improved visual effects based on tracing and a special rendering method with DLSS anti-aliasing. Recently, we analyzed the features of these technologies using the example of three games.
The game Control is coming out the other day, which will have reflections and shadows based on tracing, and a special update for Wolfenstein: Youngblood is in the pipeline. There will be new effects in Watch Dogs: Legion, Call of Duty: Modern Warfare, Vampire: The Masquerade — Bloodlines 2 and many other projects. In addition to supporting new effects, Turing solutions have a number of important improvements and optimizations, which makes GeForce RTX graphics cards the fastest and most advanced at the moment.
Super models are accelerated by increasing the number of active compute units, increasing operating frequencies and switching to faster memory.
For clarity, we have summarized the characteristics of the updated and old video cards in one table.
|Video adapter||GeForce RTX 2080 Super||GeForce RTX 2080||GeForce RTX 2070 Super||GeForce RTX 2070||GeForce RTX 2060 Super||GeForce RTX 2060|
|Number of transistors, million pieces||13600||13600||13600||10800||10800||10800|
|Process technology, nm||12||12||12||12||12||12|
|Core area, sq. mm||545||545||545||445||445||445|
|Number of CUDA Stream Processors||3072||2944||2560||2304||2176||1920|
|Number of tensor cores||384||368||320||288||272||240|
|Number of RT cores||48||46||40||36||34||30|
|Number of texture blocks||192||184||160||144||136||120|
|Number of render units||64||64||64||64||64||48|
|Core frequency Base, MHz||1650||1515||1605||1410||1470||1365|
|Boost core frequency, MHz||1815||1710||1770||1620||1650||1680|
|Memory bus, bit||256||256||256||256||256||192|
|Memory frequency, MHz||15500||14000||14000||14000||14000||14000|
|Memory size, GB||8||8||8||8||8||6|
|Supported version of DirectX||12 (12_1)||12 (12_1)||12 (12_1)||12 (12_1)||12 (12_1)||12 (12_1)|
|Interface||PCI-E 3.0||PCI-E 3.0||PCI-E 3.0||PCI-E 3.0||PCI-E 3.0||PCI-E 3.0|
Previously, video cards used processors with partially disabled blocks. The GeForce RTX 2080 Super now features the full-featured TU104 GPU with all 3072 stream processors and 192 texture units, plus 384 tensor cores for deep learning and 48 RT cores for accelerated tracing. Curiously, NVIDIA mentions the superiority of this video card over the Titan Xp. The base frequency has been increased from 1515 to 1650 MHz, and the average Boost from 1710 to 1815 MHz. It should be noted that there were also GeForce RTX 2080 Founders Edition with Boost Clock 1800 MHz, and with them the frequency difference is already very small. The memory is accelerated to 15500 MHz instead of the standard value of 14000 MHz.
If the simple GeForce RTX 2070 was based on the TU106 processor, then the new GeForce RTX 2070 Super switched to a stripped-down version of the TU104. This made it possible to open 2560 stream processors and 160 texture units instead of 2304 and 144 units in the original version. There are now 160 tensor cores, and 40 RT blocks. In fact, we have an intermediate configuration between the GeForce RTX 2080 and GeForce RTX 2070. Plus, the core frequencies have been increased — the base has grown from 1410 to 1605 MHz, and the Boost Clock has reached 1770 MHz instead of 1620 MHz. The memory parameters remained at the old level — these are 8 GB GDDR6 modules with a data exchange rate of 14 GHz.
The GeForce RTX 2060 update came out the most interesting. The super version received a block configuration close to the GeForce RTX 2070 — this is the TU106 processor with 2176 stream processors, 136 texture units, 272 tensor units and 34 RT cores. GPU frequencies increased from 1365/1680 MHz to 1470/1650 MHz. But the most important thing is that the amount of memory has been increased. Now there is a full 8 GB on board on a 256-bit bus instead of 6 GB on a 192-bit bus.
Such an upgrade did not go unnoticed for TDP. Therefore, a higher power level is declared for all Supers. We will talk about the performance difference after our testing, where we compare the GeForce RTX 2070 Super with the GeForce RTX 2070, old GeForce models and new competitors from AMD.
ASUS Dual GeForce RTX 2070 SUPER EVO OC (ASUS DUAL-RTX2070S-O8G-EVO)
The video card comes in a large box.
ASUS looks simple and bulky. Occupies three slots. The cooling system uses two fans.
Full length up to 27 cm.
On the side face there is a large inscription GeForce RTX ASUS. The backlight is limited to a small oblique insert in the case. In the corner there are two power connectors — six and eight pins.
The reverse side of the video card is covered with a metal plate. It has a protective function, it does not remove heat from the elements. The presence of numerous slots will not interfere with the natural convection of warm air.
There are three DisplayPort ports and one HDMI on the back panel.
Let’s take a look at the cooling system. The radiator consists of two sections and four heat pipes. The top sections are blown by 90 mm fans (total impeller diameter 88 mm). The fans themselves are made using Axial-tech technology — a special shape of the blades and a ring around the edges allow you to increase the air flow and reduce noise at high speeds. Such fans are used in the new ASUS Strix video cards.
It is worth noting that the ASUS Dual GeForce RTX 2070 video card, with similar dimensions, used a radiator without heat pipes, compensating for this with its large size and the presence of an evaporation chamber.
This design resembles a stripped-down heatsink from Strix, where there is more free space, which, given the presence of two fans, is quite justified. One part of the heatsink removes heat from the GPU and memory chips, the second section removes heat from the VRM.
Contact with the surface of the graphic crystal is carried out through a flat base. Heat pipes are fixed in the channels between the base and the block of plates.
The design of the printed circuit board resembles the older GeForce RTX 2080, which is quite logical, since this is a stripped-down version of the GeForce RTX 2080.
The right side of the board is occupied by a large power system, which has eight phases for the GPU and two phases for powering the memory.
Full GPU marking — TU104-410-A1.
Soldered Micron 9EA77 D9WCW memory chips.
ASUS DUAL-RTX2070S-O8G-EVO is factory overclocked. With the standard base value of 1605 MHz, the average Boost is increased to 1815 MHz.
The actual Boost frequencies turned out to be slightly higher than 1900 MHz. In Final Fantasy XV Windows Edition, the main operating range is 1920-1965 MHz, in The Division 2 there were drawdowns up to 1890 MHz. In game mode, the core warmed up to 72 °C on an open stand at a room temperature of up to 27 °C. The fans spin up to 1800-1900 rpm, the noise is moderate.
Since we are interested in the performance of simple versions of the GeForce RTX 2070 Super, we need a slight frequency correction. By lowering the base value by 50 MHz, we got a working frequency range of 1845-1905 MHz, which roughly corresponds to the frequencies of the reference video cards of the new series.
It’s time to talk about overclocking. We note right away that the frequency potential of the updated video adapters remained at the standard level. And given the initially increased frequencies, the potential for additional acceleration has decreased. Our ASUS Dual managed to overclock by 45 MHz at the base frequency.
When the power limit was increased to the maximum, this acceleration made it possible to achieve Boost frequencies of 1995-2010 MHz with peak Boost up to 2085 MHz. In some games, the frequency aspired to 2025 MHz, there were rare drawdowns up to 1980 MHz. The memory was overclocked to 7726 (15452) MHz.
When the fan speed increased to 2400 rpm, the core temperature reached 78 °C. At the end of the tests there was one hang. This means that the video adapter worked at the limit, where air cooling is barely enough to operate at maximum frequency. If you try to achieve a balance between boosting frequencies and comfortable acoustic mode, then acceleration will be very small.