The controversy surrounding Turing Nvidia's graphic architecture seems to be divided into two camps. On the one hand, critics are generally dissatisfied about the way the GeForce RTX 20-series rates compared to the previous gene board. Or, they are not happy that Nvidia has committed so much resources to features that appear in the future that may not have a large visual impact or run smoothly, except for the ultra-high-end models.
The GeForce RTX 2060 sets out to change minds on both fronts. At a price of $ 350 / £ 330, it landed below the launch price of the GeForce GTX 1070. However, the geometric average of the average frame rate shows the GeForce RTX 2060 faster than the GeForce GTX 1070 Ti! What's more, the latest updates to Battlefield V make Nvidia's real-time ray rasterization / tracking approach playable on lower-end hardware. We averaged 68 frames per second (FPS) in 2060 through our benchmark at 1920x1080 with graphic options, including DXR Reflection Quality, set to Ultra.
NVIDIA GEFORCE RTX 2060 FE
- Better performance than GeForce GTX 1070 Ti
- Fair price point
- High-quality cooler
- Playable performance in Battlefield V with DXR enabled
- Higher power consumption than previous-gen cards it replaces
- Axial fans exhaust waste heat into your case
The GeForce RTX 2060 Founders Edition found itself surrounded by competing competitors on the GeForce GTX 1070, GTX 1070 Ti, Radeon RX Vega 56 and Vega 64. Most outperformed them all and at a lower price point.
Meet TU106, Again: Impressive, Even After Cutting Hair
We were first introduced to the Turing10 based Turing processor when Nvidia launched its GeForce RTX 2070. The Founders Edition board is a little faster than the GeForce GTX 1080, but seems too expensive at $ 600.
The TU106 version used on the RTX 2060 boasts 83 percent of the CUDA 2070 core, Tensor core, and RT core; 75 percent ROP, memory bus, GDDR6, and L2 cache; and operates at 98 percent of the GPU Boost clock rate. However, at a price of $ 350 / £ 330, 2060 sells for around 60 percent of the price of the RTX 2070 Founder Edition. With numbers, the story of the value of this card will definitely be much better.
In architecture, the TU106 is very similar to the GeForce RTX 2080 TU104 processor and GeForce RTX 2080 Ti TU102. Nvidia only makes chips use fewer modular blocks than those that go into high-end graphics cards. TU106 consists of three Graph Processing Clusters, each with six Texture Processing Clusters. Throughout the Turing family, TPC includes two Multiprocessor Streaming, adding up to 36 BC in the TU106 which was not cut. Nvidia deactivated six SM for RTX 2060, so the card ended with 1920 CUDA cores, 240 Tensor cores, 30 RT cores and 120 texture units. The Core Tensor and RT are typical features of this generation. So, if you miss our deep dive into their destination, be sure to check out the Explored Turing Nvidia Architecture: Inside the GeForce RTX 2080.
Nvidia's basic clock speed for 2060 is 1,365 MHz (compared to 2070 1,410 MHz). Furthermore, the GPU Enhancement rating is 1,680 MHz (versus 1,710 MHz). The Founders Edition card uses TDP 160W, down 25W from 2070, but goes up compared to 1060 (120W) and even 1070 (150W).
Without being cut off, TU106 maintains the same 256-bit memory bus as TU104. However 2060 lost two of the 32-bit GPU memory controllers, producing a total of 192-bit buses that were filled with 6GB of GDDR6 14 Gb / s modules that were capable of moving up to 336 GB / s. However, apart from the pared-back configuration, 2060 provides more memory bandwidth for the processor than the last generation 1070 Ti.
In disabling the memory controller, Nvidia also turned off a pair of ROP TU106 partitions and two 512KB L2 cache slices. What started as a GPU with 64 ROP and 4MB L2 consequently was trimmed to 48 ROP and 3MB cache for 2060. Like 2070, NVLink support has long been lost at this price point.
Although the TU106 is the most complex Turing-based GPU so far, the 445 mm² die contains no less than 10.8 billion transistors. It is still remarkable for what Nvidia might have considered as the middle part of her portfolio. By comparison, the GP106 - "Pascal mid-range" - is a 200 mm² chip with 4.4 billion transistors in it.
Based on the initial benchmark results in Battlefield V with DXR enabled, it appears as if real-time beam search is not feasible on models below 2070. However, the combination of weight optimization from EA DICE and relatively light cuts to the TU106 means that 2060 maintains playable performance in one game that currently utilizes Turing's most praised progress.
Meet GeForce RTX 2060 Founders Edition: The Little Engine That Could (Ray Trace)
From the outside, 2060 Founders Edition is exactly like the Nvidia version itself in 2070. This is the same height (11.5cm), width (3.9cm), and length (23cm). The compact dimensions and coolers that are less massive than the larger Ti 2080 and 2080 implementations also reduce weight. 2060 Founders Edition weighs 968g, places it in 2g from Nvidia 2070.
Up front, Nvidia uses the same pair of 8.5 cm, 13-blade axial fan, which blows through a stack of vertically oriented fins underneath. Even though we still lost the centrifugal fan and acrylic window of the Nvidia reference card beforehand, there is no doubt that the 2060 Founders Edition is made firmly and with high-quality materials.
Inverted, the plate covers the back of the PCB and wraps it around, touching the front sheath at both ends to make a clean and unbroken line. The back plate makes direct contact with several board components with thermal pads in between to facilitate better heat transfer.
The top of this card is far more attractive than the 2080 Founders Edition or 2080 Ti Founders Edition. Because 2060 does not support SLI through NVLink, there is no interface there. Furthermore, the additional power input moves backwards, providing cleaner aesthetics in windowed cases.
The single eight-pin power connector on the back will not cause problems right. Because 2060 Founders Edition is on the short side, the cable coming out of its back will not collide with a low hanging hard drive cage. And it certainly looks better than the front protruding connector.
Gamers who buy high-end graphics cards tend to have monitors with modern display inputs. At least that is what we got from Nvidia's decision to arm RTX 2080 and 2080 Ti with three DisplayPort connectors and one HDMI output, while 2060 Founders Edition features two DisplayPort 1.4 interfaces, HDMI 2.0b and a rare DVI-DL port. It still supports VirtualLink via a single USB Type-C card connector, and, like other 20-series RTX cards, you can use all four display outputs simultaneously for multi-monitor arrays.
With the fan hood disengaged, the RTX 2060 cooling unit extends from one end of the card, passes through the 19cm PCB and to the stabilizing side of the frame. The cable for the fan and lighting goes through the channel in the center, where the circuit board is visible, to the connector underneath. Whereas RTX 2080 and 2080 Ti share common thermal solutions, Founders Edition 2060 uses a copper sink with pipes that remove heat through all the fins.
The frame is held tightly to the PCB with a hex-head screw. After you release it, the frame can be lifted and lifted carefully. Be careful not to damage the thermal pad.
The 2060 power requirements are not as precise as the 2080s or 2080s, so smaller PCBs can be understood not too complicated. In fact, the 2060 Founders Edition is based on the same board as Nvidia's 2070, minus some components installed on the surface to take into account the fewer memory ICs and lower thermal ceilings.
Given the higher price platform 2060 and the mainstream performance level, we expect the Founders Edition cards to be a temporary replacement until additional board partners increase their own production of 2060. If you remember returning to 2016, Nvidia did something similar to 1060 Founders Edition, which turned out become a very limited product.
Like the 2070 Founders Edition, reference 2060 uses six phases of power for TU106 and two phases for GDDR6. While uP9512 uPI can technically control eight phases, it is easily adapted to a less sophisticated 2060 power supply. A simpler two-phase uP1666Q buck controller is sufficient for memory.
Gone are ON Semiconductor FDMF3170 Smart Power Stage modules with PowerTrench MOSFET and integrated driver IC, found in 2080 and 2080 Ti. Even cheaper NCP302155s used in 2070 (and capable of up to 55A) were lost, replaced by ON-semiconductor ON NCP302150s low with high and low side MOSFETs, plus drivers, each rated for 45A. Currently the GPU is smoothed by the usual 220mH coil, while the same memory uses a 470mH inductor.
Rotating the power connector is rotated 90 degrees and stretching the back of the frame of this card requires extension of type. Again, it's identical to the Founding Edition of 2070 that we reviewed last year. Instead of the eight pin connector that gives ~ 11.1A through three + 12V red lines, we measure ~ 8.4A thanks to reduced power consumption in 2060.
How We Test the GeForce RTX 2060
2060 is a graphics card that is more common than other Turing-based boards that we have reviewed. Thus, our graphics workstation, which is based on the MSI Z170 Gaming M7 motherboard and Intel Core i7-7700K CPU at 4.2 GHz, is very precise. This processor is equipped with a G.Kill F4-3000C15Q-16GRR memory kit. Crucial MX200 SSD is here, joining Intel DC P3700 1.4TB which is loaded with games.
As far as the competition goes, 2060 goes up against the GeForce GTX 1070 and 1070 Ti. Of course, comparison with the GeForce GTX 1060 is also unavoidable. All of these cards are included in our ranks, such as the GeForce RTX 2070. On the AMD side, we are very interested in the Radeon RX Vega 64 and the Radeon RX Vega 56, although the newer Radeon RX 590 seems to draw a comparison too.