Nvidia GeForce RTX 2060 FE in the test

– With the GeForce RTX 2060, Nvidia is bringing the new Turing architecture to the middle class. At an RRP of 369 euros, it achieves the performance of the more expensive GeForce GTX 1070 Ti in the test and is only just behind the 150 euros more expensive GeForce RTX 2070. With 6 GB instead of 8 GB of memory, however, it sacrifices future security.

GeForce RTX 2060: The Turing middle class

With the new GeForce RTX 2060, Nvidia is launching the first Turing card under the GeForce RTX 2070 (test), GeForce RTX 2080 and GeForce RTX 2080 Ti (test), the prices of which exceed 500 euros. The price for the GeForce RTX 2060 for the variant with 6 GB is 369 euros. Turing is therefore becoming cheaper, but at least in parts remains true to its pricing.

Because the GeForce RTX 2060 is not only faster than its predecessor, the GeForce GTX 1060, but also more expensive: it costs 90 euros, half of the price gap to the GeForce GTX 1070 Ti is thus closed. Unlike the predecessor, the new graphics card on the GPU is not slimmed down significantly, so the differences to the larger model are small at this point.

There is a damper already known from Turing: In fact, the GeForce RTX 2060 only has a maximum of 6 GB of memory. Like the GeForce RTX 2080 Ti, 2080 and 2070, the new GeForce RTX 2060 does not offer more memory than its predecessor by name, although the GPU performance makes a leap in the test. The rumor mill also speaks of other memory designs (of 3 GB, 4 GB and 5 GB and optionally also GDDR5X), but today only the expansion level with 6,144 MB is presented.

The GeForce RTX 2060 in the test

The Nvidia GeForce RTX 2060 Founders Edition clarifies how the GeForce RTX 2060 with its comparatively strong GPU performs in classic games, but also in the first RTX application Battlefield V.

A stripped down TU106 in use

Like the GeForce RTX 2070, the GeForce RTX 2060 is based on the TU106 GPU, but the 445 mm² and 10.8 billion transistor heavy chip is no longer used in its full configuration. 6 of 36 streaming multiprocessors, or 17 percent, were deleted, leaving 1,920 shader units (from 2,304) and 120 texture mapping units (from 144) left.

However, the GeForce RTX 2060 is significantly less cropped than the GeForce GTX 1060 compared to the GTX 1070: the Pascal predecessor was 50 percent different. At that time, Nvidia therefore used two GPUs for this leap in class: GP106 and GP104. The GeForce RTX 2060 thus adheres to the pattern known from Turing, because the distances between the Turing GPUs TU106 (RTX 2060 & RTX 2070) and TU104 (RTX 2080) are significantly smaller than between the Pascal derivatives.

Much more raw power and memory bandwidth

Nvidia specifies a base clock of 1,365 MHz for the GeForce RTX 2060, the average turbo should be 1,680 MHz. The computing power of the GeForce RTX 2060 is clearly superior to that of the GeForce GTX 1060.

This is also evident in the memory bandwidth. The memory interface remains at 192 bits (two 32-bit controllers are switched off on the TU106) – this means that the number of ROPs stagnates at 48 – but Nvidia uses significantly faster memory. As with the large Turing models, at least the fastest subject uses 7,000 MHz faster GDDR6-VRAM. This gives the new model 75 percent more memory bandwidth and about 48 percent more computing power.

Standstill during storage expansion

The slimmed-down interface automatically results in a smaller memory compared to the GeForce RTX 2070: The GeForce RTX 2060 only has a 6 GB VRAM. As a result, there has been no progress in memory size in the entire Turing product portfolio for players, which is a clear minus point two and a half years after the introduction of Pascal.

Nvidia specifies a TDP of 160 watts for the GeForce RTX 2060. This is 15 watts less than the GeForce RTX 2070, but 40 watts more than the GeForce GTX 1060 – despite the more efficient 12 nm production. Here the new graphics card has to pay tribute to the much larger GPU.

RTX 2060 uses the same technology as RTX 2070, 2080 & 2080 Ti

The bottom line of the technology of the GeForce RTX 2060 is unchanged compared to the larger Turing models. More about Turing and the changes compared to Pascal can be found in the article "Turing TU102, -104, -106: The technology of the Nvidia GeForce RTX 2080 Ti, 2080 & 2070".

The GeForce RTX 2060 Founders Edition in detail

Like the larger representatives of Nvidia, the GeForce RTX 2060 is again available as a Founders Edition, but differs in one point from the FE models with GeForce RTX 2070, RTX 2080 and RTX 2080 Ti: It runs at the reference clock speeds and is not faster than the slowest custom designs of the partners. The price for the model is 369 euros. The graphics card will be available in stores from January 15th.

PCB and cooler are known from the RTX 2070 FE

At first glance, the GeForce RTX 2060 FE looks identical to the GeForce RTX 2070 FE, but that is actually wrong. Nvidia does not use aluminum for the cover and instead uses plastic, which is hardly noticeable visually. The dual-slot cooling system, on the other hand, comes with two axial fans with a diameter of 85 mm, which run at 1,200 revolutions per minute under Windows. The vapor chamber on the bottom is almost as long as the card itself, there are no heat pipes. A backplate is installed on the back.

The PCB is also known: it comes from the GeForce RTX 2070 FE – considering the same GPU, this is not a surprise.

There is no RGB on the Founders Edition

The NVLink connection for SLI known from the two fastest models does not exist on the GeForce RTX 2060, because the TU106 GPU does not support NVLink. An RGB-LED lighting is not available, but the illuminated GeForce RTX lettering can now be adjusted (brightness, permanently illuminated, flashing).

The base clock of the GeForce RTX 2060 Founders Edition is 1,365 MHz, the average turbo is specified at 1,680 MHz. The 6,144 MB GDDR6 memory works at 7,000 MHz – as fast as on any Founders Edition with a Turing GPU so far. The TDP is 160 watts, which is why Nvidia provides an eight-pin power connector. This is placed at the rear end and points towards the drive cages – a rather unusual position.

DVI returns

Monitors on the graphics card can be controlled via two DisplayPorts 1.4, an HDMI 2.0b and a dual-link DVI output – the faster Turing models have so far dispensed with the latter. There is also a USB Type-C port that is designed as a VirtualLink for future VR headsets. These then only need a cable.

On the next page: test system and test results