Nvidia GeForce RTX 3070 : Review |Test |CUP | Specs |Config

Nvidia GeForce RTX 3070 “titleonly”
: Test |CUP | Specs |Config

tl;dr: After the GeForce RTX 3080 with 10 GB for 700 euros and the GeForce RTX 3090 with 24 GB for 1,500 euros, Nvidia is sending the 500 euros GeForce RTX 3070 with 8 GB into the race. In the test, it had a hard exchange of blows with the GeForce RTX 2080 Ti – with advantages and disadvantages for both. The Founders Edition is convincing.

The GeForce RTX 3070 Founders Edition in the test

The GeForce RTX 3080 and GeForce RTX 3090 showed a very high performance with and without ray tracing in the test, whereby the GeForce RTX 3080 in particular made a big leap in performance compared to its predecessor and at the same time came up with an attractive price of 699 euros – if the card would be deliverable (at this price).

What about the availability?

The availability of the first two gaming ampere graphics cards worldwide is catastrophically poor even weeks after the launch and Nvidia itself does not expect that anything will change significantly this year. Against this background, the start of the GeForce RTX 3070 was pushed back two weeks in order to have more goods in stock. Next Thursday it will show what the relationship between supply and demand will be in this case.

At the launch in September, the interest in the GeForce RTX 3070 was enormous. With an MSRP of 499 euros, the model penetrates an even more attractive price region and is expected to perform as much as the GeForce RTX 2080 Ti, which cost well over 1,000 euros, for two years before. That would give a decent performance boost compared to its direct predecessor, the GeForce RTX 2070 (Great) to speak, which was also more expensive. With the next smaller GA104 GPU used for the first time, the power consumption will drop significantly compared to the larger models. On the opposite side is a memory of only 8 GB, which has remained unchanged for two generations or the GeForce GTX 1070 from 2016.

Founders Edition today, custom designs on Thursday

ComputerBase uses a Founders Edition model to determine whether the overall package of the GeForce RTX 3070 is convincing. At first glance, it is surprisingly small, optically and haptically but still as high-quality as the RTX 3080 and RTX 3090. A test with the first three custom models will follow in time for the market launch on October 29th.

Nvidia GeForce RTX 3070 FE
Nvidia GeForce RTX 3070 FE “titleonly”
Nvidia GeForce RTX 3070 FE
Nvidia GeForce RTX 3070 FE “titleonly”

GA104: Premiere on the RTX 3070

The GeForce RTX 3080 and GeForce RTX 3090 are equipped with the large GA102 GPU, while the GeForce RTX 3070 is based on the smaller GA104, which with 393 mm² and 17.4 billion transistors is significantly less complex than its bigger brother.

46 streaming multiprocessors and thus still 5,888 FP32-ALUs are left on the GeForce RTX 3070, the GeForce RTX 3090 is based on 82 SMs and 10,496 ALUs in comparison. Like the consumer flagship, the GeForce RTX 3070 does not make full use of the GPU, because 48 SMs (6,144 ALUs) are physically available.

The power consumption drops significantly with the performance

The significantly reduced execution units mean that the power requirement of the GeForce RTX 3070 with a TDP of only 220 watts is significantly lower. This allows custom designs with two 8-pin power plugs or even just one. Nvidia still wants to keep the clock speeds high, because the base clock is 1,500 MHz and the average turbo of 1,730 MHz is slightly higher than in the two other models.

In terms of memory, the GeForce RTX 3070 also falls significantly behind the two large variants. There are only eight 32-bit wide memory controllers that result in a 256-bit memory interface. This also automatically reduces the memory expansion to 8 GB – 16 GB would be technically possible as an alternative, but a corresponding model has apparently been deleted. In addition, the GeForce RTX 3070 no longer relies on extremely high-clocked GDDR6X, but on classic GDDR6 memory. Its clock rate is 7,000 MHz, resulting in a data throughput of 14 Gbps.

This means that the GeForce RTX 3070 has 41 percent less memory bandwidth available than the GeForce RTX 3080, while the theoretical computing power is only 32 percent less.

Nvidia GeForce RTX 3070 FE (links) vs. RTX 3080 FE (rechts)
Nvidia GeForce RTX 3070 FE (links) vs. RTX 3080 FE (rechts)
Nvidia GeForce RTX 3070 FE (links) vs. RTX 3080 FE (rechts)
Nvidia GeForce RTX 3070 FE (links) vs. RTX 3080 FE (rechts)
Nvidia GeForce RTX 3070 FE
Nvidia GeForce RTX 3070 FE “titleonly”

Details on the ampere architecture can be found in the technology article

If you want to find out more details about the technology of the GeForce RTX 3070, you should take a look at the technology article on the GeForce RTX 3000. In addition to the GeForce RTX 3070, this also covers the GeForce RTX 3080 and the flagship GeForce RTX 3090 in great detail.

The technical specifications of the GeForce RTX 3080 at a glance

New offerings in the GeForce ecosystem

Nvidia has not only upgraded Ampere’s technology, the ecosystem has also been equipped with new features. This includes broadcast, which uses the tensor cores of the Turing and Ampere GPUs to facilitate and improve (game) streaming to the Internet. Broadcast, for example, offers AI-supported noise suppression that was previously known as RTX Voice. In addition, there is the option of removing the background of the webcam feed and replacing it with an alternative image or just a blurring effect – a green screen without a green screen via AI, so to speak. Broadcast is primarily aimed at streaming games, but normal video streaming, for example at conferences, should also benefit from the feature. Also new to the ecosystem is Nvidia Reflex (test), which is supposed to lower latencies in esports titles via SDK and to make them measurable for the first time with the Reflex Latency Analyzer on new G-Sync displays at 360 Hz. Reflex is compatible with all GPUs from Maxwell onwards.

Die Nvidia GeForce RTX 3070 Founders Edition im Detail

Nvidia is also sending its own Founders Editions into the race for the GeForce RTX 3070, which changes hands from 499 euros (RRP). It is currently unclear how the availability will be. Nvidia will sell the Founders Edition in Germany both in its own shop and through partner shops.

Nvidia GeForce RTX 3070 FE
Nvidia GeForce RTX 3070 FE “titleonly”

The GeForce RTX 3070 FE is based on the look and feel of the two larger models and thus once again gives a first-class impression, even if the weight is significantly lower. The GeForce RTX 3070 FE is a lot more compact with a length of not quite 25 cm and a height of 11 cm. The GeForce RTX 3070 FE takes its own, very classic approach to the cooler.

With a classic cooling concept

The GeForce RTX 3070 Founders Edition also uses two axial fans with a diameter of 85 mm that are as large as the GeForce RTX 3080 FE. But in this case they are both classically placed on the front and side by side. The fan on the left blows the cool air against the PCB and tries to get it out of the housing as well as possible through the slot cover. The one on the right uses the PCB, which is much shorter than the cooler, and blows air through the cooling fins once without a PCB causing turbulence and counter pressure. The new fan stop at low temperatures is also part of the game. The GeForce RTX 3070 FE therefore also works silently on the Windows desktop.

Weight of the previously tested GeForce RTX 3070/3080/3090 graphics cards

With a new 12-pin power connector

The GeForce RTX 3070 FE is configured with a power limit of 220 watts, which can be manually increased to a maximum of 240 watts. The new 12-pin power connector is again used as the power connection, with Nvidia including an adapter to a simple 8-pin connector from the Founders Edition. As a result, Nvidia is working to the limit of the specifications and sometimes a little beyond that. More details on this in the section on power consumption.

Because the PCB is significantly shorter than the cooler and Nvidia works without an extension, the 12-pin power plug is almost centrally located between the two fans and – unlike the GeForce RTX 3080 FE – parallel to the PCB. When the adapter is used, it is not only clearly visible from the card, but also quickly hits the housing door. Something else is beautiful.

The 8-pin adapter on the GeForce RTX 3070 FE
The 8-pin adapter on the GeForce RTX 3070 FE
Monitor connections on the GeForce RTX 3070 FE
Monitor connections on the GeForce RTX 3070 FE
Nvidia GeForce RTX 3070 FE (unten) vs. RTX 3080 FE (oben)
Nvidia GeForce RTX 3070 FE (unten) vs. RTX 3080 FE (oben)

The clock rates of the GeForce RTX 3070 Founders Edition, like the TDP, adhere to Nvidia’s reference specifications. The base clock is thus 1,500 MHz and the average turbo is 1,730 MHz. The 8 GB GDDR6 memory works at 7,000 MHz. Monitors can be connected to three DisplayPort 1.4 interfaces with DSC or to an HDMI 2.1 output.

On the next page: clock rates, benchmarks in Full HD, WQHD, UHD

Leave a Reply

Your email address will not be published. Required fields are marked *