[ad_1]
After announcing the first GeForce RTX 2070 graphics cards and reaching the market, we devoted ourselves to testing the production of KFA2. We looked closely at the KFA2 GeForce RTX 2070 in "OC" version and compared it to other popular competitors. The performance of the card in terms of performance, power consumption, overclocking and cooling clarifies this test.
Page 1
KFA2 GeForce RTX 2070 OC in detail
The GeForce RTX 2070 is designed to meet the high demands of gamers who do not want to spend anywhere from $ 600 to $ 1,000, but benefit from the next generation and features like Raytracing and DLSS cores. Unlike the GTX1070 and GTX1070 TI, the new RTX 2070 Founders Edition, which sells Nvidia directly to customers, costs € 629, much more than the previous generation. The models of the partners of the board of directors are well below the bar of 600 €, that is why Nvidia stopped distributing the reference cards to the editorial offices. the KFA2 GeForce RTX 2070 OC is currently one of the cheapest cards, but must not hide from the competition!
GeForce RTX 2070 in detail
The RTX-20 series is based on TSMC's 12-nanometer Turing FFN chips. The TU106-400A-A1 chip supplied with RTX 2070 cards is slightly "fine-tuned". The chip contains 10.6 billion transistors over an area of 445 mm². Unlike big brothers, this only includes 2304 shader units, 144 texture units and 64 render units. The map also includes 288 killing AI nucleus nuclei and 36 new ray tracking nuclei. For the first time, the new GDDR6 memory is used in RTX-20 cards. The RTX 2070 has 8GB of memory in total. Our sample contains Micron GDDR6 chips with the designation "8RA77 D9WCQ". These factory clocks with 1750 MHz and reached through the 256-bit memory connection a theoretical bandwidth of 448 GB / s.
Design and construction
Nvidia reference cards set the tone. The standard graphics processor clock is specified on all RTX 2070 cards at 1410 MHz. Boost brings out the first differences compared to the reference. Not all cards in the market reach Nvidia's Foundry Edition boost rate, which is at 1,710 MHz. Our test card surpbades this one, but with 1,815 MHz significantly. The memory clock is not much. Like almost all RTX 2070 cards, our test card is powered by an 8-pin PCIe connector and a 6-pin PCIe connector. When you recover the card for the first time, you will be surprised by its weight, because the cooler under the plastic cover is more mbadive than you think. Although KFA2 has set the clock much higher from the factory, but only 8 phases instead of 10 have been extended. On the printed circuit board, you will find two fan connectors and an underutilized RGB header pin, since only half of the pins were used.
On the connection side, the card has an HDMI 2.0b connection, three DisplayPort 1.4 and a USB-C connection with DisplayPort 1.4 and VirtualLink. The latter will be used in the future for the simple connection of virtual reality glbades. Instead of two or three cables, you only use one.
Page 2
Specifications and delivery
specifications:
|
|
|
|
||||
| chip | ||||||
| Shadereinheiten | ||||||
| Smart clock (stick) | ||||||
| Smart clock (boost) | ||||||
| Verbruach | ||||||
| Vertigung | ||||||
| Pitches blocked | ||||||
| memory size | ||||||
| memory interface | ||||||
| Memory bandwidth | ||||||
| fan | ||||||
| PCI Express | ||||||
| PCIe power socket | ||||||
| RGB Beleuchutng | ||||||
| DirectX version supported | ||||||
| price (EIA) |
||||||
delivery
The KFA2 GeForce RTX 2070 OC is supplied to the buyer in a very elaborate box. The typical Nvidia brand is implemented as usual and seems very consistent. The graphics card itself is carefully packaged in a very discreet black cardboard box. This opens very stiff, because a cardboard insert with the glossy lettering "GeForce RTX" is moved. This gives a sense of value, which corresponds to the price range. Under a foam insert are the graphics card, the manual and a small elongated case. This includes both PCIe to Molex adapters. Otherwise, nothing else is included.
We received our graphic card test template from the KFA2 Public Relations Agency. This provision did not affect the evaluation and objectivity of the test.
the system requirements
The following requirements are mentioned on the product packaging:
- PCI Express compatible motherboard with 16 PCIe slots
- 550 watt AC adapter
- 1.5 GB of disk space
- 8 GB of RAM (16 GB recommended
- Microsoft Windows 10 64-bit (Updated April 2018), Microsoft Windows 7 64-bit or 64-bit Linux as the operating system
- A 6-pin and 8-pin PCIe power cable
Page 3
Raytracing and DLSS
Real-time ray tracing in the future
At the press conference that preceded GamesCom 2018, Nvidia spoke about ray throwing almost all the time, the excellent quality of the new maps in real time and how they will change the game world. , the cards are on the market and there is not a single demo or technological reference to experiment this technology in action! Even Shadow of The Tomb Raider, which has been featured as the "RTX On" demo, has not yet received a ray cast patch. UL Benchmark, the company behind the famous 3DMark and PCMark benchmarks, does not yet have an appropriate benchmark.
There remains only the theory and the hope that Nvidia will have it again this year to present the RTX titles. Metro Exodus is Nvidia's next workhorse to support RTX. Until now, the game is not published. The videos shown here arouse the expectation and the hope that you will get more realism, especially in titles such as Battlefield 5, RTX, gives you a definite advantage, because you can recognize the opponents that lie actually behind the player. gives a definite advantage.
The GeForce RTX 2070 OC has 36 ray tracing cores and is designed to provide realistic highlights and shadows in the future. According to Nvidia, the radius calculation has been improved by a factor of 8 compared to the predecessor architecture "Pascal".
Super Sampling Deep Learning (DLSS)
In addition, Nvidia donated the new Turing architecture as well as second-generation tensor cores. The GeForce RTX 2070 has 288, while the GeForce RTX 2080 Ti has 544 tensor cores. Tensor cores are responsible for deep learning, that is, some artificial intelligence applications used in conjunction with RTX cores. They process partially completed raytracing images for "denoising". The most important task is to replace the anti-aliasing by Deep Learning Super Sampling (DLSS). Once the supercomputer has defined the best method, resolution and optimization of the image quality for the game, calculate the rest of the Tensor cores locally on the map, all of which must be done "over the air".
How does this technology affect reality and what advantage does it have with an RTX-20 graphics card compared to another RT-less card, you can not say exactly now. DLSS and RTX must be accepted only as future technology that you can not even use on the new cards, for lack of compatible games and demos.
Page 4
Test system and overclocking
The test system
The tests are done again in our ITX test system. The structure looks like this:
- CPU: AMD Ryzen 5 2600 at 3.7 GHz, 1.27850v
- motherboard: Asrock AB350 game ITX / ac
- SSD: Samsung 840 PRO 256 GB SSD, Samsung 850 EVO 500 GB
- m.2 SSD: Samsung 970 Evo MZ-V7E250BW 250 GB
- RAM: 16GB Patriot Viper 4 Kit, DDR4-3000, CL16-16-16-36 (PV416G300C6K)
- CPU Cooler: CRYORIG C7
- Feed: Corsair SF 450 80+ Gold with 450 watts
- fan: Fan 1 x 80 mm
- housing: Kolink rocket
Overclocking the GeForce RTX 2070 OC
For overclocking, we have the original software again "Xtreme tunerUsed by KFA2. For RTX cards, there is a new remote version, still in beta. The most important functions are already included, so nothing overclocking forbids.
We have slowly reached the limits of the map and have increased the GPU and the memory clock. But we also Power target from 100% to 110% increases. Properly stable and without artifacts, the card worked with an increase in GPU clock 111 MHz to 1521 MHz, what Increase of about 7.5% equivalent. The boost clock has increased accordingly from 1,815 to 1,926 MHz. Since the card may have an even longer clock based on usage and circumstances, it has reached a maximum clock of 2.055 MHz. the GDDR6 memory let yourself be changed 603 MHz to 1,901 MHz or effectively 7.604 MHz, which corresponds to an increase of 8.6%.
Page 5
Synthetic markers
So far, the GeForce RTX 2070 OC beats the latest generation an average of 11%If you compare the values after overclocking, you get on average 13%. Even the GTX 1080 beats the RTX 2070 OC, even though the Lead at 2.5 or 5 percent (at OC) is.
With Sky Diver and Cloud Gate, the RTX 2070 OC is almost identical to the GTX 1080.
At Cinebench R15, the RTX 2070 weakens and ranks behind the GTX 1070 and GTX 1080.
In the OpenCL test of Geekbench 4, the results show clear differences between the cards. The GeForce RTX 2070 OC hangs with the GTX 1080 36% advantage from. After overclocking, it even reaches 43%. The distance to the GTX 1070 is included 33% and 36% respectively after overclocking.
Even the reference "Superposition" of Unigine differs from more than 25% of the previous generation. For the GeForce GTX 1070 in the normal clock, it is 35%.
The old reference, Ungine Valley, seems more balanced, although the KFA 2 GeForce RTX 2070 OC still has a good lead.
Page 6
Virtual reality benchmarks
Virtual reality benchmarks also had to support the GeForce RTX 2070 OC. We rely on VR Mark and the Steam VR test.
In the Orange Room, the RTX 2070 OC is behind the GTX 1080 and is comparable to the GTX 1070.
The Blue Chamber demands much stronger cards. The RTX 2070 OC is far ahead.
The normal Steam rating ends at 11 and there are no surprises. If you look at the number of frames, the RTX 2070 OC clearly has the advantage.
Page 7
Reference Games
Games are much more versatile than synthetic benchmarks, and here too, the KFA2 GeForce RTX 2070 OC must prove itself.
(average images per second / frames per second, plus it's better)
Overall, the GeForce RTX 2070 OC is similar to the previous ones. It is at least on par with the GeForce GTX 1080 in most games and behaves much better than the GTX 1070.
Page 8
Cooling, temperatures, RGB lighting and power consumption
Mbadive cooling construction with two 90 mm fans
As soon as you have the card in your hand for the first time, you immediately realize that it is much more mbadive and heavy is as expected. After dismantling the cooling structure, you also know why. You have the impression that the builders did not want you to ever remove the cooler!
The two 90mm fans with clear fan blades at the front are bolted and covered with a layer of foam for the radiator. This reduces the transmission of vibrations. The first cover is made of clear plastic and looks pretty chic, but it is mainly used for RGB lighting. The black part below contains everything.
First of all, you have to all screws in the PCI 2-slot metal cover solve. Then you can plastic cover silver screw Remove. Then you can lift these and the disconnect two fan connections and the RGB ribbon cable, Subsequently, the four silver screws with springs to be removed on the back. They are centered and not neglected. Only then will you get the solid radiator block from. The large backplate and front plate are screwed together by many screws and cover just about the entire board, but also GDDR6 memory and voltage transformers.
The main cooler is particularly interesting. It consists of numerous aluminum fins and a huge copper block covering the entire map. Those are the Vapor Champer Technology, which should distribute more quickly the heat lost between the fins located above and better on the whole of the radiator. This quickly shows that the cooler is really expensive by using so much copper and the steam chamber. Judging by the images on the KFA2 website, the same cooler also seems to be used on the GeForce RTX 2080 and RTX 2080 TI.
The cooling fins are aligned at very small distances from each other. This means that the radiator is not designed for low speeds. On the contrary, the waste heat should be dissipated as quickly and efficiently as possible. The fans run faster at high load and are visible even in standby mode in the case of the Rocket ITX. Temperatures were measured with an open lateral part because the maximum clock had to be felt first. The KFA2 GeForce RTX 2070 OC is overclocked at the factory and therefore produces more heat loss than cards with reference clock.
At room temperature of 25 ° C
RGB lighting should not be missed
Compared to the first generation GeForce GTX 1080 EXOC, KFA2 has learned to do this in terms of RGB lighting. The transparent cover diffuses the light and 6 LEDs are inserted by fan hub. They mainly illuminate the fans. In addition, the slogan "What is your game?" Is also illuminated. The lighting makes a clear impression, unfortunately, the LEDs can not be controlled individually. The LEDs are not "addressable" and the individual fans and slogan can not be lit with different colors or combined with effects. The lighting is set under Xtreme tuner. The software has fields for 3 LED zones, but you can only use one. If later individual elements can be controlled, we do not know it until now.
energy consumption
Because of its high factory overclocking, the GeForce RTX 2070 OC is not at the level of the GTX 1070, it is much more similar to the EXOC GTX1080. But even that is good because it allows more energy with a similar consumption. The measured values in the diagram again show the entire system, with the Ryzen 2600 clocked at 1600 MHz at 1.2 volts and the fan fan speed and the CRYORIG C7 also minimized at 10% of the speed.
Page 9
conclusion
The currently most affordable graphics card in the new RTX-20 series is positioned much better on the market than the much more expensive big sisters. the GeForce RTX 2070 OC from KFA2 away from reference and clocks 105 MHz higher than the GeForce RTX 2070 "Founders Edition" at the factory from Nvidia. By the time, it is well ahead of most graphics cards in the market, especially if you have the current price 569 € considered.
The cooling design of the GeForce RTX 2070 OC is really amazing! the large-scale use of copper and a steam chamber certainly not cheap, but offers good reserves for overclocking. The card was stronger than the GeForce GTX 2080 EXOC of KFA2, but beat it in almost all the benchmarks. Thanks to the mbadive radiator, it was generally noticeable, but not disturbing. At full load, it became clearly noticeable.
Ray Runs and DLSS – Nile
What Nvidia announces the most can not be used with the launch of GeForce RTX 2070 cards. Games and benchmarks hardly exist and are not available to consumers. So you have to trust the purchase of Nvidia's RTX-20 voice cards to say that the features will soon be complete.
GeForce RTX 2070 no replacement for GTX 1070
Even though the GeForce GTX 1070 has been clearly suspended, the extra burden of the new generation should not be underestimated. Thus, the GeForce RTX 2070 competes with the GeForce 1080, which costs around € 530. By overclocking, you get even more power without the consumption increases significantly. This gives the GeForce RTX 2070 OC from KFA2 a very balanced impression. The RGB lighting is well implemented, even though it's not very sophisticated.
Looking for every aspect of KFA2 GeForce RTX 2070 OC you can recommend them to users with look for a new graphics card and with a GTX 1080 card have. Here, the GeForce RTX 2070 OC is a better choice if you have the small change available. With a budget of about 400 €, you will always have a better chance of acquiring a GTX 1070 or a Vega 56. Thanks to the exorbitant prices practiced by Nvidia, the graphics cards Vega 56 and Vega 64 become a good alternative.
Source link
