Review EVGA GeForce RTX 2070 XC ULTRA GAMING 8GB [08G-P4-2173-KR]

Review EVGA GeForce RTX 2070 XC ULTRA GAMING 8GB [08G-P4-2173-KR]  – EVGA has been one of the faithful manufacturers dedicated to graphic processors of NVIDIA, one of which unlike others have managed to deliver an attractive product, with sufficient coverage and availability to be present in much of the market, markets such as the Chilean which It tends to be lacking in several major manufacturers.

A series of changes have undergone EVGA since the last generations, clearly the changes have been largely linked to their cooling systems. Now in the last generations of video cards we find more robust and aesthetically more attractive cooling systems

Turing architecture.

New transmission multiprocessor (SM).

Turing presents a new processor architecture, Turing SM, offering a dramatic increase in shading efficiency, achieving a 50% improvement in performance delivered by CUDA Core compared to the Pascal generation. These improvements are enabled by two key architectural changes. First, Turing SM adds a new independent integer data path that can execute instructions simultaneously with the floating-point mathematical data path. In previous generations, the execution of these instructions would have blocked the issuance of floating point instructions. Second, the SM memory path has been redesigned to unify shared memory, texture caching and memory caching in a drive. This translates into 2 times more bandwidth and more than 2 times more available capacity for L1 cache in common workloads.

Turing Tensor Cores.

Tensor cores are specialized execution units, specifically designed to perform tensor / matrix operations that are the main computing function used in “Deep Learning”. Similar to the Volta Tensor Cores, the Turing Tensor Cores provide high accelerations for matrix calculations at the heart of deep neural network learning and inference operations. The Turing GPUs include a new version of the Tensor core design that has been improved for inference. Turing cores add new INT8 and INT4 precision modes for inference of workloads that can tolerate quantification and do not require FP16 accuracy. Turing Tensor offers for the first time new artificial intelligence capabilities based on deep learning for gaming PCs with GeForce and Quadro-based workstations. A new technique called Deep Learning Super Sampling (DLSS) is driven by Tensor Cores. DLSS leverages a deep neural network to extract the multidimensional features of the rendered scene and intelligently combine the details of multiple frames to build a high-quality final image. DLSS uses fewer input samples than traditional techniques such as TAA, while avoiding the algorithmic difficulties such techniques face with transparency and other complex scene elements.

Ray Tracing Acceleration in real time.

Turing introduces real-time ray tracing or Ray Tracing that allows a single GPU to render visually realistic 3D games and complex professional models with physically accurate shadows, reflections and refractions. Turing’s new RT Cores accelerate ray tracing and are leveraged by systems and interfaces such as NVIDIA Ray Tracing RTX technology and APIs such as Microsoft DXR, NVIDIA OptiX ™ and Ray Tracing Vulkan to deliver a real-time Ray Tracing experience .

New shading advances:

Mesh shading.

Mesh shading advances the NVIDIA geometry processing architecture by offering a new shading model for the shading stages of vertices, tessellations and geometries of the graphics pipe, which supports more flexible and efficient approaches to geometry calculation . This more flexible model makes it possible, for example, to admit an order of magnitude with more objects per scene, moving the bottleneck of the object lists outside the CPU towards highly parallel GPU mesh shading programs. Mesh shading also allows new algorithms for advanced geometric synthesis and LOD object handling.

Variable Rate Shading (VRS)

VRS allows developers to dynamically control the shading rate, shading only once every sixteen pixels or up to eight times per pixel. The application specifies the shading rate using a combination of a shading surface and a “per-primitive” value (triangle). VRS is a very powerful tool that allows developers to shade more efficiently, reducing work in regions of the screen where total resolution shading would not give any visible benefit to image quality, and therefore, would improve frame rate . Several classes of VRS-based algorithms have already been identified, which can vary the shading job according to the level of detail of the content (adaptive content shading), the speed of content movement (shading adaptable to the movement) and applications of virtual reality, lens resolution and eye position (Foveated Rendering).

Texture and space shading

With texture space shading, objects are shaded in a private coordinate space (a texture space) that is stored in memory, and pixel shaders sample that space instead of directly evaluating the results. With the ability to cache shading results in memory and reuse / resample them, developers can eliminate duplicate shading work or use different sampling approaches that improve quality.

Multiple view representation (MVR)

MVR powerfully extends Pascal’s single stereo step (SPS). While SPS allowed the representation of two views that were common, except for an X offset, MVR allows the representation of multiple views in a single pass, even if the views are based on completely different origin positions or view directions. Access is made through a simple programming model in which the compiler automatically considers the code independent of the view, while identifying the attributes dependent on the view for optimal execution.

Deep learning functions for graphics

NVIDIA NGX ™ is the new neural graphics framework based on deep learning of NVIDIA RTX technology. NVIDIA NGX uses deep neural networks (DNN) and a set of “neural services” to perform functions based on artificial intelligence that accelerate and improve graphics, representation and other client-side applications. NGX employs Turing Tensor Core for deep learning operations and accelerates the delivery of NVIDIA deep learning research directly to the end user. Features include ultra-high-quality NGX DLSS (Deep Learning Super-Sampling), AI InPainting image replacement with sensitive content, very high quality Slow-Mo AI and smooth slow motion, and AI Super Rez intelligent image change resolution .

Deep learning functions for inference

Turing GPUs offer exceptional inference performance. The Turing Tensor cores, together with the continuous improvements in the TensorRT (NVIDIA runtime inference framework), CUDA and CuDNN libraries, allow Turing GPUs to offer exceptional performance for inference applications. Turing Tensor Cores also add support for fast INT8 matrix operations to significantly accelerate inference performance with minimal loss of accuracy. New low precision INT4 matrix operations are now possible with Turing Tensor Cores and will allow research and development in sub 8-bit neural networks.

GDDR6 high performance memory subsystem

Turing is the first GPU architecture compatible with GDDR6 memory. GDDR6 is the next major advance in the design of DRAM GDDR high bandwidth memory. The GDDR6 memory interface circuits in the Turing GPUs have been completely redesigned for speed, energy efficiency and noise reduction, achieving transfer rates of 14 Gbps with an improved energy efficiency of 20% compared to the GDDR5X memory used in Pascal’s GPUs.

USB-C and VirtualLink

Turing GPUs include hardware support for USB Type-C ™ and VirtualLink ™ 4. VirtualLink is a new open industry standard that is being developed to meet the power, screen and bandwidth demands of upcoming VR headsets generation through a single USB-C connector. In addition to easing the inconvenience of configuration present in VR headsets today, VirtualLink will bring virtual reality to more devices.

Specs.

First look.

Because of the robust cooling system, this RTX 2070 model will occupy 3 slots in your expansion area.

The length of the video card is around 26.5 cms.

In a later view we see the much needed backplate for this model, very necessary when considering the robust cooling system of this model.

2x 8 + 6 pin connectors as PCI Express auxiliary power.

Under the cooling system, an aluminum block covers practically the entire PCB by contacting in the most temperature sensitive areas offering dissipation area. In the VRM area we see a differentiated heatsink and slightly thicker than the rest.

The configuration in which the main cooling system is composed is mainly with 6 nickel-plated copper heatpipe, distributed throughout the entire length of the heatsink.

Under the backplate, we can see how the contact of the thermal pads of this are subsequently related to the most temperature sensitive areas. We see how the density of electrical elements populates the large part of the PCB.

The power phases are arranged in an 8 + 2 configuration. The 8 GDDR6 memory chips are manufactured by Micron.

Both the backplate and the block on the PCB have thermal pads, elements that in some cases serve as protection, in this case, they protect and dissipate temperature. In front of all the elements that are seen in the refrigeration area, we see a great effort by EVGA in giving a great focus on keeping the video card fresh.

Testing and Methodology Platform.

• Windows 10 Pro x64 operating system.
• The tests were performed in an environment with a temperature of approximately 25 ° C.
• The platform was used without a cabinet.
• The drivers used for the NVIDIA graphics cards: 418.91 WHQL
• The resolutions of the synthetic tests are the default ones for each of the benchmarks.
• The real test resolutions are 1920 × 1080 and 3820 × 2160 with all the maximum available graphics in each game.

Synthetic tests

Real tests

Overclocking.

In the overclocking section and in view of encountering high potential refrigeration components, we wanted to take this RTX 2070 model at a higher frequency than the one we experienced with the version of ASUS we tested earlier. Adjusting the Clock GPU to 1580 MHz (+ 12%) and memories to 1930 MHz (+ 10%) without modifying voltages or additional parameters.

Temperature.

Undoubtedly, the triple slot in the cooling system is noticeable, with only 62º at full load the GPU stays totally cool, leading fans to only develop a speed of 50% in their fans.

Consumption.

Considering that the TDP of an RTX 2070 is 175W, the consumption of this version of EVGA achieves 189W in load, which reflects that due to the overclocking that it brings by default it would be increasing consumption, however it is not so far from what bring as TDP. When we see the consumption after manually overclocking, it reaches 220W, which in relation to the temperature that was seen in the previous graph, this consumption is dissipated without problem by the large heatsink that this RTX 2070 model brings .

Conclusion.

At the moment an RTX 2070 model is a GPU capable of delivering the best gaming experience in virtually all video game titles that today we can find in ultra quality at 2K resolution and in some cases it could be part for resolutions in 4K.

The design of this EVGA model generates great appeal, but at the same time it can be totally the opposite. The presentation of this specimen is very extreme, it can generate a delight for some eyes, while for others, due to its very aggressive terminations and design characteristics, they could be the opposite of attractive. The material in this case is up to par, and for us when the video card is seen as part of the platform it gives an aspect that attracts a lot of attention, but as always this is a point that each one must analyze, But what we can tell you is that the materials used and the aesthetic finishes generate confidence in their durability.

The construction of the cooling system of this EVGA model achieves great results in the dissipation of this NVIDIA GPU, values ​​that slightly exceed 62º at high load, are very impressive parameters especially when we can that the fans only developed 50% of its maximum speed, speed that at the noise level is almost imperceptible.

The price of this model if it falls slightly above the average of the RTX 2070 models, so to have this robust EVGA model you will have to spend something more than what you had budgeted for this NVIDIA model, it will depend on how Enthusiast be the buyer finally.

Do not forget that this model has a width of 3 slots in the expansion area so you should consider a cabinet and the arrangement of the PCI Express on the motherboard if it also occupies other elements in this area in addition to the video card .

The temperature and overclocking potential offered by these models are undoubtedly the most notable factors, after this it could be the design that breaks the scheme that we are currently used to see on video cards.