Gigabyte RTX 3070 Gaming OC graphics card test, a first custom...

A totally Gaming RTX 3070!

Two days after the enema of the RTX 3070 Founders Edition, the custom versions enter the ball, the opportunity for us to test the RTX 3070 Gaming OC of Gigabyte. The card therefore integrates the range Gaming OC, but offers more contained dimensions than the RTX 3080 and 3090, 28.6 x 11.5 x 5.1 centimeters for the newcomer, against 31.7 x 12.8 x 5.5 centimeters for its large sisters. The RTX 3070 is also satisfied with a 6-pin connector and an 8-pin connector. Visually, we find the OC Gaming label but in a more contained design, the PSB no longer offers an appendage at the top and the three fans have the same size. The RTX 3070 also features WindForce 3X cooling, with graphite nano-lubricant.

Let’s focus on NVIDIA’s new Ampere architecture, it’s the second architecture that natively grants Ray Tracing technology. In September 2018, NVIDIA offered its Turing architecture, which had the right to a refresh during the summer of 2019. Two years after Turing, NVIDIA switched to Ampere and offered a leap in certain characteristics. The RTXs resulting from the Ampere architecture will benefit from an engraving fineness of 8 nm, carried out by Samsung, concretely it would be an optimization of the engraving fineness of 10 nm, specially designed for NVIDIA, which allows a gain As a result of transistors, Turing proposed 18.6 billion at most (12 billion for the previous generation, Pascal), Ampere announces 28 billion. The most visible consequence is the explosion in the number of Cuda Core, the RTX 3080 announces 8704, where the RTX 2080 Ti offered “only” 4352 and the RTX 3090 passes the 10,000 Cuda Core milestone! The other Cores do not suffer the same increase with this new architecture, NVIDIA has preferred to give priority to units dedicated to floating-point number calculations, claiming that these are the most promising in the field of video games. The Tensor Core and RT Core are evolving for their part and NVIDIA says that this will translate into an increase in performance when the Ray Tracing and DLSS technologies are activated.

Generation RTX 3000 will benefit from second generation RT Core, NVIDIA claims they are twice as efficient compared to the previous generation and third generation Tensor Core, NVIDIA specifies that this generation would be 4-8 times more efficient compared to the previous generation. previous generation. These new generations allow NVIDIA not to need to increase the number of RT Cores and even to decrease the number of Tensor Cores, while announcing a performance gain.

The announcement of this new generation was accompanied by the invocation of RTX IO, based on Microsoft’s Directx Storage API, API expected for 2021. The goal is to significantly reduce loading times, the basic finding is simple , compressed game data is conventionally sent to the CPU, in order to be decompressed, then sent to the GPU, so that it is displayed, which constitutes a lot of bottlenecks. The RTX IO opts for a more direct solution, sending the compressed data to the GPU, which will then decompress and display it. This would go from a loading time of 3.8 to 0.5 seconds for an Nvme SSD with a PCie Gen 4.0 interface and from 6.6 to 3.3 seconds for an SSD with a SATA interface. Load times on an HDD would drop from 18.9 seconds to 11.9 seconds.

The new RTX 3080 and 3090 graphics cards from NVIDIA will benefit from a new GDDR6X memory standard, which promises higher operating frequencies and logically increased bandwidth as well. The GDDR6X standard allows the processing of more signals at the same time, 4 against 2 for the GDDR6 standard. The RTX 3070 is “content” to memory the GDDR6 standard, standard which equipped RTX 20×0.

Finally, the Ampere architecture inaugurates the HDMI 2.1 standard, which makes it possible to offer an output signal in 8K 60 images per second, with a single cable. The cards also support acceleration of the AV1 codec to 8K, HDR 60 frames per second.

The Ampere architecture is quite close to that of Turing, we discover large GPC blocks, which include TPC blocks, where the SM blocks are located. Each SM block has Cuda Core, Tensor Core and RT Core. The GA102 in its most successful version, in the RTX 3090 therefore, offers 82 SM, each SM offers 128 Cuda Core, so we end up with 10,496 Cuda Core in total.

NVIDIA also announces:

• ● NVIDIA Reflex : Instantly become more competitive. Gamers strive for the lowest possible latency in competitive games as it allows the PC to respond to their commands faster, allowing them to play with greater accuracy. NVIDIA Reflex is a new suite of technologies that optimize and measure system latency. These include NVIDIA Reflex Low-Latency Mode, a technology built into popular esports games such as Apex Legends, Call of Duty: Warzone, Fortnite, and Valorant, which reduces latency by up to 50%, and NVIDIA Reflex. Latency Analyzer, which detects input from the mouse and then measures the time it takes for the resulting pixels (for example, a gun barrel flash) to change the screen. The Reflex Latency Analyzer is built into the new Esports 360Hz NVIDIA G-SYNC displays, arriving this fall from Acer, Alienware, ASUS and MSI and supported by ASUS ‘leading esports devices, Logitech, Razer and SteelSeries.
• ● NVIDIA Broadcast : Stream like a Pro. Over 20 million live streamers around the world can turn their homes into a broadcast studio with NVIDIA Broadcast, a universal plugin that improves the quality of microphones and webcams through RTX-accelerated AI effects, such as audio noise cancellation, virtual background effects and webcam auto-frame.
• ● NVIDIA Omniverse Machinima : Enrich a new art form. Modern games continue to expand the genre of storytelling, in which game resources are used to create cinematic masterpieces. Omniverse Machinima makes this job easier, providing a path-traced visualization tool and engine designed for physical precision, simulating light, physics, materials and AI. Users can take resources from supported games and use their webcam and AI to create characters, add high-fidelity physics and face and voice animations, and publish cinematic-quality cinematics using the power. rendering of their RTX 30 series GPU.

Introduction The map GPU-Z 3DMark Extreme 3DMark Ultra YEAR 1800 Assassin’s Creed Odyssey Far Cry 5 Gears 5 Ghost Recon Wildlands Grid (2019) GTA V Hitman 2 Red Dead Redemption 2 Shadow of the Tomb Raider The Division 2 Jeux Ray Tracing et DLSS – I Ray Tracing and DLSS Games – II Performance index Consumption and temperature Conclusion