Where the hell is it? NV New Flagship GTX 1080 detailed interpretation

Last night 21 o'clock, NVIDIA GeForce GTX 1080 formally unveiled all the veils. How strong is this card? Is it really faster than Titan X? What are the technical improvements to the new graphics card? This article will answer for you.

NVIDIA GeForce GTX Founders Edition

Unmatched gaming performance

The important thing to say first. The GeForce GTX 1080 is based on the more mature 16nm FinFET process, built with Nvidia's latest Pascal GPU architecture. It is built with 7.2 billion transistors and has 2,560 cuda cores. to maximize performance, NVIDIA engineers optimize every detail of the circuit to ensure that every watt of power can perform its maximum performance. As a result, the GeForce GTX 1080 can operate stably at a 1607MHz UHF rate.

At the same time, we have enhanced GPU auto-overclocking performance and launched GPU Boost 3.0 technology. Its overclocking amplitude will no longer be a fixed value, but will run under the core voltage, the core can safely withstand the approximate highest frequency. we give a typical GPU boost frequency of 1733MHz, in fact, it can reach 1800MHz, or even higher.

Based on the test results, theGeForce GTX Founders Edition has a game performance up to 1.25 times times Titan X, substantially more than the GeForce GTX 980 Ti, and is similarly slightly above the GeForce GTX 980 2-way SLI.

In the "blockade of the whole", "GTA5", "Crysis 3", "Tomb Raider: Rise", "Wizard 3: Wild Hunting" and other games, with 4 K resolution, the highest picture quality settings, GeForce GTX Founders Edition are significantly superior to Titan X, performance is excellent.

1.7 times times the memory performance

As we all know, in GeForce GTX 1080 above, we use the latest gddr5x memory, capacity 8GB, bit width 256bit. It has an ultra-high operating speed of up to 10Gbps, with a memory bandwidth of up to 320gb/s, far exceeding the 224gb/s of the GTX 980.

At the same time, we have enhanced the lossless memory compression technology since the Maxwell architecture to achieve a more efficient 2:1 compression ratio, as well as new 4:1, even 8:1 compression performance, reduced memory data writing, as well as video memory and L2 cache transmission, resulting in more memory performance.

Uncompressed splash screen

Uncompressed splash screen

The pink part is the area that can be compressed in the picture, the left is the Maxwell architecture, the right is the Pascal architecture

VR performance of twice times Titan X

The VR performance of the GeForce GTX 1080 is twice times that of Titan X, combined with a lower power consumption of the graphics card, and VR performance consumes three times times more power than Titan x!

We have added a hardware unit called "Simultaneous multi-projection (synchronous multi-projection)". It can be used in multi-screen splicing, curved screen and other surround features of the display device, for the user (the position of the eye, what we call "point of view") to provide the correct surround image, to avoid the traditional plane image projection to the curved surface of the bending and abnormal distortion occurs. It simulates up to two viewpoints, and the exact picture of up to 16 different angles per viewpoint, without the need to re-render each image.

As a result, when we apply this technology to VR, we can render the left-eye image at the same time, accurately projecting the right eye image, to avoid the previous GPU in the VR application of two times rendering.

One render, the left and right eye screen simultaneously generated

In addition, in VR mode, the GPU in the process of projecting a picture for each eye (viewpoint), by dividing the viewpoint screen into 4 different regions, we can directly get a picture with the approximate distortion effect, which can be applied to the lens of VR helmet directly after simple processing. The unnecessary rendering waste is further reduced.

Approximate shape of VR screen is obtained by rendering four quadrilateral regions

The performance gains from architecture and process advantages, combined with the special improvements brought by simultaneous multi-projection, make the GeForce GTX 1080 show unprecedented ultra-strong VR performance, twice as much as Titan x!

More excellent VR experience

In addition to VR performance enhancements, Nvidia has introduced new vrworks technologies, including Vrworks Audio, and the integration of PHYSX technologies that can simulate real-world physical feedback.

In the past, the sound of the game is often straight, there is no reflection of sound waves, absorption and other phenomena, sounds very direct, rough. With Vrworks audio technology, developers will be able to track the propagation of sound waves in virtual scenes, and apply the results to sound rendering compositions, resulting in sound changes and echoes.

Sound retransmission is more real, reflective, and absorbed.

In addition, we are familiar with the Nvidia PhysX, Flex, Destruction, Hairworks, Flow, cloth and other technologies, are also used in the vrworks inside. With Nvidia's newly released VR Funhouse application, you will be able to fully experience the tremendous efforts Nvidia has made towards VR.

NVIDIA Vrworks is supported by a wide range of developers

Asynchronous computation

Today, GPUs are no longer pure image rendering, especially in the DirectX 12 era, where GPUs are often given more work, such as computational tasks and so on.

The NVIDIA GeForce GTX 1080 supports asynchronous computing, which enables fast and efficient multitasking in parallel, dynamic load balancing, and pixel-level task "preemption" for greater operational efficiency.

In static mode, when the GPU handles both the graphics rendering and the calculation of two tasks, the completion time of the two tasks is often different. As the static partitioning shows, the graphical task is completed at 60% time, and the remaining 40% time waits for the calculation task to complete before the next task can be performed. This will cause a lot of GPU idle conditions to occur.

In the dynamic load Balancing mode (Balancing section), the graphics task takes the lead, and the GPU resources used for this part can take over the rest of the computation immediately and shorten the task completion time.

Preemption (preemption) refers to the ability of the GPU to turn off less important work and prioritize important work so that important work can be done quickly. Pascal is also the first GPU architecture to support pixel-level preemption, which allows the GPU to preempt faster and with less performance overhead.

Other important features

PlayReady 3.0 (SL3000) Specification: Support the highest level of streaming media technology, high-quality 4K online video!

HDR: Two-fold visible color, more saturated color output, 10,000:1 contrast, support for HDR games, HDR video playback ([email protected] 10/12b HEVC decoding), HDR video/streaming ([email protected] 10b HEVC , streaming requires the use of Shield consoles) and supports the latest HDR TVs.

Fast Sync: A new vertical synchronization technology, when the game is rendered at a higher frame rate, Fast sync only transmits the frame that is closest to the display refresh time to the display, avoiding the display tearing caused by excessive frame-screen signals.

High-performance SLI Bridge: delivers higher GeForce GTX 2-way SLI productivity and more stable performance. As for 3-way and 4-way, it will also be available in some ways, more suitable for extreme gamers.

GeForce GTX 2-way SLI

Finally, with the release of GeForce GTX 10,801, there is also a tool called "Ansel". By collaborating with game developers to incorporate Ansel into the game, you'll be able to freeze the game scene at any time, save screenshots at any camera location/perspective, and get a super-clear picture of up to 32 times times the maximum screen resolution. In addition, Ansel supports filters, EXR format, 360-degree scene capture, and more, and is very powerful.

Where the hell is it? NV New Flagship GTX 1080 detailed interpretation