Graphics computing power 1080ti2060

1. The computing power of the project, gtx1080ti, is 6.1 / 3.7, about 1.64 times that of K80
at present, the popular cards for deep learning are rtx2080ti and rtx2070 (multi-channel), and it is not easy to buy more than one of them when gtx1080ti is delisted Second hand another)
* for the deep learning of CUDA platform, the graphics card mainly focuses on: single precision floating-point operation, video memory, tensor core (Turing architecture and volt architecture, RTX series and Titan V)
* in terms of the main stability and some special functions of Tesla, double precision (currently this deep learning is less used), single precision and half precision floating-point operation have little advantage, It's expensive (if you want to surpass gtx1080ti, you need tens of thousands of Tesla V100)

2.

From the overall performance data analysis, considering the unified drive problem, gtx1080 is 5% - 10% better than rtx2060. From other aspects, the details are as follows:

1, rtx2060 and gtx1080 have different types of graphics cards, rtx2060 belongs to the fever level graphics card, gtx1080 belongs to the mainstream level graphics card

The frequency of rtx2060 and gtx1080 are 7010mhz and 8008mhz respectively

The maximum resolution of rtx2060 is 4096 × 2160, gtx1080 graphics card maximum support 7680 × 4320

4. Rtx2060 3D API supports DirectX 11.2, gtx1080 3D API supports DirectX 12 / OpenGL 4.5

extended information:

about VR:

gtx1080 and rtx2060 are very close, but there are many new Pascal technologies and lower power consumption, so gtx1080 will be a little bit higher than rtx2060

e to the limitation of VR devices, as long as the graphics card reaches the standard line of VR, it can play smoothly, and gtx970 is the standard line of N card, so gtx1060 can also run HTC vive and oculus rift without pressure

in addition, Pascal core has its own VR black technologies (Ansel, SMP, VR works audio, VR touch, etc.), so gtx1060's VR additional experience will be better than gtx970 / 980

3.

The graphics card 2060 is good

In terms of graphics card parameters, rtx2060, as the next generation version of gtx10606g, has obvious upgrade in terms of parameter comparison. Mainly in architecture, process, stream processor, texture unit, ROPS, tensor cores, video frequency / bit width / bandwidth, video type and so on

e to the small difference between gtx1060 and rtx2060 in mainstream 1080p image quality, the main test is the frame rate difference between 2k and 4K Ultra HD image quality

extended data:

precautions:

1. Slot. Check whether the motherboard supports PCI-E and AGP at the same time. If only one is supported, select the appropriate type of graphics card

2. Resolution: the resolution of the graphics card is not the higher the better, but should correspond to the highest resolution of the computer display used by the user, if the display only supports 1280 × 1024, so it's not worthwhile for users to invest more money in high-resolution graphics cards

3. Try to buy the procts of large companies with R & D capabilities, because these manufacturers will never use immature public board design, and will improve their circuit layout and materials to make them more stable, but often the procts come to market later

4. Rtx2060 with ray tracing, gtx1080ti without ray tracing, these two graphics cards are better than 1080ti
rtx2060 is better than 1060, and dx12 is better than 1070.

5. At present, there is no 2060ti graphics card, only 2060 and 2060super graphics card. Gtx1080ti 11g graphics card is available. In terms of two performances, it must be 1080ti, the flagship graphics card of the previous generation, which is equivalent to the position between rtx2070 and 2080. It is completely strong for rtx2060super. However, 1080ti has been discontinued. It is better to use 2070super graphics card

6. No, this generation of rtx2000 series does not have the gtx900-gtx1000 to improve. Rtx2080 is barely better than gtx1080ti. Rtx2060 can only reach the level of gtx1070ti, which is far from gtx1080ti

7. It depends on the price you can buy. These graphics cards are divided into different 3g6g, and you can choose according to the price

8. If the recommended budget is enough, go to 3070 and buy new instead of old. If not, consider 2070s. In terms of performance, 1080ti is basically the same as 2070s, which is better than 2070. However, 1080ti has been out for a long time, so it's not recommended to buy. 2070s is recommended

9.

The main difference between the two is the chip. 1080ti is not an upgrade to 1080. The specific differences are as follows:

1. They are both 16nm process and Pascal architecture. The difference is that GTX 1080 is the core of gp104-400 and GTX 1080 ti is the core of gp102-350. The core area of GTX 1080 ti is 50% larger than that of GTX 1080, the number of CUDA cores is nearly 67%, and the grating and texture units are nearly 40% higher

2. GTX 1080 ti is not as high as GTX 1080. Its basic frequency is 1480mhz and boost frequency is 1584mhz, which is about 9% lower than GTX 1080. In terms of video memory, GTX 1080 Ti and GTX 1080 are both gddr5x. However, the capacity, frequency and bit width of video memory have increased significantly, which also makes the video memory bandwidth of GTX 1080 Ti reach an amazing 484gb / s, which is 4GB / s higher than that of Titan x Pascal< 3. Different power consumption. GTX 1080 ti is rated at 250W, 70W more than GTX 1080. Therefore, 8 + 6 pin external auxiliary power supply is required to provide 225W (150 + 75) in addition to PCI-E slot 75W< 4. Different prices. GTX 1080 ti is sincere, but it is only 400 yuan more expensive than GTX 1080 in terms of specifications. In addition, Titan x Pascal, which is similar to GTX 1080 Ti, is 4000 yuan more expensive. However, after GTX 1080 Ti was listed, NVIDIA chose to rece the price of GTX 1080 by US $100, which is also a gap with GTX 1080 Ti

the specific data are as follows:

extended data

NVIDIA (NASDAQ Stock Code: NVDA) is an intelligent computing company. Founded in 1993, the company is headquartered in Santa Clara, California, USA. Jensen Huang is the founder and CEO

In 1999, NVIDIA invented GPU, which greatly promoted the development of PC game market, redefined modern computer graphics technology, and completely changed parallel computing

on February 8, 2018 Pacific time, NVIDIA announced its fourth quarter and full year results. According to the financial report, the annual revenue of NVIDIA hit a record high last year, reaching US $9.71 billion, with a year-on-year growth of 41% and a profit growth of 83%. NVIDIA's annual GAAP earnings per share reached $4.82, up 88% year on year. In the past fourth quarter, NVIDIA's revenue reached $2.91 billion, up 34% year on year. GPU revenue increased 33% year on year to $2.46 billion

in June 2017, he was selected into the "top 50 Smartest Companies in the world" list of MIT Technology Review in 2017