Idle graphics GPU makes money
21.5 * 3600 seconds * 24 hours = 1857600 M / day
1857600 / 1024 / 1024 = 1.7715 T / day
a t of computing power / day can dig bitcoin worth 4.03 yuan
1.7715 * 4.03 = 7.14 yuan / day
since 2017, bitcoin has reached new heights. Yesterday (May 23), the value of bitcoin broke through the ￥ 16000 mark, hyping virtual currency, becoming a matter that the general public can participate in, And a little bit of technical level, both fried money, but also began to have the technical content of mining
as we all know, virtual currency has a numerical peak, the more difficult it is to dig. In fact, it's hard to judge whether to lose money or not if we continue to dig. Therefore, in the end how to dig, dig what can be the fastest back to this, this paper studies the above two core issues based on the current market situation
CUDA has two meanings
The first is CUDA core. That is CUDA core, NVIDIA graphics card from Fermi architecture began to use, Kepler architecture is also used. CUDA core is the most core part of the architecture, which is also the most dominant part in quantity. It is actually an ALU, which can also be called stream processor, and is the most basic operation unit of N card. Fermi architecture usually contains 48 cudacores in each group of SM units, while Kepler architecture contains 192 cudacores in each group of SMX. The number and scale of cudacore directly determine the operation scale of the graphics card, and also directly affect the performance of the graphics card. The number in "shaders" of N card in gpu-z is the number of CUDA cores (note only for Fermi and Kepler architectures). The following is a brief illustration of gk104 core, in which each green square represents a CUDA core. If you want to deeply understand the working principle of CUDA core, you also need to understand the architecture principle, which will not be introced here 
the second is CUDA environment, CUDA full name, is a general computing architecture implemented by NVIDIA, and is also a kind of API like OpenCL and directcompute. Based on C language development, especially for GPU development of CUDA core architecture, it contains a variety of advanced general technologies, such as parallel architecture and so on. Program developers can use CUDA platform to achieve general computing of NVIDIA graphics card, and use GPU to participate in large-scale computing
CUDA computing is mostly used in professional fields, but not in consumer market. It's just that the salesperson is playing with this concept
exchange of fire means that two a cards are used at the same time. Generally, it's better to use the same model
n card. Similarly, SLI technology, such as two 1080 cards, will be 666.
wash and sleep with the same card. If you don't want to waste it, you can set up a No.2 machine
it's equivalent to human labor. It's good to participate in activities, but it's not good when it's often heavily loaded
it's a good thing to occupy it occasionally. If it's frequently occupied, it will heat up, and accelerate aging
the occasional occupation will rece the possibility of damp hardware
the CPU idle rate is so high ring the game that it is not necessary to assist the graphics card. The functions of the two are different and cannot be assisted
the high idle rate of the game is e to the poor SIMD performance of the CPU in the past, which leads to the significantly lower floating-point computing capacity than the GPU. Therefore, the GPU is allocated a load suitable for parallel, resulting in the CPU does not have a high parallel load. The older graphics API in the game is also very difficult to deploy the CPU for intensive computing, but the SIMD performance of the CPU has been greatly improved, The application of AVX / avx2 is also relatively mature. Taking advantage of CPU's memory access advantage, even if the theoretical floating-point peak value is not high, it has good adaptability to handle the mixed load of branch, logic and intensive computing
I'm afraid I can't feel the drag of CPU on graphics card in my daily life, but I still need to pay attention to the balance between CPU and graphics card. In addition to games, more and more professional software also has high requirements for CPU
GTX titanx
double way gtx1070
Titan x Pacal: 11T
gtx1080: 8.9t
gtx1070: 6.5T
CPU = central processing unit = CPU
at present, GPU and CPU are divided into two completely different functions. Even if the computing power of GPU exceeds that of CPU, the system can't run on its architecture. It's just a babble to replace it. Supercomputers use a lot of GPUs just to improve their computing power, Its core part is composed of CPU<
APU = accelerated processing unit = Accelerated processor
is a proct that AMD integrates graphics core into CPU, and its graphics core is completely independent
in addition to Intel's core display, Intel hopes that GPU integrated in CPU can share part of bus resources with CPU in the future, or even share physical transistor resources after integration in the future, so that CPU and GPU will be integrated into one
in addition, APU is CPU centralized display, and there is no such saying as check display. Please respect yourself!