Phi mining
Publish: 2021-03-23 22:37:45
1. Because amd graphics card provides a large number of simple computing resources, which is suitable for high-intensity and heavy load computing,
while NVIDIA graphics card provides less computing resources, mainly through software optimization to achieve game acceleration. In addition, game n card simplifies some moles used in scientific computing (mining just belongs to this kind of Computing),
therefore, in the face of mining, which is a simple and high load operation, the N card with weak body will not work
for now, the design between AMD and NVIDIA is more and more similar. The a-card part absorbs the advantages of serial and branch judgment of n-card, and the n-card part also adopts the simple violence stack design of a-card. So who has the advantage is a mathematical proportion problem, whose graphics card is more accurate to the proportion planning of game resources, whose performance is better
so mining depends on the efficiency of the algorithm. A card is really good at an algorithm called SHA-256, which was used in violent decoding before (so a card also has a good performance in violent decoding)
unfortunately, in order to rece the power consumption, n card has actually simplified some scientific computing moles. As we said downstairs, the CPU is good at the work of branch judgment. N card no longer relies on itself, instead, it has been deleted, leaving this part of the work to the CPU to the processor, so the general computing performance is weaker than a card. As for the advantages of N card in the field of professional card for a card mentioned by some people, it mainly depends on the early investment and construction in the application of software in the past, that is, the early optimization is still in place, and now the old customers are still eating well
it is true that Tianhe first used 4870 as accelerator card, but there may be a mistake, that is, it was not NVIDIA's Tesla that was used by new Tianhe, but Titan, another supercomputer, which was not entirely Tesla processor, but also AMD's Haolong processor for collaborative computing. In fact, tianhe-2 uses Intel's Xeon Phi processor, neither amd nor NVIDIA.
while NVIDIA graphics card provides less computing resources, mainly through software optimization to achieve game acceleration. In addition, game n card simplifies some moles used in scientific computing (mining just belongs to this kind of Computing),
therefore, in the face of mining, which is a simple and high load operation, the N card with weak body will not work
for now, the design between AMD and NVIDIA is more and more similar. The a-card part absorbs the advantages of serial and branch judgment of n-card, and the n-card part also adopts the simple violence stack design of a-card. So who has the advantage is a mathematical proportion problem, whose graphics card is more accurate to the proportion planning of game resources, whose performance is better
so mining depends on the efficiency of the algorithm. A card is really good at an algorithm called SHA-256, which was used in violent decoding before (so a card also has a good performance in violent decoding)
unfortunately, in order to rece the power consumption, n card has actually simplified some scientific computing moles. As we said downstairs, the CPU is good at the work of branch judgment. N card no longer relies on itself, instead, it has been deleted, leaving this part of the work to the CPU to the processor, so the general computing performance is weaker than a card. As for the advantages of N card in the field of professional card for a card mentioned by some people, it mainly depends on the early investment and construction in the application of software in the past, that is, the early optimization is still in place, and now the old customers are still eating well
it is true that Tianhe first used 4870 as accelerator card, but there may be a mistake, that is, it was not NVIDIA's Tesla that was used by new Tianhe, but Titan, another supercomputer, which was not entirely Tesla processor, but also AMD's Haolong processor for collaborative computing. In fact, tianhe-2 uses Intel's Xeon Phi processor, neither amd nor NVIDIA.
2. First of all, applications that can be implemented with CUDA can be transplanted to OpenCL< However, OpenCL has many advantages that CUDA does not have, and it is supported by a wide range of manufacturers:
Desktop GPU: NVIDIA, amd
Desktop, notebook and server CPU: Intel, AMD, arm
server coprocessor: NVIDIA Tesla series, Intel Xeon Phi
mobile processor: Mobile CPU, mobile GPU (with Qualcomm snapdragon, Samsung exynos, NVIDIA Tegra et al.
FPGA: Xilinx, Altera
briefly talk about the applications I know
digital image processing
Computer Vision
machine learning
augmented reality
virtual reality
medical image processing
large scale data analysis
bitcoin mining
biological computing
Computational photography
physical simulation
numerical analysis
camera pipeline accelerator
Desktop GPU: NVIDIA, amd
Desktop, notebook and server CPU: Intel, AMD, arm
server coprocessor: NVIDIA Tesla series, Intel Xeon Phi
mobile processor: Mobile CPU, mobile GPU (with Qualcomm snapdragon, Samsung exynos, NVIDIA Tegra et al.
FPGA: Xilinx, Altera
briefly talk about the applications I know
digital image processing
Computer Vision
machine learning
augmented reality
virtual reality
medical image processing
large scale data analysis
bitcoin mining
biological computing
Computational photography
physical simulation
numerical analysis
camera pipeline accelerator
3. It's best to default. Otherwise unstable, if you memory 2333 or above, you can modify the line overclocking
Hot content