As a result of the rise of artificial intelligence, NVIDIA has gained revenue and stock price growth with its GPU in the field, and Intel is questioned that it may not be able to meet the challenge of NVIDIA in the field, but Microsoft has recently adopted Intel's FPGA chip as a real-time AI cloud platform Project Brainwave's DPU, seems to show that Intel is expected to change the disadvantage.
NVIDIA Challenge Intel
NVIDIA's GPU has the advantage in the field of artificial intelligence, is that it has more than Intel's CPU more parallel computing, low-precision computing and other performance, due to this advantage most of the current use of NVIDIA's GPU training neural network.
With this advantage of GPU, NVIDIA began to enter the server market. In the current server market, Intel's X86 processor occupies an absolute advantage in market share, gaining 99% market share. NVIDIA to enter the server market is the choice of ARM architecture, with the hope that the ARM architecture has a low power, low cost advantage to challenge Intel in the server chip market monopoly.
NVIDIA's fiscal first quarter results show that its data center business revenue reached $ 409 million, an increase of 186%, showing a strong offensive. NVIDIA also acquired the acquisition of ARM's Softbank support, in May this year, Softbank holds NVIDIA 4.9% stake to become its fourth largest shareholder.
Softbank after the acquisition of ARM is expected to monopolize the mobile market to the server chip market development, expected by 2020 to occupy the server chip market share of 20%, in addition to NVIDIA, Qualcomm, AMD has released their ARM architecture server chip The
But in the situation under a good, NVIDIA in the field of artificial intelligence also has its short board, in dealing with the delay CPU CPU more advantages than the GPU, GPU processing is the primary goal of computing and data throughput, and CPU internal transistor The primary purpose is to reduce the processing delay, which is a major weakness for the GPU for industries with high latency requirements for autopilot.
Intel's efforts to change the vulnerability in the field of artificial intelligence
Intel lacks its own GPU, in order to change the disadvantages in the field of artificial intelligence it has acquired FPGA maker Altera. FPGAs are originally semi-custom programmable circuits developed from application-specific integrated circuits that have very high flexibility and adaptability by generating different circuit functions with different programming data, so it can be used as a special The task reprogrammable chip is used in machine learning.
Compared to the advantages of the GPU is that the FPGA in the factory can be in the hardware circuit, the corresponding program, change the FPGA hardware structure so that FPGA can support the corresponding artificial intelligence operation, so the efficiency will be higher than the GPU. Google's TPU is also similar to FPGAs, a chip tailored for its deep learning language Tensor Flow, except that the TPU can no longer be programmed to change, and the FPGA can therefore have greater flexibility.
Intel also acquired Nervana Systems, a deep learning technology provider, an artificial intelligence startup that is said to have developed a faster processor than NVIDIA's top-of-the-line graphics processor when running neural networks The performance.
Intel hopes to integrate its own server chip Xeon processor and Nervana's accelerated processor in the field of artificial intelligence with more performance while making it easier to program it with its acquisition of Altera to provide a programmable universal AI chip Is interlinked.
Now Microsoft introduced Intel's FPGA chip for its real-time AI platform, which means Intel in the field of artificial intelligence may be a breakthrough. Synergy Research said Microsoft is the world's second largest cloud platform occupies 11% of the market share, Amazon is the largest cloud platform has 34% market share, but the two companies in the cloud business in the near future cooperation Both companies want to package artificial intelligence services in cloud services to better serve other companies.
Intel's FPGA chip is used in Microsoft's second largest cloud platform in the world, if achieved good results will undoubtedly play a great role in demonstrating, Intel in the artificial intelligence market to win more market share, which for NVIDIA to Said it was not good news.
Intel's global market share of the server chip market is also conducive to its advantage in the field of artificial intelligence, artificial intelligence needs large data, and Intel's server chip is widely used in the data center, but previously because of Intel lack of GPU and difficult NVIDIA contends that in the acquisition of Altera and Nervana it is making up this short board.
In contrast, although NVIDIA hopes to develop ARM-based server chip, but its current server chip market compared to Intel is still too weak, ARM is also hoping to win a limited market share by 2020, NVIDIA in the Server chip market is more difficult to challenge Intel. The situation seems to have begun to favor Intel's direction.