Keywords: artificial intelligence
Fierce competition in the field of artificial intelligence in the two giants - nvidia and Intel are busy crazy SPAR verbally.
Last week, Intel (Intel) in Intel Developer Forum (IDF) annual developers conference announced that it launched in 2017 specially designed for deep learning of artificial intelligence and the latest generation product, Intel Xeon Phi processor, codenamed Knights Mill, according to its operation ability more than two times faster than rival products, directed at NVIDIA (NVIDIA).
Nvidia also not to be outdone. On IDF meeting held on the same day, nvidia released a "powerful" blog, post title is clear indication position: the correct Intel on deep learning benchmark ".
Nvidia's article, Intel lost facts in deep learning benchmark, Intel said in its latest Xeon Phi processor speed deep learning performance when using the outdated data, it also used the in depth study performance test Kepler GPU nvidia many years ago old products series, this belongs to intentionally mislead consumers, without giving the GPU performance a fair evaluation. And these two benchmark "mistake" fix up is easy.
Intel now delve into the depth of study, this is very good. When we step into artificial intelligence era, this is the most important change of operation, a deep learning is important, that cannot be ignored. However, they shall examine the facts involved.
Nvidia vice President in charge of accelerating deep learning Ian Buck also added fuel to the fire, along with the deep learning system and the rapid change of artificial intelligence, the new comer in this field may not be aware of the aspects of hardware and software change all of the development."
Soon, Intel data center group vice President Jason Waxman in recently published a blog post hit back, said "public debate available performance benchmark is a waste of time." He said firmly defend their own company's test data:
Artificial intelligence is a new thing. As a catalyst for innovation, broaden the field of artificial intelligence Intel's leadership will be crucial.
Although a lot of people are talking about gpus now machine deep learning value, but in fact, last year for deep learning and research and development of the processor, the proportion of using the GPU is less than 3%.
The nvidia is a global leader in the GPU production "eldest brother". So, Jason Waxman implying object clear but anymore.
And better known than CPU and GPU have what advantage? GPU Chinese translation for the graphics processor, its performance is excellent in terms of visual processing, luminescent spot lies in its strong ability of parallel computing, parallel processing a large number of trivial information. In addition, the GPU processing speed is very fast. The best of the domestic hardware of artificial intelligence research project "Cambrian" team the latest research results show that the GPU can provide an average of 58.82 X times the CPU speed.
Nvidia CEO jen-hsun huang said, using the GPU massive processing power to learn artificial intelligence algorithm is a new kind of calculation model. Analysts believe that with the development of artificial intelligence and deep learning, GPU and some chips have reverse the potential CPU (central processing unit).
In the field of computer processors, the rapid rise of Intel for nvidia under pressure. The old CPU manufacturer in busy playing spat at the same time also took action, paid $400 million for deep learning technology company Nervana Systems, the goal is to achieve its expected 2017 deep learning accelerator chips, to challenge the nvidia in artificial intelligence market dominant position. At that time, if chip performance as expected, Intel is expected to be in the field of deep learning processor beyond the nvidia GPU.
Nvidia in the company's overall profit and deep learning progress quickly, the second quarter, the company net profit of $253 million, compared with $26 million, up 873%; Data center business revenue of $151 million, more than doubled over the previous year, of which a large part of the contribution is deep learning system. Apparently Wall Street is approved for nvidia development strategy. Nvidia current share price is in its historic peak level.
Intel's second quarter net profit, by contrast, are affected by the drag of restructuring write-downs of $1.4 billion in "down", to $1.3 billion. The giants in the transformation of the two growth point of profit, data center and business income of the Internet of things are not as good as expected, the data center of net operating income of $4.03 billion, market expectations of $4.16 billion.
Obviously, the two giant Intel and nvidia competition in the field of artificial intelligence is far from over, and the recent spat whether is the final say.
