[PConline News] NVIDIA not only has an advantage in the high-end graphics card market, but also has started early in the field of deep learning, AI artificial intelligence and other areas, and AMD has made slow progress in this area. Therefore, almost all GPUs use deep learning and AI operations. Only NVIDIA has chosen. In order to save money, some people and departments chose to purchase GeForce game cards and run deep learning and AI-like operations on game cards. However, the situation has changed in the past few days. NIVIDA has quietly updated its user agreement to explicitly prohibit data centers from using GeForce graphics cards to run deep learning. This way, data center operators can only choose to buy expensive Tesla cards to run. The cost may be more than 10 times before.
This matter has been fermenting for several days. The company that was first warned by NVIDIA was a company in Japan. The boss of Miotsukushi Analytics, the country's boss, wrote an article on wirelesswire website a few days ago and criticized them for suddenly modifying the user. The agreement (EULA) prohibits data centers from using GeForce graphics cards to run deep learning, forcing them to buy expensive Tesla cards. Interestingly, this article described NVIDIA's approach as the rise of the First Order in the Star Wars. Nvidia turned into an evil black warrior.
So what exactly is it saying? Deep learning is a hot area now, but for the average person this concept is a bit far, let's talk about the impact of this matter - some years ago some people tried to use the GeForce graphics card as Quadro professional card, crazy even when Some manufacturers have made jumpers directly, which can make consumer-level graphics cards become professional cards, so that some users can use inexpensive GeForce graphics cards as expensive Quadro professional cards, which can save a lot of money! Now that NVIDIA bans deep learning of GeForce graphics cards, it is similar in nature, because in order to save costs, some operation centers use GeForce graphics cards to deploy deep learning so that they don't need to buy expensive Tesla graphics cards.
This is why NVIDIA's move has stimulated many data center operators. NVIDIA has changed this rule. For them, it is only one doomsday, and the operating cost is greatly increased. However, it should be noted that NVIDIA’s blocking strategy is also very clever. First of all, they banned the data center from using GeForce graphics card to run deep learning, and did not restrict the use of personal, academic teams, so the impact on personality is not, the greatest impact is those A business organization that provides deep learning operations. Second, NVIDIA is not directly blocking the use of GeForce graphics cards in the data center. Their blocking is controlled by the driver, that is, they do not provide driver support, so they do not need the official CUDA tools. Third, NVIDIA blocked the data center from using GeForce graphics card to run deep learning, but the block is not within the limits, which means mining is not affected.
Regardless, NVIDIA's move will leave people with the impression of overbearing and abuse of rights. However, NVIDIA is certainly not immune to the use of consumer-grade graphics cards in the data center to reduce costs and improve profitability. It is necessary to know GeForce consumer-grade graphics cards. The price is quite cheap, TX 1080 is only 499 US dollars, while the Titan X is also only about 1,000 US dollars, but the deep learning special card such as Tesla P100 is expensive, despite no public offer, but earlier broke the analysis of the card price To about 10,000 US dollars, the cost of 10 times the spread is very large for the data center, and it is not surprising that these operators are so opposed to NVIDIA's approach.
This fairy fight has nothing to do with ordinary consumers. Both sides are fighting for their own interests. We can't say who is right or wrong. NVIDIA has always positioned the GeForce graphics card as a game card, not to mention that the data center is used to run deep learning. Even mining is prohibited by them. Officials have expressed their position more than once. When this year's mining hotspots, NVIDIA even stated that it would limit game card mining. Of course, NVIDIA doesn't actually prevent video card users from mining. Their approach is to launch a dedicated mining card with a warranty of only three months and no video output. It is just too normal to block data centers using GeForce graphics cards now.
Stainless Steel Piano Wire,Ss Piano Wire,Stainless Steel Spring Piano Wire,Piano Wire Stainless Steel
ShenZhen Haofa Metal Precision Parts Technology Co., Ltd. , https://www.haofametals.com