Recently, revelations about the NVIDIA RTX 40 series graphics cards have emerged one after another, and many of them have claimed that the new generation has paid a relatively high price for power consumption in order to improve performance.and even the flagship model has more than 600W.
Although the wind of 600W faded later, 450W began to become mainstream, which is still a figure that should not be underestimated.
On the topic of graphics card power consumption, a few days ago, AMD senior vice president Sam Naffziger used a few pages of PPT to do some analysis and introduction.
Sam Naffziger pointed out that the actual computing demand has exceeded the evolution of Moore’s Law. However, AMD has always been very concerned about energy efficiency parameters. For example, the unit performance of the RDNA 3 architecture is 50% higher than that of the RDNA 2 architecture.
Having said that, another PPT shows that since about 2015, the TDP of high-efficiency GPUs has begun to rise dramatically. Around 2023, there will be 700W graphics cards, and this year, there will be 450W and 550W graphics cards. left and right display cards.
Of course, the concept of GPU is very large. In fact, it is not only game cards, but also professional cards, accelerator cards, etc. Although AMD did not map 700W to a specific product this time, it also shows that Moore’s Law is slowing down, that is At the moment when the chip process is “squeezing toothpaste”, in order to respond to customers’ pursuit of performance, at least in the next few years, it will have to sacrifice power consumption to a certain extent.
source of information
Further reading: