Welcome to ShenZhen Modern Chip Mall Internet Technology CO., LTD | Register

Home > Industry Information > AMD Will Increase FPGA And AI Chip Output

AMD Will Increase FPGA And AI Chip Output

Auth:现代芯城元器件采购平台 Date:2023/9/11 Source:现代芯城元器件采购平台 Visit:1109 Related Key Words: AMD will increase FPGA and AI chip output

AMD will increase FPGA and AI chip output


    Italian technology website chips ' n bits believe,AMD  It will sacrifice the output of TSMC's next -generation Radeon game GPU (RX 8000) to increase the output of FPGA and GPGPUs.
    This may be more or less can be attributed to allocating more capacity to artificial intelligencechipNotGraphics cardEssenceThis has also promoted people's concerns about the supply of GPU, which is not unreasonable.
    With the upsurge of cryptocurrencies, and then epidemic and current artificial intelligence, gamers are indeed very difficult to live.During the epidemic, the price of graphics cards soared to the real ridiculous level in about 18 months.Even now, the value of many current graphics cards still looks bad, although AMD's new Radeon RX 7800 XT is helpful.
    At the same time, we have seen reports about AMD canceling its next -generation RDNA 4 series high -end model, which may be called the Radeon RX 8000 series.In fact, some sources reported that NVIDIA has actually stopped the current RTX 40 GPU production and has changed to produce more H100 AI chips that can make money.So this has happened.
    at least, this is a hint.But is it really like this?If so, how important is this?There is no doubt that NVIDIA uses H100 GPU to create artificial intelligence reasoning.There is no doubt that NVIDIA will favor TSMC's production capacity to produce those chips with higher profits.TSMC is a Taiwan chip foundry, producing all GPUs of AMD and NVIDIA.
    TSMC's production capacity is limited, NVIDIA cannot currently meet the needs of H100.Therefore, every GPU for games can eat more profitable profits that artificial intelligence chips can bring.
    however, if you check it online, you can buy any NVIDIA RTX 40 GPU without difficulty.From the entry -level RTX 4060 weak to the RTX 4090 monster, they are widely available.
    Of course, some people think that the price of RTX 40 is so high because of the vigorous development of artificial intelligence.This may have some impact, but at best it is just indirect.AI devices will not purchase the RTX 4060 Ti and RTX 4070  MotherboardThis did happen during the peak of the cryptocurrency mining boom.
    therefore, the supply and demand of the current game graphics card does not have serious distortion.But if the artificial intelligence boom continues, there will be almost a change in priority.
    Earlier this year, we reported that Nvidia's own roadmap shows that its next -generation RTX 50 GPU has been postponed to 2025.Considering the normal GPU cycle, you expect RTX 50 cards to appear next year.
    It is very likely that at least part of the reason is that TSMC's upcoming 3N node production capacity is more used for more profitable artificial intelligence chips.By 2025, TSMC should have established a large number of 3N production capacity. Perhaps the launch of the increasingly infamous Arizona Worse Factory enables NVIDIA to meet the needs of artificial intelligence while producing games GPU.
    If so, AMD will undoubtedly have a similar process.It wants to participate in artificial intelligence games, if it is 146 billionTransistor  Instinct Mi300 artificial intelligence beasts have received attention, it will definitely be affected by the same market forces as NVIDIA.
    but things are like this.If the facts prove that artificial intelligence chips occupy the capacity of all wafers in all wafers within an foreseeable time, then this will be a sexual blow.The GPU of the game can be returned to a node behind the AI chip, or a cycle is suspended according to the way you want, and then moved forward.
    As early as 2014, NVIDIA had taken a very good step on the same node as the same generation of GPUs as the Maxwell.They retain the same 28 nanometer production technology as Kepler GPU.However, the attention of Maxwell's attention to design and efficiency still brings the performance improvement you usually expect to be reduced through nodes.
    in other words, even if the game GPU is forced to return a node to allow more favorable chips to occupy all production capacity, this sexual strike does not actually miss the faster GPU of the new generation.
    It is important that continuous artificial intelligence boom does not mean that the GPU pricing will reappear the absurd trend.Therefore, things are not ideal.The graphics card is still too expensive, and the prosperity of artificial intelligence is not available.But there may be no reason to panic.

Industry Information

Product Index :