不管後面那些SF
大話怕計數,這根本亂吹
>像是為ChatGPT 提供支援的大型語言模型GPT-3，光是在訓練頭幾天就要消耗10 GWh 的能源
當年GPT-3用的是1萬張A100,按最大功耗250W計算也就是2.5 MWh,跟10GWh差距沒有億點點,也有千點點了
>比歐洲城市一整年耗費的能源還要多六千倍
不知是哪個「歐洲城市」,以香港來說,去年用掉163.3拍焦耳(拍=peta=1000T),相當於45444GWh。換句話說,那「歐洲城市」用電(每3天10/3GWh的六千分之一)是香港226216分之一,如果該「歐洲城市」人均用電量是香港的一半,那該「歐洲城市」人口是……66.3人
大話怕計數,這根本亂吹
>像是為ChatGPT 提供支援的大型語言模型GPT-3，光是在訓練頭幾天就要消耗10 GWh 的能源
當年GPT-3用的是1萬張A100,按最大功耗250W計算也就是2.5 MWh,跟10GWh差距沒有億點點,也有千點點了
>比歐洲城市一整年耗費的能源還要多六千倍
不知是哪個「歐洲城市」,以香港來說,去年用掉163.3拍焦耳(拍=peta=1000T),相當於45444GWh。換句話說,那「歐洲城市」用電(每3天10/3GWh的六千分之一)是香港226216分之一,如果該「歐洲城市」人均用電量是香港的一半,那該「歐洲城市」人口是……66.3人
Quoting official data:
ChatGPT: Energy Consumption in MWh = 1,248 MWh
ChatGPT: Energy Consumption in MWh = 1,248 MWh
50 GWh
The Cost of Training GPT-4
OpenAI has revealed that it cost them $100 million and took 100 days, utilizing 25,000 NVIDIA A100 GPUs. Servers with these GPUs use about 6.5 kW each, resulting in an estimated 50 GWh of energy usage during training.8 Oct 2023
The Cost of Training GPT-4
OpenAI has revealed that it cost them $100 million and took 100 days, utilizing 25,000 NVIDIA A100 GPUs. Servers with these GPUs use about 6.5 kW each, resulting in an estimated 50 GWh of energy usage during training.8 Oct 2023
