{"id":1108583,"name":"Computational performance per dollar","unit":"FLOP/s/$","createdAt":"2025-10-14T11:42:43.000Z","updatedAt":"2025-10-22T19:13:51.000Z","coverage":"","timespan":"","datasetId":7234,"columnOrder":0,"shortName":"comp_performance_per_dollar","catalogPath":"grapher/artificial_intelligence/2025-10-10/epoch_gpus/epoch_gpus#comp_performance_per_dollar","descriptionShort":"Hardware computational performance shown in [floating-point operations](#dod:flop) per second (FLOP/s) per US dollar, adjusted for inflation.","descriptionProcessing":"- Reporting a time series of AI investments in nominal prices (i.e., without adjusting for inflation) means it makes little sense to compare observations across time; it is therefore not very useful. To make comparisons across time possible, one has to take into account that prices change (e.g., there is inflation).\n- It is not obvious how to adjust this time series for inflation, and we debated it at some length within our team.\n- It would be straightforward to adjust the time series for price changes if we knew the prices of the specific goods and services that these investments purchased. This would make it possible to calculate a volume measure of AI investments, and it would tell us how much these investments bought. But such a metric is not available. While a comprehensive price index is not available, we know that the cost for some crucial AI technology has fallen rapidly in price.\n- In the absence of a comprehensive price index that captures the price of AI-specific goods and services, one has to rely on one of the available metrics for the price of a bundle of goods and services. In the end we decided to use the US Consumer Price Index (CPI).\n- The US CPI does not provide us with a volume measure of AI goods and services, but it does capture the opportunity costs of these investments. The inflation adjustment of this time series of AI investments therefore lets us understand the size of these investments relative to whatever else these sums of money could have purchased.","type":"int","grapherConfigIdETL":"0199e287-a105-7d33-bc58-f2949ad90e99","dataChecksum":"10940366440225583694","metadataChecksum":"-34971804681253011","datasetName":"Machine Learning Hardware","updatePeriodDays":365,"datasetVersion":"2025-10-10","nonRedistributable":false,"display":{"unit":"FLOP/s/$","zeroDay":"2000-01-01","yearIsDay":true,"numDecimalPlaces":0},"schemaVersion":2,"processingLevel":"major","presentation":{"topicTagsLinks":["Artificial Intelligence"]},"descriptionKey":["Measures how much computing power is delivered per dollar when purchasing a GPU.","GPUs (graphics processing units) are specialized computer chips that can perform many calculations simultaneously, which makes them widely used for training AI systems.","Only includes GPUs that have been used to train notable AI models or are specifically designed for machine learning tasks. Excludes other processor types like TPUs and CPUs.","Performance is measured using AI-specific benchmark tests; prices are the launch prices when GPUs first became available.","Values are shown as FLOP/s per dollar—meaning how many floating-point operations per second the GPU can perform for each dollar of its price. All prices are adjusted for inflation to 2024 US dollars using the Consumer Price Index.","Performance is measured using FP32 (32-bit floating-point), a standard precision level that represents numbers to about 7 decimal places. This ensures consistent comparisons across different hardware, though newer GPUs often support faster but less precise formats.","As this metric has improved over time, the cost of training AI systems has decreased."],"dimensions":{"years":{"values":[{"id":3089},{"id":3965},{"id":4699},{"id":4798},{"id":4808},{"id":4891},{"id":4952},{"id":5074},{"id":5162},{"id":5375},{"id":5434},{"id":5554},{"id":5792},{"id":5939},{"id":5991},{"id":6100},{"id":6118},{"id":6246},{"id":6247},{"id":6278},{"id":6305},{"id":6381},{"id":6391},{"id":6544},{"id":6660},{"id":6799},{"id":6837},{"id":6891},{"id":7439},{"id":7549},{"id":7583},{"id":7772},{"id":7982},{"id":8111},{"id":8123},{"id":8298},{"id":8342},{"id":8372},{"id":8659},{"id":8713},{"id":8740}]},"entities":{"values":[{"id":309878,"name":"NVIDIA GeForce GTX 280","code":null},{"id":309821,"name":"NVIDIA GeForce GTX 580","code":null},{"id":309908,"name":"NVIDIA Tesla K20X","code":null},{"id":309986,"name":"NVIDIA Tesla K20c","code":null},{"id":309962,"name":"NVIDIA GeForce GTX TITAN","code":null},{"id":310029,"name":"NVIDIA Quadro K4000","code":null},{"id":309895,"name":"NVIDIA GeForce GTX 780","code":null},{"id":309844,"name":"NVIDIA Quadro K6000","code":null},{"id":370015,"name":"NVIDIA Tesla K40s","code":null},{"id":372226,"name":"NVIDIA Tesla K40t","code":null},{"id":370016,"name":"NVIDIA GTX Titan Black","code":null},{"id":309910,"name":"NVIDIA GeForce GTX 980","code":null},{"id":370020,"name":"NVIDIA Tesla K80","code":null},{"id":309845,"name":"NVIDIA GeForce GTX TITAN X","code":null},{"id":370023,"name":"NVIDIA M40","code":null},{"id":370018,"name":"NVIDIA P100","code":null},{"id":309693,"name":"NVIDIA GeForce GTX 1080","code":null},{"id":370034,"name":"NVIDIA P40","code":null},{"id":309694,"name":"NVIDIA Quadro P5000","code":null},{"id":309712,"name":"NVIDIA Quadro P6000","code":null},{"id":309783,"name":"NVIDIA Quadro P4000","code":null},{"id":372176,"name":"NVIDIA Quadro P600","code":null},{"id":309705,"name":"NVIDIA GeForce GTX 1080 Ti","code":null},{"id":309708,"name":"NVIDIA TITAN Xp","code":null},{"id":372180,"name":"NVIDIA Tesla V100 PCIe 16 GB","code":null},{"id":372220,"name":"NVIDIA Tesla V100 SXM2","code":null},{"id":370025,"name":"NVIDIA Titan V","code":null},{"id":372221,"name":"NVIDIA Tesla V100 PCIe 32 GB","code":null},{"id":372222,"name":"NVIDIA Tesla V100 SXM2 32 GB","code":null},{"id":309724,"name":"NVIDIA Quadro RTX 5000","code":null},{"id":309748,"name":"NVIDIA Quadro RTX 6000","code":null},{"id":372223,"name":"NVIDIA GeForce RTX 2080 Ti 11GB","code":null},{"id":309695,"name":"NVIDIA Quadro RTX 4000","code":null},{"id":372240,"name":"NVIDIA A100 SXM4 40 GB","code":null},{"id":309738,"name":"NVIDIA GeForce RTX 3080","code":null},{"id":309752,"name":"NVIDIA GeForce RTX 3090","code":null},{"id":309754,"name":"NVIDIA RTX A6000","code":null},{"id":372210,"name":"NVIDIA A10 PCIe","code":null},{"id":370037,"name":"NVIDIA RTX A4000","code":null},{"id":370029,"name":"NVIDIA RTX A5000","code":null},{"id":370044,"name":"AMD Radeon Instinct MI250X","code":null},{"id":372209,"name":"Huawei Ascend 910B","code":null},{"id":370041,"name":"NVIDIA GeForce RTX 3090 Ti","code":null},{"id":370038,"name":"NVIDIA GeForce RTX 4080","code":null},{"id":370014,"name":"NVIDIA GeForce RTX 4090","code":null},{"id":372177,"name":"NVIDIA H100 SXM5 80GB","code":null},{"id":370027,"name":"AMD Radeon RX 7900 XTX","code":null},{"id":372181,"name":"NVIDIA RTX 6000 Ada Generation","code":null},{"id":372203,"name":"NVIDIA L20 PCle","code":null},{"id":372196,"name":"NVIDIA HGX H20","code":null},{"id":372186,"name":"AMD Instinct MI300X","code":null}]}},"origins":[{"id":8863,"title":"Machine Learning Hardware","description":"This dataset contains detailed information about machine learning hardware, including GPUs, NPUs, and other specialized AI chips. It includes technical specifications such as computational performance across different precision levels (FP64, FP32, FP16, INT8, etc.), memory configurations, release dates, pricing, and manufacturing details.","producer":"Epoch AI","citationFull":"Epoch AI, 'Data on Machine Learning Hardware'. Published online at epoch.ai. Retrieved from 'https://epoch.ai/data/machine-learning-hardware' [online resource].","attributionShort":"Epoch AI","urlMain":"https://epoch.ai/data/machine-learning-hardware","urlDownload":"https://epoch.ai/data/generated/ml_hardware.zip","dateAccessed":"2025-10-10","datePublished":"2025-10-10","license":{"url":"https://epoch.ai/data/machine-learning-hardware","name":"CC BY 4.0"}},{"id":3393,"title":"US consumer prices","description":"The Bureau of Labor Statistics reports the monthly Consumer Price Index (CPI) of individual goods and services for urban consumers at the national, city, and state levels. CPI is presented on an annual basis, which we have derived as the average of the monthly CPIs in a given year.","producer":"U.S. Bureau of Labor Statistics","citationFull":"U.S. Bureau of Labor Statistics","urlMain":"https://www.bls.gov/data/tools.htm","dateAccessed":"2025-04-12","datePublished":"2025","license":{"url":"https://www.bls.gov/opub/copyright-information.htm","name":"Public domain"}}]}