Blackwell is 2 times more efficient than its predecessors
Nvidia has unveiled new Blackwell chips that significantly speed up the process of training large artificial intelligence (AI) models. According to new data published by the non-profit organization MLCommons, the number of chips required to train large language models has significantly decreased.
MLCommons specializes in comparing the performance of AI hardware. The latest report included testing results from Nvidia and AMD chips, as well as other manufacturers. At the same time, an important indicator remains the efficiency of training AI models – the stage when systems receive huge amounts of data for training. Although many now pay attention to the speed of AI when responding to users, it is the number and power of chips involved in training that remains a critical factor.
The tests were conducted using the Llama 3.1 405B* model, a complex open AI module with a huge number of parameters developed by Meta*. Training data for this model was provided only by Nvidia and its partners. The results showed that the new Blackwell chips are twice as fast as the previous Hopper series per chip.
It took 2,496 Blackwell chips to complete the test, which took 27 minutes. That’s better than the previous chips, which were more than three times as many.
Experts note a change in approaches in the industry: instead of huge homogeneous clusters of tens of thousands of chips, more compact subsystems are increasingly used for individual AI training tasks. This method allows for speeding up the process and reducing the time for training models with trillions of parameters.
* belongs to the Meta company, the organization is recognized as extremist, its activities are prohibited in Russia
Anticipated New Releases As we look towards Spring 2025, the television landscape is set to be…
Manage Cookie Consent
To provide the best experiences, we use technologies like cookies to store and/or access device information. Consenting to these technologies will allow us to process data such as browsing behavior or unique IDs on this site. Not consenting or withdrawing consent, may adversely affect certain features and functions.
Functional
Always active
The technical storage or access is strictly necessary for the legitimate purpose of enabling the use of a specific service explicitly requested by the subscriber or user, or for the sole purpose of carrying out the transmission of a communication over an electronic communications network.
Preferences
The technical storage or access is necessary for the legitimate purpose of storing preferences that are not requested by the subscriber or user.
Statistics
The technical storage or access that is used exclusively for statistical purposes.The technical storage or access that is used exclusively for anonymous statistical purposes. Without a subpoena, voluntary compliance on the part of your Internet Service Provider, or additional records from a third party, information stored or retrieved for this purpose alone cannot usually be used to identify you.
Marketing
The technical storage or access is required to create user profiles to send advertising, or to track the user on a website or across several websites for similar marketing purposes.
/imgs/2025/06/04/17/6839950/bdc642f99da9d76ff48c0c0f569f220925382ca7.jpg)