20x Performance for the Same HW Price: GPGPU-based Machine Learning at Yandex

Victor Haydin
Head of Products and Services at ELEKS

Russian search-engine Yandex has disclosed some details about their machine-learning framework, FML. The most interesting detail is that it runs on 80 TFLOPS cluster powered by GPGPU. This is quite unusual application for GPU, as ML algorithms are usually hard to be paralleled. However they have managed to adapt their decision tree algorithm for high-level of parallelism. As a result Yandex has achieved more than 20x speed-up for the same hardware price.
They are going to upgrade their cluster to 300 TFLOPS. Yandex expects its cluster to be in the list of top 100 most powerful supercomputers in the world after that upgrade.



  • Can you give out Difference’s whats the benefit we will get for Complete Online Video Transcoding usign GPGPU hardware you are referring. Also

    • Hi, Eric

      In general, you can get quite a good performance of video transcoding on GPGPU, sometimes much better than on CPU. But it all depends on details of your particular task: what kind of video do you want to transcode, what are the source and target codecs and so on. Should you provide us with more details I will be happy to answer all your questions. Please don’t hesitate to contact me via e-mail: victor.haydin@eleks.com

      Thank you,