I’m using whisper to transcribe some mp3s.
The performances are terrible so I decided to buy a video card to solve the problem.
Unfortunately online you can find benchmarks with games but not with AI applications.
I wonder if there is a real difference between a Nvidia Tesla T4 ($2000) and a RTX 3060 Ti ($400).
Or maybe two Nvidia Tesla K80s are better?
If it is an option for you, you can try various GPUs at runpod.io and compare the results.
I am not affiliated with that company or service in any way.
Here is a list of Graphics Cards with Whisper performance measured in words per minute. A base line is that an average human talks at 150 words per minute with a typical +/- 50wpm variance.
1 Like
Tnx for your answers.
I know this page and also tomshardware.
In theory Tesla T4 is specific to work with AI but I can’t find any comparative benchmarks.