Tesla M40 24GB GPU: very poor machine-learning performance?

• ai-art-generator

  3 627 0.0 Python

  For automating the creation of large batches of AI-generated artwork locally.

I'm a software engineer, but a complete machine-learning noob (not exactly a linux guru, either). I'm trying to use the GPU for VQGAN+CLIP image generation. Running on an RTX 3060, I get almost 4 iterations per second, so a 512x512 image takes about 2 minutes to create with default settings. Running on the Tesla M40, I get about 0.4 iterations per second (~22 minutes per 512x512 image at the same settings). A full order of magnitude slower! I'd read that older Tesla GPUs are some of the top value picks when it comes to ML applications, but obviously with this level of performance that isn't the case at all. I figure I must be going wrong somewhere.

• InfluxDB

  www.influxdata.com featured

  Power Real-Time Data Analytics at Scale. Get real-time insights from all types of time series data with InfluxDB. Ingest, query, and analyze billions of data points in real-time with unbounded cardinality.

  

Suggest a related project