Our organization has decided to use automation with UiPath. We handle a large number of documents (nearly 100 per day), including structured, semi-structured, and unstructured formats. We have been advised to use CPUs instead of GPUs. Has anyone worked on a similar scenario? Please share your experience.
Welcome to the community
few models need gpu to be enabled…if speed is not a concern then cpu also should do the job for you
cheers
@Donamol_Kurian Welcome to community!!
I got some article from other, go throw this it may help you..
When processing a large volume of documents (like your case with nearly 100 per day), CPU-based automation in UiPath has been effective. Here’s how you can optimize:
- OCR: For structured and semi-structured documents, OCR tools (e.g., UiPath’s native OCR, Google OCR, or Abbyy FlexiCapture) are used for extracting data. These tools are generally optimized for CPU usage.
- Parallelization: While CPUs are preferred, UiPath does allow for parallel processing (running multiple robots concurrently), which can make handling large volumes of documents more efficient. This is especially useful when you have a high throughput of documents.
- AI and ML Models: If you’re dealing with highly unstructured documents and need to implement AI/ML models, you might want to test a combination of local CPU-based processing and cloud-based models (which are optimized for inference but don’t always require GPUs).
Ideally GPU is recommended for training any AI models. With CPU - it is possible, but you might face some challenges when dealing with huge training data / specific model in long run
In my view - you can start with CPU if GPU is not possible at the moment but definitely consider enhancing to GPU at least in future