Tech tip: When running AI locally, use the graphics card

[AlexB23]

Hello folks. A few days ago, I found out that the AI front-end system is much faster at processing or "learning" PDF documents when I set it to GPU, as it takes only an hour or so to learn 10 million words of documents compared to over a day using CPU. In other words, I do not have to wait as much for the machine to learn documents. Which means if you want me to use the AI more, I can.

 

[Reluctant Theologian]

Hello folks. A few days ago, I found out that the AI front-end system is much faster at processing or "learning" PDF documents when I set it to GPU, as it takes only an hour or so to learn 10 million words of documents compared to over a day using CPU. In other words, I do not have to wait as much for the machine to learn documents. Which means if you want me to use the AI more, I can.

View attachment 356562

travelling soon?

 

[AlexB23]

travelling soon?

Yep. 11/4-11/18

 

[Reluctant Theologian]

Yep. 11/4-11/18

Have a great journey - I'm sure you will see plenty of interesting stuff besides family. I'm in NL now as well ... Let me know if/when you may need any assistance

 

[AlexB23]

Have a great journey - I'm sure you will see plenty of interesting stuff besides family. I'm in NL now as well ... Let me know if/when you may need any assistance

Thank you brother. God bless.

 

[timewerx]

Hello folks. A few days ago, I found out that the AI front-end system is much faster at processing or "learning" PDF documents when I set it to GPU, as it takes only an hour or so to learn 10 million words of documents compared to over a day using CPU. In other words, I do not have to wait as much for the machine to learn documents. Which means if you want me to use the AI more, I can.

View attachment 356562

I told you this a long time ago.

You were complaining so often about your old laptop computer how it's so slow with the low RAM and I told you to use the GPU.

Modern data-crunching apps tend to have the option to use GPU including open-source apps. You just need to explore the options/settings/preferences/documentation/google it to see if the app has it.

I always look for that option whenever I'm using app that's starting to feel slow in relation to the volume of data being handled.

 

[AlexB23]

I told you this a long time ago.

You were complaining so often about your old laptop computer how it's so slow with the low RAM and I told you to use the GPU.

Modern data-crunching apps tend to have the option to use GPU including open-source apps. You just need to explore the options/settings/preferences/documentation/google it to see if the app has it.

I always look for that option whenever I'm using app that's starting to feel slow in relation to the volume of data being handled.

Yeah, I did not know how to do that at the time, but you were right, even an NVIDIA 1050M GPU has the power to do this.

In other words, I should have taken your advice. I do not need an NVIDIA 3090, and am really surprised that a 1050 mobile GPU on a laptop can be so fast. The GPU is a lifesaver. My GPU is roughly 2 TeraFLOPS which is slow by today's standards, but is dozens of times faster compared to my CPU.

 

[timewerx]

Yeah, I did not know how to do that at the time, but you were right, even an NVIDIA 1050M GPU has the power to do this.

In other words, I should have taken your advice. I do not need an NVIDIA 3090, and am really surprised that a 1050 mobile GPU on a laptop can be so fast. The GPU is a lifesaver. My GPU is roughly 2 TeraFLOPS which is slow by today's standards, but is dozens of times faster compared to my CPU.

Your laptop is still way more capable than mine that still uses non-Vulkan Intel GPU.

This is why Nvidia is one of the companies at the forefront of research with driverless vehicles and machine learning.

 

[AlexB23]

Your laptop is still way more capable than mine that still uses non-Vulkan Intel GPU.

This is why Nvidia is one of the companies at the forefront of research with driverless vehicles and machine learning.

Yeah, mine uses Vulkan or CUDA.