how much difference does it make focusing on AI hardware for ai processes?
I am having trouble understanding this question. Can you rephrase it? or goo deeper into the question/idea you have?
Running an AI model on your PC takes a LOT of operations. Generally speaking, to run even a basic text-chat style AI model, you need somewhere in the 10 to 45 TOPS range. Things become more intense when you want to perform generative AI, like image or video. You need either a GPU with a sizeable amount of VRAM (12GB is pretty much a bare minimum) or you need a CPU with a built-in NPU and plenty of main system RAM.
1 Like
Thank Doug. this is EXACTLY what I’m looking for.
@TheZacAttack
I’m trying to see what difference it makes to have a computer w/ a strong NPU or a strong GPU for AI programs
thank you guys for responding!
What would you say would be the minimum CPU and GPU you would need to run a AI model? Most likely an Nvidia GPU, but what Intel or AMD processors would you want? This is just out of.curiosity. I have no interest in building an AI capable PC. Plus, how much RAM would you want realistically. Include a text chat model and agenerative AI model please.
1 Like
Text AI models, it all depends on how large of a language set you want the AI model to know. There’s different ones that require different amounts of RAM - and there’s plenty of text based AI that can run just on CPU power. Yeah it’ll take a few seconds to generate a response, but it can work. If you’re not using a GPU to do the text chat, you want the most CPU possible. Realistically, 8 cores at least. Ryzen 7 5700X or so is probably the lowest I’d go for that, and then you want a bunch of main RAM, cuz that’s what it’s going to use.
Generative AI you want as much GPU power as you can throw at it. You’ll have a way different experience with generative AI on an RTX 3060 12GB than you would on an RTX 3090. VRAM and AI TOPS are the name of the game for generative AI, legit as much as you can afford. Running multiple 96GB RTX Pro 6000 Blackwell cards? Go for it. I would never run generative AI on anything below the 12GB 3060, and certainly not with anything with under 12GB total.
2 Likes
Doesnt ChatGPT have a free open source version users can experiment with do their own?
What would I need to make my own AI and teach it things myself?
I know content Creator Vedal creating a Vtuber AI named Neuro-sama ( and another daughter in an attempt to recreate Neuro. He named her Evilyn ). I htink her LLM was Chat-GPT but he has some implementation of Nvidia AI Deep Learning mixed in so she can read the screen and reply in real time. He wont tell anyone cause some companies want to buy the rights to the AI but he wont sell out. All we know is She is made with Python and C#.
What hardware should we look at to make our own AI? Language? Are there open source data bases we can use to upload massive amounts of data that isnt behind copyright?
2 Likes
AI enhancement upgrades do work but the results depend on your tasks. A better NPU can speed up AI workloads like image processing, voice recognition, and data analysis. It helps the CPU and GPU handle other jobs more efficiently. For regular use, the difference may be small, but for AI-focused work, it can be significant. Many new chips are designed to handle AI better and use less power. The technology is still improving fast. If you use AI tools often, getting hardware with strong AI support is a smart move.
1 Like
good info. good info.
1 Like
thanks
1 Like