So i am finally gettng around to learning how to program AI. Want to do local LLMs of some sort. but being a newbie to this whole area, i am not sure what i need. the options are:
1) Get an AMD CPU and an AMD graphics card in a pc or laptop, use some of the AMD software, for instance, one recent headline i saw was: "DeepSeek performing very well on AMD Radeon 7900 XTX". Is that a way to go? Is there a better choice of AMD hardware that is available today? Any recommendations on RAM, etc? There seem to be ways of porting CUDA onto an AMD graphics card too.
2) Wait until March, and try to get an NVDA "Project Digits" desktop with a GPU in it. I would have to brush up on my Linix Ubuntu to use it, but am workng on that now. The price expected ($3000) was not outrageous for what would be basically a supercomputer on my desktop.
3) is there ANY Intel based PC/Laptop that is going to be capable of AI LLMs? Seems like Intel has made a series of performance and reliability mis steps, but they DO apparently have new hardware coming out, that they were pushng at the CES show. It seems like they might be more oriented to a microsoft co-pilot type version of AI? Is intel worth a look???
i am enticed by statements like: "Microsoft rolls out DeepSeek's AI model on Azure"
I am not sure i am even asking the right questions here. but would like to end up with an AI ready computer of some sort soon, that would not be obsolete for at least the next year.
Should i even be thinkng of using deepseeks LLMs, since they will be much more unstable and harder to use than the entire NVDA CUDA stack? Seems like NVDA is more user friendly, although requires better hardware.
Please educate me. i am not sure where to start
1) Get an AMD CPU and an AMD graphics card in a pc or laptop, use some of the AMD software, for instance, one recent headline i saw was: "DeepSeek performing very well on AMD Radeon 7900 XTX". Is that a way to go? Is there a better choice of AMD hardware that is available today? Any recommendations on RAM, etc? There seem to be ways of porting CUDA onto an AMD graphics card too.
2) Wait until March, and try to get an NVDA "Project Digits" desktop with a GPU in it. I would have to brush up on my Linix Ubuntu to use it, but am workng on that now. The price expected ($3000) was not outrageous for what would be basically a supercomputer on my desktop.
3) is there ANY Intel based PC/Laptop that is going to be capable of AI LLMs? Seems like Intel has made a series of performance and reliability mis steps, but they DO apparently have new hardware coming out, that they were pushng at the CES show. It seems like they might be more oriented to a microsoft co-pilot type version of AI? Is intel worth a look???
i am enticed by statements like: "Microsoft rolls out DeepSeek's AI model on Azure"
I am not sure i am even asking the right questions here. but would like to end up with an AI ready computer of some sort soon, that would not be obsolete for at least the next year.
Should i even be thinkng of using deepseeks LLMs, since they will be much more unstable and harder to use than the entire NVDA CUDA stack? Seems like NVDA is more user friendly, although requires better hardware.
Please educate me. i am not sure where to start
Last edited:
Facebook
Twitter
Instagram