What Windows version, how much memory?

Windows 11 Pro, 16GB

How about an error message? 

This is an impossible thread to contribute to until you provide comprehensive information about your environment, errors, general setup and anything that you've changed/modified recently that may have caused problems afterwards.  Can you post a useful summary please?

gehenna:

 

This is an impossible thread to contribute to until you provide comprehensive information about your environment, errors, general setup and anything that you've changed/modified recently that may have caused problems afterwards.  Can you post a useful summary please?

 

You said it far more gracefully than I. 

Windows 11 pro will be fine, so long as it is 22H2 or later (I use 24H2).

Batman: Windows 11 Pro, 16GB

After playing around with running a bunch of AI models I think you're going to struggle getting anything really running with 16gb of ram. These are incredibly memory hungry. You should chuck in an extra 16gb.

Have a play around with LM Studio instead: https://lmstudio.ai/ - it's more user friendly and can run everything including DeepSeek and Llama totally fine. The more ram the better along with having a dedicated GPU.

I was experimenting around tonight (crap where did that time go) with a few models running on a server with 32 CPU cores and assigned 16gb of ram and even then AI models do run slow on CPU alone however you chuck in a decent GPU (the GTX 1070 may work, but could be a little too old) and performance improves greatly.

michaelmurfy:

 

Batman: Windows 11 Pro, 16GB

 

After playing around with running a bunch of AI models I think you're going to struggle getting anything really running with 16gb of ram. These are incredibly memory hungry. You should chuck in an extra 16gb.

 

Have a play around with LM Studio instead: https://lmstudio.ai/ - it's more user friendly and can run everything including DeepSeek and Llama totally fine. The more ram the better along with having a dedicated GPU.

 

I was experimenting around tonight (crap where did that time go) with a few models running on a server with 32 CPU cores and assigned 16gb of ram and even then AI models do run slow on CPU alone however you chuck in a decent GPU (the GTX 1070 may work, but could be a little too old) and performance improves greatly.

 

thanks, lmstudio is amazing, just one click and everything is installed

interestingly, 13900k + 4090 vs MBPro vs 5500 + 1070 is roughly the same speed for deepseek in gwen aka rather slow.

llama is a lot faster, not sure if that means anything.

have deleted docker as i don't want windows virtual Machine turned on (apparently makes games slower?)

gehenna:

 

This is an impossible thread to contribute to until you provide comprehensive information about your environment, errors, general setup and anything that you've changed/modified recently that may have caused problems afterwards.  Can you post a useful summary please?

 

well it didn't have that when it said failed. after some digging around it says i need to turn on virtual machine.

i've deleted docker and now running LM Studio