VertexAI models for tabular data ran locally, PC recources question

[jeden8l]

Hi everyone. 
This is not an issue related thread, I'm just seeking advice. 

My Vertex AI regression model for tabular data is running on my local PC. 

Single prediction request takes about ~28ms (message printed within Google SDK Shell). 
I need to go on that as low as possible. 
What should I consider doing first?

Increasing CPU threads and speed? AMD or Intel? 
Increasing memory size and it's speed? 
Regarding GPU utilization, model will be used on the rented server, so not really an option to have GPU available. 

Just need some guide to whether focus on the best and fastest CPU available, or rather balance it between CPU and memory recources. 

Any thoughts please?
I will be thankful for any examples, like CPU and memory spec, data labels quantity input for model and models prediction time you are getting. 

[nimrah-waqar · Accepted solution]

For optimizing prediction time on your local PC for a Vertex AI regression model:

Prioritize CPUs with high single-thread performance (Intel Core i9 or AMD Ryzen 9).

Aim for at least 16GB of RAM.

Benchmark different hardware configurations to find the optimal balance between cost and performance for your specific use case.