Hi techy savvy friends, I'm hitting some bottlenecks in Python (mostly trying to data mine stock data / models).
My mobo> ASUS TUF GAMING B660M-PLUS D4 LGA 1700 Micro-ATX Motherboard
Current RAM> Corsair Vengeance RGB PRO 16GB (2x 8GB) DDR4 3200MHz
What's a good buy atm to speed things up, 2x64 ddr4 for a few hundred bucks or 2x32gb?
You sure it is a speed things up problem?
Check you system to see if you are utilising the full 16GB. I think it is amount of ram problem not ram speed problem.
16GB will not be great when trying to process allot of data, but you need to investigate where the bottle neck is as it could be ram or it could be that ram is borderline and when it does occur then your HDD/SSD is slow.
First do some memory investigation to see if you are running low/swapping memory and if you are then gran 2 x 32 GB sticks at the sweet $ spot for the speed and give it a go.
If you are trying to process allot of data and the disk it is being read from is slow then the this will be a bottle neck.
Python speed has improved allot of the last few yeas, so check you are running a recent version.
Hitting CPU and Ram bottlenecks
AI in second monitor seems to take up a huge % of memory
Depends on budget. You have four slots to use, maybe 4x16gb? That’ll be $160-180 for a Corsair kit from Amazon
Ah, the old "lets fix the software problem by throwing more hardware at it".
Sure you can double you RAM, but fixing software might drop the demands by an order of magnitude or two.
AI in second monitor
Ah … so your real problem is with a local AI model? Do you want a bigger model? It is really RAM limited?
And "seems to take up a huge % of memory" may well be normal. You want it to uses as much RAM as you give it, in many cases. That does not mean more will solve your problem.
Maybe you need a better GPU, to optimise the scripts, or choose a more suitable model.
I'm not an I.T. expert but from what I read and other comments seem to be missing is for running data/AI models you need a expensive GPU with lots of RAM (more the better).
comments said using CPU and it's RAM is a very inefficient way of running models. so that's why you are probably getting slowing down and or starting to lag/freeze.
if you are spending few hundred on RAM might as well spend a bit more for older model GPU.
Dodge Ram