Newbie Question About the Best NVIDIA On-Premises Solution To Give The Highest Speed For My Deep Learning Training and Inference

I have a deep learning model that uses the Pytorch and TensorFlow for predicting prices that we want to be using in our new start-up.

I currently use my MacBook Pro for both training and prediction, but it is terribly slow.

I want an in-office machine (a server, sort of) that will be dedicated to training the model and making inference from it, and that will take the training and inference process to the highest possible speed. I think NVIDIA holds an answer to my question, but I don’t have any idea which of their products I will need or even how their products work.

I will really appreciate it if someone can volunteer a gentle guide on which of NVIDIA products I should go for and how it works.

Thanking you in anticipation of your kind assistance

Welcome to the community and congratulations on your new start up.
Firstly I would encourage all startups to check out our inception program: Inception Program for Startups | NVIDIA
The program will give you access to expertise you don’t even know you need.

As for your specific question, there are so many options and directions you could go there is no simple answer.
I would suggest you try out some of the GPU accelerated nodes available on AWS or Azure, and that will give you an idea of the performance you can expect for different GPUs, based on that - and your budget, you can start to architecture a solution. If the objective is to get a single GPU workstation which you plan to run 24/7 - then you should consider our professional RTX series Graphics Cards for Professional Desktop Workstations | NVIDIA, they are more costly than GeForce, but designed for 24/7 operation and some have ECC memory too.
If you have secured some funding and you expect you need alot of AI power then you can consider a DGX Station, DGX Systems : Built for the Unique Demands of AI | NVIDIA , that’s our most powerful workstation class solution.

Sorry I couldn’t help you in a more precise way. Best of luck !

Thanks a lot for the detailed answer @nadeemm

You are right, my objective right now is to get a single GPU workstation which I plan to run 24/7.

I guess that means the Graphics Cards for Professional Desktop Workstations | NVIDIA is my best bet.

The highest solution from the link you sent seems to be the NVIDIA RTX A6000, correct me if I’m wrong.

However, I’ve always seen write-ups relate the NVIDIA TITANS as the best when it has to do with Data Science, AI and Deep Learning, which is my use case. Are they the same thing? Again, I might be getting it totally mix up.

Finally, whatever your prescription is between the above mentioned work stations, could you please help suggest where and how to quickly buy one for someone in Abuja, the Federal Capital City of Nigeria?

Also, I understand that there may be some configurations that needs to be done before I can use it with my MacBook Pro.

Just to mention that I am already aware of the Mac Studio M1 Ultra, but a dedicated 24/7 workstation seems to be just what I need at the moment.

Looking forward to your kind response.

The Titan cards blurred the lines between Geforce and the Professional cards when they were launched, but you would have notices there are not titan cards based on our current latest architecture , Ampere. So the highest end consumer card would be Geforce 3090 Ti , not any Titan card.
The RTX A6000 would be the most capable GPU of the professional series. You will find a number of our workstation partners have developed full workstations around this GPU, you can find more information about some of them here:

With these high performance solutions - its important to ensure all the core elements of the system are well chosen - a system is only as good as its weakest component - and I don’t want to be the person recommending you simply spend your limited funding on the most expensive solution. You will have to figure out what parts of the configuration will give you the best value.
When you do make a decision let us and the community know what you settled on and why - perhaps it will help the next person facing the same dilemma !
thanks