Hello,
I am data scientist. I run Deep learning and LLM mode. I am using Lenovo 11th Gen Intel(R) Core™ i7-11800H @ 2.30GHz, 2304 Mhz, 8 Core(s), 16 Logical Processor(s). I have integrated GPU Intel(R) UHD Graphics. Another GPU come with Laptop GPU Nvidia T1200 Laptop GPU. Recently I bought GeForce Nvidia RTX 4070 external GPU to increase cuda core. It seems become faster. But in some point it freezed. Then I talked with NVidia customer support. They told me Quadro cards are not recommended to be used with Geforce (or vice versa). They suggest me to use one GPU. and for Cuda related I should go with Quadro based cards as it is Quadro cards are designed for for CUDA tool kits, also as you are using this for AI / Deep Learning / Machine Learning related work, Geforce cards will not be suitable as the cards are not designed to work with the APIs required by Deelp learning. I am little bit confused as RTX 4070 has 5888 cuda cores where T1200 has 1024.
My questions is
- for deep learning which is better?
- Is it true I can not use both GPU simultaneously? if it is true should I remove one device? or just enable or disable from device manager when I ran to make sure using one GPU? My understanding is that even I ran something windows automatically choose one GPU for one applicaiton. Correct me if I am wrong.