Among the below options, which one is better for deep learning research?
- ASUS TUF Gaming F15 FX506HE Core i5 11th Gen RTX 3050 Ti 4GB Graphics
- Asus TUF Dash F15 FX516PE Core i5 11th Gen RTX 3050Ti 4GB Graphics
- Asus TUF Dash F15 FX516PE Core i7 11th Gen RTX 3050Ti 4GB Graphics
- Asus TUF Dash F15 FX516PM Core i5 11th Gen RTX3060 6GB Graphics
- MSI GP65 Leopard 10SEK Core i7 10th Gen RTX 2060 6GB Graphics
- MSI GF65 THIN 10UE Core i5 10th Gen RTX 3060 MAX-Q 6GB Graphics
Hi Hasin,
What is your main use case: training or inference?
Do you have any idea regarding which DNN models you will be working with? Do you know these models' training / inference memory footprints?
Generally speaking, 3050Ti + 4GB DRAM would be just fine for inference / model evaluation. However, you would have a lot of problems if you attempt to use it for training medium to large DNN models. (4GB is too low)
-good luck
Dear Hasin Rehana
It is okay to have NVIDIA GeForce RTX 3050 Ti 4GB GDDR6 for normal usage but for deep learning research I would recommend Intel 5 processor with atleast 12 GB Ram in your laptop otherwise with low configuration laptop while compiling your program it will stuck in the process.
Kind Regards
Qamar Ul Islam
It would be important to know what is the budget!
As long as budget is not specified, I would recommend to just use a supercomputer ...
Within budget boundaries, the experts might be able to give you a good recommendation - if the concrete goal is known as well.
Both Dell and AMD have desktop work stations that would offer much better performance for "deep learning research". However if your using an existing laptop or prefer one, I'd recommend combining a core i9 processor with lots of RAM and the best GPU you can afford. - Usually the 17 in models offer the better performing options. If you need a 15 inch model definitely chose the best i7 you can afford. With, if possible, the RTX 3060Ti 6GB or optionally, the RTX 3050Ti 4GB.
Since you are talking about RESEARCH, I am assuming you are training a model, not just inference. The critical questions to ask are:
1. Budget
2. Language/DL toolset - eg python/tensorflow/pytorch? Java DL4J? etc
3. Model sizes - 4Gb is really tight for training the sample models - 6Gb would be better. Your choice now can't be upgraded unlike normal RAM!
4. Model design - are you using CNN or RNN (or other) networks? Usually I found RNN use less memory and your constraints are more CPU bottlenecked feed into GPU limitations. But even a lower i7 is ok, most DNN are GPU constrained that's why Lenovo/Dell/HP DL workstations don't bother with i9.
5. If you don't have a working model yet, estimate what it would be closest to eg IMDB LSTM or VGG16 etc and use plaidbench (see https://openbenchmarking.org/test/pts/plaidml) to chose the right GPU.
6. Overheating is a problem with laptops - many have thermal throttling problems - often better to get big heavy gaming laptop or desktop replacement, rather than light thin ones if you intend to run models for hours. Otherwise your CPU/GPU can also be damaged by overheating.
I recommend you try Google Colab first before making this decision, observe your sample models and how much GPU time and memory they use. This will better inform your decision.
Don't forget that if you are a current student or educator you can also get big discounts from the DELL, Lenovo and HP education stores.
I don't know if your country has big sales, but often there is one in Black Friday in late November.
Hi, Hasin. Did you buy the RTX 3050? I am interested in deep learning research and have the same question.
- Badges
- Science topic
- Similar topics
- Filing