# How to utilize gpu **URL:** \ **Category:** AI Discussions\ **Tags:** ai-discussions\ **Created:** [September 7, 2024, 4:29am UTC](https://community.deeplearning.ai/t/how-to-utilize-gpu/688937 "2024-09-07T04:29:14Z")\ **Posts on this page:** 17\ **Page:** 1
**Author:** ![SASWATA\_MAITY](https://avatars.discourse-cdn.com/v4/letter/s/f05b48/32.png) [@SASWATA\_MAITY](https://community.deeplearning.ai/u/SASWATA_MAITY)\ **Post date:** [September 7, 2024, 4:29am UTC](https://community.deeplearning.ai/t/how-to-utilize-gpu/688937/1 "2024-09-07T04:29:14Z")
![Screenshot 2024-09-07 095508](https://global.discourse-cdn.com/dlai/original/3X/5/3/5368207524d5c063f16ce6243c94bdcc080d33ad.png) this is my current laptop config can i add an external gpu for faster training process will it work? how to utlize all the gpu ?? ---
**Author:** ![Nevermnd](https://sea1.discourse-cdn.com/dlai/user_avatar/community.deeplearning.ai/nevermnd/32/126021_2.png) [@Nevermnd](https://community.deeplearning.ai/u/Nevermnd)\ **Post date:** [September 7, 2024, 11:07am UTC](https://community.deeplearning.ai/t/how-to-utilize-gpu/688937/2 "2024-09-07T11:07:45Z")
@SASWATA_MAITY it all depends on what program/software you are running, or even whether it is one you built/wrote yourself. For example: Plain vanilla Python won’t just ‘magically’ start using your GPU unless you write a program that tells it to. ---
**Author:** ![TMosh](https://sea1.discourse-cdn.com/dlai/user_avatar/community.deeplearning.ai/tmosh/32/9321_2.png) [@TMosh](https://community.deeplearning.ai/u/TMosh)\ **Post date:** [September 7, 2024, 3:05pm UTC](https://community.deeplearning.ai/t/how-to-utilize-gpu/688937/3 "2024-09-07T15:05:02Z")
Here is some very sketchy background on machine learning and GPUs. I don’t know very much about it. Some GPUs provide an interface called CUDA. Some machine learning tools (like TensorFlow and PyTorch) can be configured to use CUDA. This results in your GPU being used for the machine learning calculations. So you can start by doing some research on your laptop’s GPU and see if it supports CUDA. If it does not support CUDA, and if your laptop has a USB-C or Lightning interface, there are external GPUs you can connect to your laptop. ---
**Author:** ![balaji.ambresh](https://sea1.discourse-cdn.com/dlai/user_avatar/community.deeplearning.ai/balaji.ambresh/32/128697_2.png) [@balaji.ambresh](https://community.deeplearning.ai/u/balaji.ambresh)\ **Post date:** [September 7, 2024, 3:15pm UTC](https://community.deeplearning.ai/t/how-to-utilize-gpu/688937/4 "2024-09-07T15:15:39Z")
[CUDA](https://developer.nvidia.com/cuda-toolkit) is for NVidia gpus and [ROCM](https://www.amd.com/en/products/software/rocm.html) is for AMD gpus. ---
**Author:** ![Nevermnd](https://sea1.discourse-cdn.com/dlai/user_avatar/community.deeplearning.ai/nevermnd/32/126021_2.png) [@Nevermnd](https://community.deeplearning.ai/u/Nevermnd)\ **Post date:** [September 7, 2024, 3:21pm UTC](https://community.deeplearning.ai/t/how-to-utilize-gpu/688937/5 "2024-09-07T15:21:08Z")
@balaji.ambresh Sorry, not to repeat somewhere else, but no longer true: > **[AMD Quietly Funded A Drop-In CUDA Implementation Built On ROCm: It's Now...](https://www.phoronix.com/review/radeon-cuda-zluda#google_vignette)** > > While there have been efforts by AMD over the years to make it easier to port codebases targeting NVIDIA's CUDA API to run atop HIP/ROCm, it still requires work on the part of developers. ---
**Author:** ![balaji.ambresh](https://sea1.discourse-cdn.com/dlai/user_avatar/community.deeplearning.ai/balaji.ambresh/32/128697_2.png) [@balaji.ambresh](https://community.deeplearning.ai/u/balaji.ambresh)\ **Post date:** [September 7, 2024, 4:12pm UTC](https://community.deeplearning.ai/t/how-to-utilize-gpu/688937/6 "2024-09-07T16:12:53Z")
@Nevermnd What do you mean by `but no longer true` ? At this point, [zluda](https://github.com/vosen/ZLUDA) is not production ready based on this note: > Warning: this is a very incomplete proof of concept. It’s probably not going to work with your application. ZLUDA currently works only with applications which use CUDA Driver API or statically-linked CUDA Runtime API - dynamically-linked CUDA Runtime API is not supported at all ---
**Author:** ![Nevermnd](https://sea1.discourse-cdn.com/dlai/user_avatar/community.deeplearning.ai/nevermnd/32/126021_2.png) [@Nevermnd](https://community.deeplearning.ai/u/Nevermnd)\ **Post date:** [September 7, 2024, 4:24pm UTC](https://community.deeplearning.ai/t/how-to-utilize-gpu/688937/7 "2024-09-07T16:24:39Z")
_shy_– So I have a childhood friend that ended up on the AMD GPU driver team-- though now he works for Qualcomm-- And I actually have an AMD chip in my desktop, which I got more confident with after the XP series-- I mean, yes they were fast, but could also readily catch on fire-- Sorry, scarry. And with that, I run an Nvidia RTX, thus I can’t completely confirm… But, unless AMD is totally out in left field at this point, this is more of a compiler issue which I don’t see as insurmountable. ---
**Author:** ![balaji.ambresh](https://sea1.discourse-cdn.com/dlai/user_avatar/community.deeplearning.ai/balaji.ambresh/32/128697_2.png) [@balaji.ambresh](https://community.deeplearning.ai/u/balaji.ambresh)\ **Post date:** [September 7, 2024, 4:41pm UTC](https://community.deeplearning.ai/t/how-to-utilize-gpu/688937/8 "2024-09-07T16:41:47Z")
Based on the work done in the repository, I don’t think anything is going to hold the developer(s) back from publishing a production ready version for public use. Until then, it’s not a great idea IMO for enterprises and learners with little experience to use such packages. ---
**Author:** ![Deepti\_Prasad](https://sea1.discourse-cdn.com/dlai/user_avatar/community.deeplearning.ai/deepti_prasad/32/165831_2.png) [@Deepti\_Prasad](https://community.deeplearning.ai/u/Deepti_Prasad)\ **Post date:** [September 7, 2024, 5:06pm UTC](https://community.deeplearning.ai/t/how-to-utilize-gpu/688937/10 "2024-09-07T17:06:07Z")
Hi @Nevermnd he isn’t pointing at your knowledge about Nvidia but @balaji.ambresh response is based learner’s image which gives information about the driver’s date as well as nvidia GeForce. Please avoid thinking everyone is arguing here, as a moderator and mentor, mentors go through training and every mentor main focus is and should be addressing learner’s query. please refer FAQ Section for community guidelines for better understanding and communication. Regards DP ---
**Author:** ![Nevermnd](https://sea1.discourse-cdn.com/dlai/user_avatar/community.deeplearning.ai/nevermnd/32/126021_2.png) [@Nevermnd](https://community.deeplearning.ai/u/Nevermnd)\ **Post date:** [September 7, 2024, 6:34pm UTC](https://community.deeplearning.ai/t/how-to-utilize-gpu/688937/11 "2024-09-07T18:34:38Z")
I was not looking to argue-- I mean this is getting a bit personal, but my ‘daily driver’ is a T470P, and honestly it is quite fine for ‘most’ of the things I want to do. If I wish to go heavy, I go to my desktop. But even with a Geforce 940MX, on the laptop even that supports CUDA… ---
**Author:** ![Deepti\_Prasad](https://sea1.discourse-cdn.com/dlai/user_avatar/community.deeplearning.ai/deepti_prasad/32/165831_2.png) [@Deepti\_Prasad](https://community.deeplearning.ai/u/Deepti_Prasad)\ **Post date:** [September 8, 2024, 1:54am UTC](https://community.deeplearning.ai/t/how-to-utilize-gpu/688937/12 "2024-09-08T01:54:31Z")
This information would be helpful to the post creator, thank you for your information. Regards DP ---
**Author:** ![Nevermnd](https://sea1.discourse-cdn.com/dlai/user_avatar/community.deeplearning.ai/nevermnd/32/126021_2.png) [@Nevermnd](https://community.deeplearning.ai/u/Nevermnd)\ **Post date:** [September 8, 2024, 12:59pm UTC](https://community.deeplearning.ai/t/how-to-utilize-gpu/688937/13 "2024-09-08T12:59:01Z")
@SASWATA_MAITY with a different mentor the other day, I also stumbled upon this: > **[CuPy](https://cupy.dev/)** > > NumPy & SciPy for GPU I… have not had a chance to play with it yet, but this could potentially be an easy way to accelerate your Python apps without having to dive into the whole ‘Numba’ business. \*Not sure if you have to download the CUDA libraries first or if it is self-contained. ---
**Author:** ![SASWATA\_MAITY](https://avatars.discourse-cdn.com/v4/letter/s/f05b48/32.png) [@SASWATA\_MAITY](https://community.deeplearning.ai/u/SASWATA_MAITY)\ **Post date:** [September 9, 2024, 5:05am UTC](https://community.deeplearning.ai/t/how-to-utilize-gpu/688937/14 "2024-09-09T05:05:19Z")
i have tried using torch but the memory isnt en ough ---
**Author:** ![SASWATA\_MAITY](https://avatars.discourse-cdn.com/v4/letter/s/f05b48/32.png) [@SASWATA\_MAITY](https://community.deeplearning.ai/u/SASWATA_MAITY)\ **Post date:** [September 9, 2024, 5:05am UTC](https://community.deeplearning.ai/t/how-to-utilize-gpu/688937/15 "2024-09-09T05:05:55Z")
external gpu i dont know what to buy ---
**Author:** ![SASWATA\_MAITY](https://avatars.discourse-cdn.com/v4/letter/s/f05b48/32.png) [@SASWATA\_MAITY](https://community.deeplearning.ai/u/SASWATA_MAITY)\ **Post date:** [September 9, 2024, 5:09am UTC](https://community.deeplearning.ai/t/how-to-utilize-gpu/688937/16 "2024-09-09T05:09:44Z")
i dont have a pc i have a loptop with an intel ghraphics and a nvdia gpu rtx1650 i dont know its not enough maybe ---
**Author:** ![SASWATA\_MAITY](https://avatars.discourse-cdn.com/v4/letter/s/f05b48/32.png) [@SASWATA\_MAITY](https://community.deeplearning.ai/u/SASWATA_MAITY)\ **Post date:** [September 9, 2024, 5:10am UTC](https://community.deeplearning.ai/t/how-to-utilize-gpu/688937/17 "2024-09-09T05:10:16Z")
let me try thanks a lot everyone ---
**Author:** ![Nevermnd](https://sea1.discourse-cdn.com/dlai/user_avatar/community.deeplearning.ai/nevermnd/32/126021_2.png) [@Nevermnd](https://community.deeplearning.ai/u/Nevermnd)\ **Post date:** [September 9, 2024, 9:57am UTC](https://community.deeplearning.ai/t/how-to-utilize-gpu/688937/18 "2024-09-09T09:57:50Z")
> [@SASWATA\_MAITY](#): > > rtx1650 Friend, I think (?) you mean GTX, not RTX, but as far as I can tell, yes, that also has CUDA: [CUDA-Enabled GeForce 1650? - CUDA Setup and Installation - NVIDIA Developer Forums](https://forums.developer.nvidia.com/t/cuda-enabled-geforce-1650/81010) \*I would say though, it is one thing if you are writing your own programs-- Or the memory issue can be a problem, though at the loss of speed you could probably solve this with HD paging. But if you are using an ‘out of the box’ package, unfortunately it may be the case you have less options.