# How to choose a GPU - Please forgive the total noob question

**URL:** <https://discourse.julialang.org/t/how-to-choose-a-gpu-please-forgive-the-total-noob-question/44339>\
**Category:** Offtopic\
**Tags:** question, gpu\
**Created:** [August 5, 2020, 3:03pm UTC](https://discourse.julialang.org/t/how-to-choose-a-gpu-please-forgive-the-total-noob-question/44339 "2020-08-05T15:03:02Z")\
**Posts on this page:** 4\
**Page:** 2

<div class="post-metadata">

**Author:** ![RussBowdrey](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/russbowdrey/32/14980_2.png) [@RussBowdrey](https://discourse.julialang.org/u/RussBowdrey)\
**Post date:** [August 5, 2020, 9:47pm UTC](https://discourse.julialang.org/t/how-to-choose-a-gpu-please-forgive-the-total-noob-question/44339/21 "2020-08-05T21:47:03Z")

</div>

> [@ChrisRackauckas](#):
>
> > [@RussBowdrey](#):
> >
> > Why is the ratio between SP and DP usually so high? And what is it about the nvidia 1660 that they had a lower ratio? (I think I found the answer to this and it’s to do with not having Tensor cores and introducing SP cores instead…)
> 
> To sell more expensive cards. Literally, at one point it wasn’t so bad, but then that cut into the data center GPUs, so Nvidia stopped people from using it for FP64 computing through a driver setting. Everyone found out and disabled that to get back the FP64 compute, so they started destroying that part of the cheaper GPUs and sells the same chips with a really high markup for the FP64 (yes, same exact GPU chip in both the $5000 card and the $500 card). Then people started using FP32 a bunch in ML, so they had to ban those from data centers using licensing. Seriously though, if you didn’t live through this story, it’s mind boggling.

This is absolute gold, thank you so much!

---

<div class="post-metadata">

**Author:** ![johnh](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/johnh/32/3615_2.png) [@johnh](https://discourse.julialang.org/u/johnh)\
**Post date:** [August 5, 2020, 10:02pm UTC](https://discourse.julialang.org/t/how-to-choose-a-gpu-please-forgive-the-total-noob-question/44339/22 "2020-08-05T22:02:31Z")

</div>

@ChrisRackauckas Thankyou for the explanation on Titan Black and the licensing.  
I worked with an oil and gas company which had literally thousands of GPUs in production.  
As these were in data centres they could not update the drivers beyond that crucial version.

I did try to evangelise Julia when I was there, and I did find out they had funded some Julia work in the seismics field a couple of years ago.

---

<div class="post-metadata">

**Author:** ![ImreSamu](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/imresamu/32/20677_2.png) [@ImreSamu](https://discourse.julialang.org/u/ImreSamu)\
**Post date:** [September 8, 2020, 3:39am UTC](https://discourse.julialang.org/t/how-to-choose-a-gpu-please-forgive-the-total-noob-question/44339/23 "2020-09-08T03:39:33Z")

</div>

new blog post:  
_“Which GPU(s) to Get for Deep Learning: My Experience and Advice for Using GPUs in Deep Learning”_  
2020-09-07 by Tim Dettmers

> **[The Best GPUs for Deep Learning in 2023 — An In-depth Analysis](https://timdettmers.com/2023/01/30/which-gpu-for-deep-learning/)**
>
> Here, I provide an in-depth analysis of GPUs for deep learning/machine learning and explain what is the best GPU for your use-case and budget.

bonus:

Tim Dettmers: _“In the community aspect, AMD is a bit like Julia vs Python. Julia has a lot of potential and many would say, and rightly so, that it is the superior programming language for scientific computing. Yet, Julia is barely used compared to Python. This is because the Python community is very strong. Numpy, SciPy, Pandas are powerful software packages that a large number of people congregate around. This is very similar to the NVIDIA vs AMD issue.”_

---

<div class="post-metadata">

**Author:** ![ToucheSir](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/touchesir/32/14411_2.png) [@ToucheSir](https://discourse.julialang.org/u/ToucheSir)\
**Post date:** [September 8, 2020, 4:27am UTC](https://discourse.julialang.org/t/how-to-choose-a-gpu-please-forgive-the-total-noob-question/44339/24 "2020-09-08T04:27:20Z")

</div>

Just saw that post as well, what an unexpected passage!

Perhaps the nuance here is that Julia does have a healthy and established community that a smaller (but still significant) number of people congregate, whereas ROCm unfortunately does not. I’d concede the last sentence if the author replaced numpy/scipy/pandas with torch/tensorflow because Julia’s deep learning community is proportionally much smaller than Python’s. Anecdotally, all of my research code is and continues to be written in Python. However, one would be hard-pressed to find a language community that has the same confluence of “backend” implementation experience, application domain knowledge and enthusiasm for implementing a competitive ML/DL ecosystem. That still leaves an organizational and human resource deficit, but we’re working on that 🙂

[Previous page](https://discourse.julialang.org/t/how-to-choose-a-gpu-please-forgive-the-total-noob-question/44339.md?page=1)