# PowerVR GPU(GX6650) supports Pytorch

**URL:** <https://forums.imgtec.com/t/powervr-gpu-gx6650-supports-pytorch/3082>\
**Category:** Uncategorized\
**Created:** [September 30, 2019, 7:48am UTC](https://forums.imgtec.com/t/powervr-gpu-gx6650-supports-pytorch/3082 "2019-09-30T07:48:46Z")\
**Posts on this page:** 4\
**Page:** 1

<div class="post-metadata">

**Author:** ![Ramanjineyulu](https://avatars.discourse-cdn.com/v4/letter/r/3d9bf3/32.png) [@Ramanjineyulu](https://forums.imgtec.com/u/Ramanjineyulu)\
**Post date:** [September 30, 2019, 7:48am UTC](https://forums.imgtec.com/t/powervr-gpu-gx6650-supports-pytorch/3082/1 "2019-09-30T07:48:46Z")

</div>

Hi,

I am working on PowerVR Series6XT (Rogue) GX6650 on Linux platform.  
Can you please let me know the possibility to install Pytorch in this GPU.  
If Yes, kindly share the step by step instructions.

Thanks,  
Ram.

---

<div class="post-metadata">

**Author:** ![CarlosSarria](https://sea1.discourse-cdn.com/flex015/user_avatar/forums.imgtec.com/carlossarria/32/263_2.png) [@CarlosSarria](https://forums.imgtec.com/u/CarlosSarria)\
**Post date:** [September 30, 2019, 8:42am UTC](https://forums.imgtec.com/t/powervr-gpu-gx6650-supports-pytorch/3082/2 "2019-09-30T08:42:54Z")

</div>

Hi Ram,

It is not recommended to run a DNN Framework directly on a mobile SoC. If you really want to do it, you should be able to install it just following Pytorch [instructions guide](https://pytorch.org/get-started/locally/). I doubt it will be possible to accelerate it using OpenCL.

Regards.

Carlos.

---

<div class="post-metadata">

**Author:** ![Ramanjineyulu](https://avatars.discourse-cdn.com/v4/letter/r/3d9bf3/32.png) [@Ramanjineyulu](https://forums.imgtec.com/u/Ramanjineyulu)\
**Post date:** [September 30, 2019, 9:08am UTC](https://forums.imgtec.com/t/powervr-gpu-gx6650-supports-pytorch/3082/3 "2019-09-30T09:08:05Z")

</div>

Hi Carlos,  
Thank you for your reply.  
Currently we are not using mobile SoC.

We are using GPU on embedded platform(Renesas RCAR-H3). Please see the below link for more information.

> **[R-Car H3 System-on-Chip (SoC)](https://www.renesas.com/sg/en/solutions/automotive/soc/r-car-h3.html)**
>
> The R-Car H3 SoC delivers computing capabilities that can be applied to in-vehicle infotainment systems as well as driving safety support systems.

Is it possible to work Pytorch on this platform?

Thanks,  
Ram.

---

<div class="post-metadata">

**Author:** ![CarlosSarria](https://sea1.discourse-cdn.com/flex015/user_avatar/forums.imgtec.com/carlossarria/32/263_2.png) [@CarlosSarria](https://forums.imgtec.com/u/CarlosSarria)\
**Post date:** [September 30, 2019, 12:51pm UTC](https://forums.imgtec.com/t/powervr-gpu-gx6650-supports-pytorch/3082/4 "2019-09-30T12:51:34Z")

</div>

Hi Ram,

Although R-CarH3 is a very capable SoC, it is still a mobile part (ARM CPU) and it is not at the level of desktop CPU/GPU capable of running a DNN Framework directly. As far as I know there is not support for aarch64 in Pytorch. You could try building it from source. We can not give you guidelines on how to do this.

Regards.

Carlos.