# GPU Hardware acceleration support

**URL:** <https://forum.clockworkpi.com/t/gpu-hardware-acceleration-support/483>\
**Category:** GameShell\
**Created:** [July 10, 2018, 7:17am UTC](https://forum.clockworkpi.com/t/gpu-hardware-acceleration-support/483 "2018-07-10T07:17:16Z")\
**Posts on this page:** 5\
**Page:** 1

<div class="post-metadata">

**Author:** ![Marc\_Sauber](https://avatars.discourse-cdn.com/v4/letter/m/278dde/32.png) [@Marc\_Sauber](https://forum.clockworkpi.com/u/Marc_Sauber)\
**Post date:** [July 10, 2018, 7:17am UTC](https://forum.clockworkpi.com/t/gpu-hardware-acceleration-support/483/1 "2018-07-10T07:17:16Z")

</div>

Hi all, i just wonder what the plan about the GPU hardware GLES support for the GS ?  
What’s the roadmap about that ?  
Thanks

---

<div class="post-metadata">

**Author:** ![macario](https://yyz1.discourse-cdn.com/flex029/user_avatar/forum.clockworkpi.com/macario/32/14564_2.png) [@macario](https://forum.clockworkpi.com/u/macario)\
**Post date:** [July 10, 2018, 9:25am UTC](https://forum.clockworkpi.com/t/gpu-hardware-acceleration-support/483/2 "2018-07-10T09:25:06Z")

</div>

The work by NextThingCo to achieve this on the PocketCHIP with a debian toolchain should come in helpful: [https://github.com/NextThingCo/chip-mali](https://github.com/NextThingCo/chip-mali)

---

<div class="post-metadata">

**Author:** ![fufu](https://avatars.discourse-cdn.com/v4/letter/f/50afbb/32.png) [@fufu](https://forum.clockworkpi.com/u/fufu)\
**Post date:** [July 13, 2018, 8:33pm UTC](https://forum.clockworkpi.com/t/gpu-hardware-acceleration-support/483/3 "2018-07-13T20:33:37Z")

</div>

is possible using it as cuda?  
for example to tensorflow?

---

<div class="post-metadata">

**Author:** ![Petrakis](https://yyz1.discourse-cdn.com/flex029/user_avatar/forum.clockworkpi.com/petrakis/32/256_2.png) [@Petrakis](https://forum.clockworkpi.com/u/Petrakis)\
**Post date:** [July 14, 2018, 1:19am UTC](https://forum.clockworkpi.com/t/gpu-hardware-acceleration-support/483/4 "2018-07-14T01:19:26Z")

</div>

Cuda is nvidia only thing.  
And there are other arms for machine learning [https://www.theverge.com/2017/5/29/15707606/arm-cortex-a75-a55-mali-g72-specs-announced](https://www.theverge.com/2017/5/29/15707606/arm-cortex-a75-a55-mali-g72-specs-announced)

---

<div class="post-metadata">

**Author:** ![Godzil](https://yyz1.discourse-cdn.com/flex029/user_avatar/forum.clockworkpi.com/godzil/32/60_2.png) [@Godzil](https://forum.clockworkpi.com/u/Godzil)\
**Post date:** [July 15, 2018, 12:09pm UTC](https://forum.clockworkpi.com/t/gpu-hardware-acceleration-support/483/5 "2018-07-15T12:09:03Z")

</div>

If you are interested in ML look at what they’ve done at [http://jevois.org](http://jevois.org), they are using an allwinner chip and use the Mali for GPU acceleration .

But don’t expect tensorflow or caffe to run (well) on a device like the GameShell… it is really not made for (like all ARM devices including the RaspberryPi)