Addition of GPU's

#1

It seems that adding GPU’s to the TF Grid is being discussed as a possible future development. It would be good to get a bit more information from those with better insight.

  1. What applications would the GPU’s be for?
  2. Do we know the profile of teh customers who might require them and use cases?
  3. Any idea which GPU’s are being considered?
  4. When might this roll-out happen (if at all)?

It could be useful information for both farmers planning future capacity addition and conversations with potential partners/end clients who might want to use them.

1 Like
#2

@colin, I’ll try to answer this partly :slightly_smiling_face:

In general, applications requiring intensive computation for graphics so gaming has to be one of them for sure.

In healthcare, medical devices that need the ability to perform real-time and advanced imaging. For 3D designing and architectural rendering tools used by civil engineers and architects.

In addition to that, GPU mining is also a strong candidate and could be lucrative. With reference to mining, here is a good link that should give you some insight to GPU mining. This link would also give you ranking by profitability for GPUs !

For question 4, if this is ever going be rolled out in ThreeFold ? This has to be answered by the management’s top tier :slightly_smiling_face:

1 Like
#3

One addition to this: GPUs are also very popular in artificial intelligence, for training deep learning models (ML).

Following link provides some insight in this:

2 Likes
#4

I would like to add, that right now there are no affordable GPU providers. And many machine learning frameworks, like Keras or PyTorch provide infrastructure to speed up neural network training up x100 using GPU. Also, they provide support for using a pool of GPUs. Such feature is provided by Google Cloud for example (I mean pool of GPUs), but prices are very high. ThreeFold could be such an affordable provider of GPUs.

1 Like
#5

Small update. I have checked the Golem project. There is a demand for GPU usage there https://github.com/anshuman73/DeML-Golem . But there are a lot of questions to be answered in Golem model for the scalability of the long-running solutions. TF seems more promising from the architectural point of view.

2 Likes