4 ms·

I wish they had a range that was affordable for a hobbyist. You can buy a cheap Nvidia card to "get your feet wet". I would like to play with these things.

dogma1138 10y ago
Intel was selling the Xeon Phi 31S1P for under 200$ (it's back to 500$ now) for a limited time. They will likely to have cheap version and promotions this time around too.
imaginenore 10y ago
Why not use Google's platform, which runs on their new custom chips (Tensor Processing Unit)?
- vonnik 10y ago
  Why use it, if it locks you in?
  - xadhominemx 10y ago
    If you're a hobbyist, why would you care about lock in?
    - flamedoge 10y ago
      Because you don't have control over what Google does. They may kill TPU altogether leaving your work irrelevant.
      - hyperbovine 10y ago
        Hobbies come, hobbies go.
      - willvarfar 10y ago
        You don't have control over what nvidia, intel or the rest of them either. If you want to get your 'feet wet', then why bother?
        flamedoge 10y ago
        Because they have client facing products and backward compatibility is important to clients.
    - lmm 10y ago
      I care about being able to carry on doing my hobby stuff 5 or 10 years from now. With NVidia I can at least be confident that as long as my graphics card keeps working (which feels like something under my control, unlike Google shutting down their products) I can keep running my code on it.
- visarga 10y ago
  Because it's not made to accelerate training, just inference. The TPU is an 8-bit fixed point processor less power hungry than GPUs, so it won't help research, only deployment for large projects, running in the cloud.
rch 10y ago
Keep an eye on Colfax - they've had some nice deals in the past: http://www.colfax-intl.com/nd/ http://www.colfax-intl.com/nd/
marmaduke 10y ago
A recent multicore i7 (e.g. 4 core Haswell, 8 way SIMD for single precision = 32 threads) is enough to prototype OpenCL code which you can then run on larger CPUs or GPUs.

dogma1138 10y ago

Intel was selling the Xeon Phi 31S1P for under 200$ (it's back to 500$ now) for a limited time. They will likely to have cheap version and promotions this time around too.

imaginenore 10y ago

Why not use Google's platform, which runs on their new custom chips (Tensor Processing Unit)?

vonnik 10y ago
Why use it, if it locks you in?
- xadhominemx 10y ago
  If you're a hobbyist, why would you care about lock in?
  - flamedoge 10y ago
    Because you don't have control over what Google does. They may kill TPU altogether leaving your work irrelevant.
    - hyperbovine 10y ago
      Hobbies come, hobbies go.
    - willvarfar 10y ago
      You don't have control over what nvidia, intel or the rest of them either. If you want to get your 'feet wet', then why bother?
      - flamedoge 10y ago
        Because they have client facing products and backward compatibility is important to clients.
  - lmm 10y ago
    I care about being able to carry on doing my hobby stuff 5 or 10 years from now. With NVidia I can at least be confident that as long as my graphics card keeps working (which feels like something under my control, unlike Google shutting down their products) I can keep running my code on it.
visarga 10y ago
Because it's not made to accelerate training, just inference. The TPU is an 8-bit fixed point processor less power hungry than GPUs, so it won't help research, only deployment for large projects, running in the cloud.

rch 10y ago

Keep an eye on Colfax - they've had some nice deals in the past: http://www.colfax-intl.com/nd/ http://www.colfax-intl.com/nd/

marmaduke 10y ago

A recent multicore i7 (e.g. 4 core Haswell, 8 way SIMD for single precision = 32 threads) is enough to prototype OpenCL code which you can then run on larger CPUs or GPUs.