0

I have a flashed a MacPro5,1 running macOS High Sierra 10.13.6 and a Geforce GTX 760 GPU with the latest supported CUDA library installed at a compute capability 3. Mathematica 11 uses the GPU quite nicely for deep learning (6X speedup well outpacing my newer Mac mini), but I'd like to look at using Python or similar ideally using TensorFlow, Keras or Pytorch, since Mthematica 11's DL support isn't great. Installing the latest versions of theses libraries obviously is not going to work since they now need a higher compute capability.

Are there any archived versions of these libraries or other deep learning tools that will install successfully?

If no such are available would it be worth upgrading from a 16 GB 4 core box to a 128 GB 12 core box, dropping the GPU and run TensorFlow or such in core only?

David Anderson
  • 40,244
Nick
  • 133
  • What is a Mac 5,1? – David Anderson Apr 19 '23 at 21:14
  • 1
    @DavidAnderson - one of this family… as they're the only ones that can fit the other criteria - https://everymac.com/systems/apple/mac_pro/specs/mac-pro-twelve-core-3.06-mid-2012-westmere-specs.html – Tetsujin Apr 20 '23 at 06:10
  • Nick - I have no idea on the maths side, but you can pick up 3.46 GHz 12-core 64GB 5,1s on eBay etc for under £500 these days. 128s are rarer, 96 might be faster [it will triple channel] but equally rare because the 16GB chips are rare in themselves. 8's are everywhere. If you have money to spend on a GPU, http://www.macvidcards.com/store/c1/Featured_Products.html is probably where to spend it [I didn't, I just have AMD HD 7950s in the Pros here, on Mojave. NVidia are limited to High Sierra] – Tetsujin Apr 21 '23 at 15:44

0 Answers0