News for hackers. Stuff that matters.

sorenbouma

Re:Gabe Newell applies for NZ residency on Tuesday October 20, 2020 @05:32AM (#24835410)
Auckland already has one of the highest house price:median income ratios in the world.
Re:Google Coral Dev Board mini SBC is now available for $100 on Saturday October 17, 2020 @05:04PM (#24813359)
That benchmark appears to compare full precision fp32 inference on the nano with uint8 inference on the coral, that floor wiping comes with a lot of caveats
Re:Nvidia to Acquire Arm for $40B on Sunday September 13, 2020 @10:09PM (#24466055)
I'm a TPU user and I'd be interested to see a specific example of something that can be done on TPU but not GPU.

Perhaps I'm just not experienced enough with the programming model, but I've found them to be strictly less flexible/more tricky than GPUs, especially for things like conditional execution, multiple graphs, variable size inputs and custom ops.

Re:How much did AlphaGo Zero cost? (2018) on Thursday June 11, 2020 @07:28AM (#23487658)
I might be wrong, but I think this cost calculation is way off:

Their running cost estimate of a single TPU in a machine with 4 "TPUs" is based off the price of a cloud TPU v2-8, but a v2-8 is actually 4 ASICS on 1 board.

Also, because of the date of publication being around the time v2s were announced, and the fact that the TPU is only used for inference and GPU is used for training, I think self play was likely done on TPU v1s, which use 5x less power per ASIC and so are likely much cheaper

I also think the way they calculated the number of TPUs required is wrong, it looks like they assume 1 machine with 4 TPUs makes 1 move in 0.4 seconds, but since making 1 move only requires a forwards pass through a moderately sized CNN with 19x19(tiny) input, 1 TPU should be able to make thousands of moves in parallel per second.