News for hackers. Stuff that matters.
Re:Google supercharges machine learning tasks with TPU custo... by fiatmoney on Wednesday May 18, 2016 @03:43PM (#11725198)
"General purpose" isn't that general, if you look at the actual operations they support and their threading model. It's already fairly optimized for these sorts of operations, and this amount of claimed headroom makes me suspicious.
Re:Google supercharges machine learning tasks with TPU custo... by Symmetry on Wednesday May 18, 2016 @06:12PM (#11726427)
Google has a lot of potential options that NVidia doesn't have. They can size their cache heirarchy to the task at hand. They can partition their memory space. They can drop scatter/gather. They can gang ALUs into dataflows that they know are the majority of machine learning workloads. They can partition their register file at the ISA level or maybe even drop it entirely. They can drop the parts of the IEEE754 floating point spec they don't need and they can size their numbers to the precision they need.
Re:Google supercharges machine learning tasks with TPU custo... by azinman2 on Wednesday May 18, 2016 @03:52PM (#11725283)
The fact that I can compile arbitrary programs for the GPGPU means it is general purpose. NVIDIA isn't writing softmax or backprop into silicon as a CPU instruction.

Look at how much faster ASICs for bitcoin mining are than the GPU... orders of magnitude.

Re:Google supercharges machine learning tasks with TPU custo... by fiatmoney on Wednesday May 18, 2016 @03:59PM (#11725359)
"Backprop" isn't even close to something that would be a "CPU instruction", it's an entire class of algorithm. It's like saying "calculus" should be a CPU instruction. Matrix multiplication & other operations, on the other hand, do neatly decompose into such instructions, which have been implemented by NVidia et al., since that's the core set of functionality they've been pushing for like a decade now.

Additional die space on additional functionality might hurt the power envelope (which is where the focus on performance / watt rather than performance kicks in) but it doesn't make your chips slower per se.

Re:Google supercharges machine learning tasks with TPU custo... by agumonkey on Wednesday May 18, 2016 @04:10PM (#11725449)
That was my impression too. ML under the hood was a lot of linear algebra, not very different than most shaders. But maybe Google decided to hardcode a few important ML primitives because the ROI was that good in terms of grabbing customers. Also they might have very large scale applications not found elsewhere that motivates this.
Re:Google supercharges machine learning tasks with TPU custo... by azinman2 on Wednesday May 18, 2016 @05:16PM (#11726037)
Ok I was obviously oversimplifying things but my point is since we can only speculate, it's clear that when you know specific algorithms/math operations/memory layouts/applications you want to optimize for you can create dedicated chips that optimize and do that quickly. That bitcoin miners are all dedicated chips and run circles around GPUs demonstrates exactly this fact.

Furthermore the fact that ML can be error tolerant means you also get to optimize certain floating point operations for speed or energy efficiency at the cost of accuracy. NVIDIA doesn't get to do this in their linear algebra support.

Re:Google supercharges machine learning tasks with TPU custo... by vintermann on Thursday May 19, 2016 @02:19AM (#11728385)
Bitcoin mining is an extremely well-defined task compared to machine learning. It remains to be seen how general these TPUs are in practice - whether they will support the neural network architectures common two years from now.
Re:Google supercharges machine learning tasks with TPU custo... by agumonkey on Wednesday May 18, 2016 @06:40PM (#11726609)
tbh I felt like realizing what you meant earlier at the end of my comment. I should have ps'd it.
Re:Google supercharges machine learning tasks with TPU custo... by emcq on Wednesday May 18, 2016 @10:14PM (#11727475)
If they balance compute to memory better than GPUs, you could definitely see a 10x. GPUs have large off chip memory and small caches (like 256kb). Cost to going to off chip memory can be 1-2 orders of magnitude more than on chip memory. You can certainly fit 4+MB on modern processors, but they likely bought designs from a company like Samsung because designing high performance, low power memory cells is tricky. I'm surprised they were able to keep things a secret.