scottlegrand

5 ms·

1. 1.▲
   by scottlegrand 10y ago

   Some of us are... https://blogs.aws.amazon.com/bigdata/post/TxGEL8IJ0CAXTK/Gen...
2. 2.▲
   by scottlegrand 10y ago

   Or as Urs Hoezel would say: "advancing Moore's Law by 7 years(tm)..." Badum ba bum bum... And given Frank Seide et al. demonstrated 1-bit SGD in 2014 ( https://www.microsoft.com/en-us/research/public
3. 3.▲
   by scottlegrand 10y ago

   Here's your answer for GROMACS (and it sucks)... http://www.prace-ri.eu/IMG/pdf/wp120.pdf Even so, right now, little would please me more technologically than a competitive Xeon Phi offering, but while KNL is
4. 4.▲
   by scottlegrand 10y ago

   I prefer to look at it the other way, why don't you point out an existing and important chemistry application where KNL bested its contemporary GPUs, say the best of Knight's Corner versus the best of Kepler (K40 or K80). I'
5. 5.▲
   by scottlegrand 10y ago

   Except that up to now at least, CUDA IMO remains the best abstraction for programming multi-core: subsuming away multiple threads, SIMD width, and multiple cores into the language definition. AMBER ( http://www.ambermd.org ) lite
6. 6.▲
   by scottlegrand 10y ago

   Not even wrong. I have two PCs with 4 Titan X (Maxwell) GPUs and a third PC with 4 Titan X (Pascal) GPUs. Both of these systems are available today (I built them myself, total BOM about $7K), and both will destroy 4 Xeon Phi servers at De
7. 7.▲
   by scottlegrand 10y ago

   Yep, NVIDIA. Nervana's dedicated ASIC will deliver 55 (mostly) int16 TOps in 2017. In contrast, the two Titan XP GPUs I bought last week for a total of $2400 deliver 44 such TOps. Next year, a single Volta GPU will deliver at least 3
8. 8.▲
   by scottlegrand 10y ago

   I figured Nervana was mostly dead if they were stuck at 28 nm I figured Intel was down and out in Santa Clara without a strong deep learning play That all changed today. Intel has been bouncing all over the place trying to break into deep l
9. 9.▲
   by scottlegrand 10y ago

   Not to mention they all solve problems for which the training sets lie on low-dimensional manifolds within a very high-dimensional space. And this brings about arbitrary failures when one goes out of sample and it also serves as the basis
10. 10.▲
    by scottlegrand 10y ago

    Absolutely 100% agree, but at the same time, I think we will ultimately need to build and evaluate models that can span the memory of more than one processor. I don't think a single GTX Titan X, GTX 1080 or even a server is enough her
11. 11.▲
    by scottlegrand 10y ago

    Thanks for that! And boy, I wish I had the resources the TensorFlow team has to build standards like this and also to write their own custom CUDA compiler. I do want the multi-dimensional indexing for RNN data though. Maybe support HDF5 d
12. 12.▲
    by scottlegrand 10y ago

    It's more than that, and it's in use in production at Amazon. 8 TitanX GPUs can contain networks with up to 6 billion weights. As Geoffrey Hinton once said: "My belief is that we’re not going to get human-level abilities un
13. 13.▲
    by scottlegrand 10y ago

    Lead author of DSSTNE here... 1. DSSTNE was designed two years ago specifically for product recommendations from Amazon's catalog. At that time, there was no TensorFlow, only Theano and Torch. DSSTNE differentiated from these two fra