2 ms·
Going by the specs, I would expect that training current models on the new Titan X will be be 1.3x to 2.0x faster than with the older Titan X. In practice, thi
by cs702 10y ago
Going by the specs, I would expect that training current models on the new Titan X will be be 1.3x to 2.0x faster than with the older Titan X. In practice, this means deep neural nets that now take four weeks to train will take two or three weeks instead. That's great news.