NBF has an interesting story up about the upcoming Summit and Sierra supercomputers, which were referred to as "AI supercomputers" at the recent SC'16 conference. This is the relevant table.
The timeline given was late 2017/early 2018. It's likely that we will see a repeat of what happened with Pascal, of big full fat Volta only going to supercomputers first. Nevertheless, that being the case, we should see Volta being introduced to the world this coming GTC in April. Possibly even during CES now in January, if Huang can tie it into cars, although given past patterns GTC is still the more likely candidate.
If NV releases a GV102 GPU in late 2017 or early 2018, AMD would fall behind even more, given that their upcoming Vega GPU is likely to struggle against even the previous-gen Pascal GPU. It could be rendered insignificant within mere months if NV accelerates the rollout and releases a "halo GPU" similar to the TXP in conjuction with the supercomputer rollout.
The timeline given was late 2017/early 2018. It's likely that we will see a repeat of what happened with Pascal, of big full fat Volta only going to supercomputers first. Nevertheless, that being the case, we should see Volta being introduced to the world this coming GTC in April. Possibly even during CES now in January, if Huang can tie it into cars, although given past patterns GTC is still the more likely candidate.
If NV releases a GV102 GPU in late 2017 or early 2018, AMD would fall behind even more, given that their upcoming Vega GPU is likely to struggle against even the previous-gen Pascal GPU. It could be rendered insignificant within mere months if NV accelerates the rollout and releases a "halo GPU" similar to the TXP in conjuction with the supercomputer rollout.
Facebook
Twitter