I don't see anything bad with this slide honestly =) as it's about halp flot FP16 performance & then about other thigns including nVLINK..who said this is going to be 10X faster on SP (FP32)?coilbook wrote:bvghn
new, huge market for nvidia & not only is this early AI, deep learning & stuff along those lines. That's where that new devices for 120k$ with 8x Pascal chips under the hood Nvidia introduced during GTC shines & that's where money goes..where new steams of income for this company is ging to be (if that lifts off i nthe future as somethig we could not live without).
Nvidia game market (& we should consider being part of it as we use GTX cards) rolls on it's own. This VR thing is going to push the need for more powerfull devices, but so far nvidia maeks that happen with a lot of software optimisations. 1080 seems to be powerfull card, but bigger unit should be even more inpressive..question when it's going to come.
updating from older cards liek 980, or even 980 might make sense if You value amount of vram. That power usage alone is something to wow about. But it might not make too much of sense for anyone who has 980Ti or TitanX (those are great SKUs & at least 12GB card is here to stay for at least half a year =) Think in these days Guys expect too much from technological progress..it's hapening, but not that fast.