We are keenly expecting the keynote speech from Nvidia as they undoubtedly are going to be talking a fair bit about Volta somewhere in the speech. SK Hynix has given us the first real information on their next generation GDDR6 memory tech at GTC 2017.
They will be showcasing full fledged Wafers and so on. And they are going to be doing at Nvidia GTC meaning that Nvidia will be among the first to leverage this new memory architecture.
This means we have got a bit of partnership going on between the SK Hynix and Nvidia.
Nvidia has been using GDDR5X up until now obviously any future chips are going to be making these shift to GDDR6 and we have seen pretty nice results with 5X memory with their work with Micron, they have managed to top out HBM and they are hoping at least to repeat history with HBM2 and to top that out again with GDDR6 and of course it will only top it out in terms of gains and performances and so on that you get from it.
In terms of speed you are looking at quite a significant jump as it has gone from 12 GigaBits per second to 16 and this is pretty impressive when you consider the fact that the Titan XP has about 547.7 GB/s using 11.4 Gbps GDDR5X.
HBM2 has pretty impressive figures itself and obviously Vega is coming along soon from AMD.
But depending on this specifications of Volta this does actually mean that Volta could actually be aligned to outdo Vega when it comes to memory bandwidth which obviously isn’t everything but it is a very important component.
It’s not just in terms of raw bandwidth or anything like that are also seeing savings in terms of power. For example: GDDR6 has a 10% power saving over the current Gen 5 memory chips and lower voltage with twice the bandwidth.
The comments that were made by the CEO which basically say that they are not concerned about Vega, they are not concerned about what AMD are doing.
Our assessment is that the competitive position is not going to change so they are pretty confident that Vega as good it is looking is not going to be a concern for them over at Nvidia.