Jen-Hsun Huang, CEO of NVIDIA, explains why they use GDDR6 memory instead of HBM2

During a question and answer event, Jen-Hsun Huang, CEO of NVIDIA, explains why they use and will use GDDR6 memory instead of HBM2.

The latest NVIDIA graphics cards are characterized, among other things, by implementing GDDR6 memory. Very different is the case of Vega 10 and Vega 20 from AMD that implements HBM2 memories. Jen-Hsun Huang, CEO of NVIDIA, has been asked about this fact. It is one of the big doubts, relatively, since GDDR6 is cheaper than HBM2. If the latter have some major improvements over GDDR6. It seems that NVIDIA have bet more to control costs than for other aspects.

Huang explains why they have opted for GDDR6 over HBM2

During a question and answer session you have been asked directly about the use of HBM2 in the future. Huang stressed that there is nothing wrong with the HBM2 memory itself, but the GDDR6 is cheaper. It should be noted that NVIDIA uses the HBM2 in the Voltas, intended for AI and Deep Learning.

Huang: It is much more expensive for consumer products. But memory prices will go down. There is nothing wrong with HBM2. I love HBM2. But I love GDDR6 (graphics memory) a lot more. "

Currently the GeForce RTX based on the Turing architecture make use of GDDR6 memory. These memories increase the bandwidth considerably compared to GDDR5. The HBM2 in turn have more bandwidth than GDDR6 memories. The differential factor is that GDDR6 is cheaper than HBM2, but despite everything GDDR6 is almost 70% more expensive than GDDR5.

It would be this price difference between GDDR5 and GDDR6 that has made RTX more expensive. In addition to influencing the factor of implementing NVLink, Tensor Cores and RT Cores.