AMD have recently clued in the public with its 7-year long innovation High Bandwidth Memory interface for its flagship graphics processing unit Radeon  R9 390X.  AMD Chief, Joe Macri didn't hold back in pointing out Nvidia's delay.

Macri explained how a High Bandwidth Memory works the same way as today's urban development. Its traditional GDDR5 is designed to occupy much space with memory chips surrounding the GPU. HBM changes this by conserving space through vertically stacked memory chips and connecting one stacked chip to the next with less wiring and therefore less used energy. In comparison, a GDDR5 chip supports a 32-bit wide memory bus with 7GBps of bandwidth while a stack of HBM RAM supports a 1024-bit wide bus with approximately 125GBps of memory bandwidth. It also has lower clock speed while having incomparable memory bandwidth. Aside from the HBM fundamentals, Macri also points out that:

Nvidia creates PowerPoints and talks in advance like they are the wonderful leader of everything," Macri said. "While they're talking, we're working."

Meanwhile, sources say that Nvidia is preparing to release its forthcoming GeForce GTX 980 Ti graphics card on the Computex trade show. Some speculations say that the new GeForce version will be similar to Titan X, Nvidia's current flagship GPU. The main difference would be its lesser VRAM of 6GB and lower clock speed. It's also expected to sell between $550 and $1000 or approximately $650 while AMD's new Radeon GPU could be at the higher end of the price spectrum.

Latest buzz on AMD Radeon R9 390X say that it's to be launched with 4GB of HBM and will retail for $849 while Fiji XT-based graphics card will have 8 GB of HBM.

Nvidia has been more popular than AMD for quite a while and perhaps it's true that its upcoming Radeon flagship will steal the spotlight. There hasn't been much heard from Nvidia lately or perhaps they are also doing their homework while AMD is taking its turn to talk.