GDDR6 was designed in Micron's Munich Development Center and it is optimized for a variety of applications, including artificial intelligence (AI), networking, automotive and graphics processing units (GPUs).
Targeting up to 64GB/s in one package, GDDR6 brings a significant improvement over the fastest available GDDR5. However, the first GDDR parts currently in production are offering up to 14GB/s (1.25-1.35V). They all use industry-standard BGA packaging.
For the graphics market, Micron is releasing chips rated for 12Gbps and 14Gbps, both of which operate at 1.35v. This is down from 1.5-1.6v for GDDR5.
Despte the slower 12Gbps speeds in some chips, GDDR6 is offering a full 50% better bandwidth than the fastest GDDR5 bins. And 14Gbps GDDR6 represents a 17% increase in bandwidth over the fastest GDDR5X SKUs available today.
The biggest difference is between GDDR5X and GDDR6 comes down to the number of channels. Whereas both GDDR5 and GDDR5X used a single 32bit channel per chip, GDDR6 instead uses a pair of 16bit channels.
For the automotive and networking markets Micron is offering the same 8Gbit capacity at speeds of 10 and 12 Gbps.
Micron has also 16Gb parts in their development pipeline. In he labs, the company has also managed to get GDDR6 up to 20Gbps.
In order to deliver this high bandwidth memory technology, Micron is working directly with ecosystem partners in order to enable learning on both pre-silicon verification as well as validation. Prior to mass production of GDDR6 memory, Micron shipped early validation silicon to our ecosystem partners to accelerate engineering efforts behind validating intellectual property and build models and toolsets in the ecosystem and deliver board layout validation.
Micron GDDR6 memory solutions are on display at ISC 2018, June 24-28, in Frankfurt, Germany.
Korean SK Hynix has also made their next generation GDDR6 memory available to customers which can include graphics card manufacturers such as NVIDIA and AMD.