Notebookcheck Logo

Micron is bringing GDDR6 memory to non-GPU products

Micron's GDDR6 are significantly slower than the ones from Samsung and Hynix. (Source: HotHardware)
Micron's GDDR6 are significantly slower than the ones from Samsung and Hynix. (Source: HotHardware)
Micron and its partners want to improve the adoption rate of GDDR6 chips, making them compatible with non-GPU products like AI neural nets, car image processors and networking devices.

Samsung has already announced plans to launch the GDDR6 memory later this year, and sources have confirmed the inclusion of the new memory standard in Nvidia’s GTX 2000 series that should be released at some point in 2018. However, Samsung is not the only GDDR6 manufacturer, as Micron also stated that its GDDR6 chips are scheduled for release in 2018. Micron’s GDDR6 chips are slower compared to the Samsung ones (14 Gb/s versus 18 Gb/s), come in smaller capacities (1 GB vs 2 GB) and use the 16 nm manufacturing process instead of 10 nm. They do require 1.35V, just like Samsung’s counterparts, but Micron’s slower chips are not actually intended for GPU integration. According to Micron, three of its IP partners – Rambus, Northwest Logic and Avery Design – are working on memory controllers that will allow a more diverse use for the GDDR6 standard.

Presumably, Micron will release high performance GDDR6 chips for GPU at a later date, but right now, the company is concentrating on increasing the adoption rate of the sixth gen chips. AMD and Nvidia have been using this type of memory since the first generation and they have built their own controllers, but Micron and its partners intend to provide controllers for a wide array for applications that include AI computing, vehicular image processing and networking devices.

For now, Micron’s catalog lists the new GDDR6 chips as sampling, with no specific ETA. The performance boost that Micron is trying to bring to other devices is commendable, but it is still unclear how big of an impact this would really have over the DDR4 RAM currently in use.


Read all 1 comments / answer
static version load dynamic
Loading Comments
Comment on this article
Please share our article, every link counts!
Bogdan Solca, 2018-01-24 (Update: 2018-01-25)