Hi guys anandtech has a new article on HBM 2 from samsung well worth a read its these chips will be coming to next gen gfx cards from AMD and Nvidea at 14/16 nm
Key points in relation to GPU memory
Samsung's 4-Stack HBM2 based on 8 Gb DRAMs
Total Capacity 16 GB
Bandwidth Per Pin 2 Gb/s
Number of Chips/Stacks 4
Bandwidth Per Chip/Stack 256 GB/s
Effective Bus Width 4096-bit
Total Bandwidth 1 TB/s
Here is the link to full article
http://www.anandtech.com/show/9969/jedec-publishes-hbm2-specification
key points
No hotlinking images, re-host them yourself next time - Davey
Its a interesting read thought I would share looks like 16 gb is going to be the mem size for next gen flagship cards if a 4 stack is used. If samsung are the maker or they use a standard JEDEC layout from another manufacturer
We may see Hynix or Micron release a larger stack for aftermarket board partners opening up 32 and 48 and 64 gb versions but looks like standard is going to be 16gb
please discuss and get your geek on
Key points in relation to GPU memory
Samsung's 4-Stack HBM2 based on 8 Gb DRAMs
Total Capacity 16 GB
Bandwidth Per Pin 2 Gb/s
Number of Chips/Stacks 4
Bandwidth Per Chip/Stack 256 GB/s
Effective Bus Width 4096-bit
Total Bandwidth 1 TB/s
Here is the link to full article
http://www.anandtech.com/show/9969/jedec-publishes-hbm2-specification
key points
No hotlinking images, re-host them yourself next time - Davey
Its a interesting read thought I would share looks like 16 gb is going to be the mem size for next gen flagship cards if a 4 stack is used. If samsung are the maker or they use a standard JEDEC layout from another manufacturer
We may see Hynix or Micron release a larger stack for aftermarket board partners opening up 32 and 48 and 64 gb versions but looks like standard is going to be 16gb
please discuss and get your geek on
Last edited: