Mainly for write buffering for optimal NAND wearing. Without DRAM, the buffer is done via slower SLC cache on the NAND packages themselves. This matters for performance when when write intensive loads or mixed read/write/erase loads. For end users which fall into that category are mainly video production where they are filming lots of content and various server workloads. For the majority of users, *IF* that SLC cache is big enough, it'll be fine for daily performance. That big IF in the previous sentence is dictated by the SSD vendor and the NAND they use, not inherently the controller.
Surprised the article didn't touch on the energy efficiency gains that this controller should have. Besides having half the channels to feed and no DRAM to worry about, moving to TSMC's 7-nm process is a HUGE jump over the previous 12-nm node. I would think this controller would be just fine running full tilt with a small heatspreader, which can't be said for the current Phison E26 SSDs.
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
Facebook
Twitter
RSS
6 Comments
Back to Article
nandnandnand - Tuesday, June 11, 2024 - link
Give me DRAM or give me death.Shouldn't these controllers be supporting 16 TB soon? The density needed is basically available.
https://www.tomshardware.com/pc-components/storage...
Tom Braider - Tuesday, June 11, 2024 - link
I am curious about the purpose of a DRAM on a fast NVMe drive. What gains are there to be explored?meacupla - Tuesday, June 11, 2024 - link
Probably the same effect. Better random and sustained performance. Potentially less wear on NAND.Kevin G - Wednesday, June 12, 2024 - link
Mainly for write buffering for optimal NAND wearing. Without DRAM, the buffer is done via slower SLC cache on the NAND packages themselves. This matters for performance when when write intensive loads or mixed read/write/erase loads. For end users which fall into that category are mainly video production where they are filming lots of content and various server workloads. For the majority of users, *IF* that SLC cache is big enough, it'll be fine for daily performance. That big IF in the previous sentence is dictated by the SSD vendor and the NAND they use, not inherently the controller.NextGen_Gamer - Tuesday, June 11, 2024 - link
Surprised the article didn't touch on the energy efficiency gains that this controller should have. Besides having half the channels to feed and no DRAM to worry about, moving to TSMC's 7-nm process is a HUGE jump over the previous 12-nm node. I would think this controller would be just fine running full tilt with a small heatspreader, which can't be said for the current Phison E26 SSDs.Terry_Craig - Tuesday, June 11, 2024 - link
Even premium(expensive) SSDs use 12nm controllers and LPDDR4, I wonder what would happen if they used the best (5nm + LPDDR5x) :P