News Kioxia Researchers Demo Hepta-Level Cell NAND Flash, Nearly Doubling the Capacity of QLC

[Admin]

Kioxia researchers have successfully built a new NAND flash technology with the potential of having nearly twice the capacity of QLC NAND flash, while being significantly cheaper to produce.

Kioxia Researchers Demo Hepta-Level Cell NAND Flash, Nearly Doubling the Capacity of QLC : Read more

 

[cyrusfox]

Jump from 4 to 7 layers! if single layer is 2^1 2 voltage states, QLC is 2^4 is 16 voltage states, we are looking at 2^7 128 distinct voltage states... That is insane and Awesome,
wonder how many cycles one should expect before this can be marketed, 50? 100??
Either way I am looking forward to affordable 20tb SSD 🧀

 

[Kamen Rider Blade]

I'm really curious as to the P/E (Program / Erase) cycle life-span.

It's only going to get worse with the more bits we pack in.

 

[lorfa]

Nope, can't deal with it being 7, must be 8 so divisible by two. OCD won't tolerate it otherwise. (In my defense one byte per cell does sound kind of kewl)

 

[InvalidError]

Current 16TB SSDs are ~$6000. Halve that, it is still ~$3000. A 16TB HDD on the other hand can be had for ~$300, 10X cheaper than the hypothetical 7-bits-per-cell SSD.

I'd be a little nervous about storing data on an SSD array that is perpetually one LN2 shortage away from spontaneously erasing itself.

 

[Alvar "Miles" Udell]

Wasn't the vaporware X-NAND supposed to solve this problem?

 

[usertests]

Nope, can't deal with it being 7, must be 8 so divisible by two. OCD won't tolerate it otherwise. (In my defense one byte per cell does sound kind of kewl)

I'll be shocked if there isn't some small performance advantage to organizing flash into 1 byte per cell. Maybe not enough to overcome the effects of moving from hepta- to octa-, but still...

 

[BillyBuerger]

Nope, can't deal with it being 7, must be 8 so divisible by two. OCD won't tolerate it otherwise. (In my defense one byte per cell does sound kind of kewl)

So no TLC NAND for you then? Must be hard to find a drive to use right now.

 

[bjk8kds]

This would be much slower sustained performance than QLC and less lifespan than QLC right?
QLC "real" speed are slower than HDD, which is sometimes avoided by consumers.
Then what's the point moving from HDD to this one?

 

[hotaru251]

This would be much slower sustained performance than QLC and less lifespan than QLC right?

correct.

why you got worse overall performance & durability (all things being same) as you went from SLC, MLC, TLC, & QLC.

they give up performance & durability for raw capacity and cheapness.

QLC "real" speed are slower than HDD

depends on how they are built really.

and iirc for just reading data QLC is actually better than TLC. So work for data storage where you arent writing just reading from. Would likely be even better as we go from QLC to 7bit a cell from 4.

Then what's the point moving from HDD to this one?

cost & Capacity in the long game.
SSD take up much smaller footprint than HDD. You can fit more into same physical space thus servers and the like benefit long term.

and personally I'd enjoy massive ssd (even if slow write) just for game library/backup drives. (and just copy/move data to faster drive when I want play it)

 

[InvalidError]

and personally I'd enjoy massive ssd (even if slow write) just for game library/backup drives. (and just copy/move data to faster drive when I want play it)

Needing to store your SSD in liquid nitrogen for data retention redefines cold data storage!

 

[usertests]

and personally I'd enjoy massive ssd (even if slow write) just for game library/backup drives. (and just copy/move data to faster drive when I want play it)

Just copy the whole game into RAM (you may need 192-256 GB of RAM).

 

[cyrusfox]

This would be much slower sustained performance than QLC and less lifespan than QLC right?
QLC "real" speed are slower than HDD, which is sometimes avoided by consumers.
Then what's the point moving from HDD to this one?

Only impact is on write speed, read speed would be comparable to TLC/QLC.
Manufacturers effectively hide the write speed penalty with psuedo SLC cache(until you fill that up then 40-200MB/s).

The main benefit over HDD is

1. much better response (Seek time order of magnitude faster)
2. Rugged/durability (no problems with vibration or taking drops)
3. Reliability for low write applications

7 Layer cells are likely in the double digits of endurance [10 to 99 cycles], but With improved manufacturing and more sensitive controllers this will improve. Just look at QLC which was first marketed by Intel @ 200 cycles (660p) in 2018 and now is capable of 400 cycles with Solidigm P41 Plus(2022) That is a doubling of endurance in 4 years. The worst in the market for QLC I saw was 100 cycles and 80 cycles for TLC (SN350 check it out).

For low write applications (media), you can arguably get away with much less endurance(Write Once Read Many WORM memory).

 

[jkflipflop98]

Raise your hand if you've ever ran a LN2 pump before!

Now imagine that in a PC case. . .

 

[npyrhone]

Current 16TB SSDs are ~$6000. Halve that, it is still ~$3000. A 16TB HDD on the other hand can be had for ~$300, 10X cheaper than the hypothetical 7-bits-per-cell SSD.

I'd be a little nervous about storing data on an SSD array that is perpetually one LN2 shortage away from spontaneously erasing itself.

This drive is not one LN2 shortage away from spontaneously erasing itself. It just becomes inoperable until the cooling solution is fixed.

 

[leclod]

Current 16TB SSDs are ~$6000. Halve that, it is still ~$3000. A 16TB HDD on the other hand can be had for ~$300, 10X cheaper than the hypothetical 7-bits-per-cell SSD.

Maybe, but a 2TB SSD drive costs less than 100.-

 

[InvalidError]

This drive is not one LN2 shortage away from spontaneously erasing itself. It just becomes inoperable until the cooling solution is fixed.

If QLC NAND needs to be refreshed every 60-90 days to guarantee data integrity, HLC would likely drop that to ~10 days. Leaving the drive "inoperable" isn't an option for long at room temperature unless they had a breakthrough in cell leakage mitigation too.

 

[Kamen Rider Blade]

7 Layer cells are likely in the double digits of endurance [10 to 99 cycles], but With improved manufacturing and more sensitive controllers this will improve. Just look at QLC which was first marketed by Intel @ 200 cycles (660p) in 2018 and now is capable of 400 cycles with Solidigm P41 Plus(2022) That is a doubling of endurance in 4 years. The worst in the market for QLC I saw was 100 cycles and 80 cycles for TLC (SN350 check it out).

For low write applications (media), you can arguably get away with much less endurance(Write Once Read Many WORM memory).

Oh great, so at 10-99 P/E cycles for 7-bit/cell, you have basically disposable media that is going to go into the landfill at very fast rates.

Even if you double the life-span to 20-200 P/E cycles, you're still creating storage that will become e-Waste in a short period of time.

If QLC NAND needs to be refreshed every 60-90 days to guarantee data integrity, HLC would likely drop that to ~10 days. Leaving the drive "inoperable" isn't an option for long at room temperature unless they had a breakthrough in cell leakage mitigation too.

I have people who forget that their data is on a SSD for years to decades on end.

If the data vanishes so soon, what's the point of the media if it doesn't have off-line longevity measured in years to decades?

Maybe we need to take a step back and stop at QLC and call it "Good Enough" and work on Longevity and raising the P/E cycles.

 

[InvalidError]

I have people who forget that their data is on a SSD for years to decades on end.

If the data vanishes so soon, what's the point of the media if it doesn't have off-line longevity measured in years to decades?

SSDs aren't intended for offline storage.

How long they will retain data depends on bitness per cell, fabrication technology, quality, wear, ambient temperature and likely a bunch more factors. For QLC, data can degrade beyond recovery within months at 40C. Keep your offline backup SSD in the fridge!

 

[Kamen Rider Blade]

SSDs aren't intended for offline storage.

How long they will retain data depends on bitness per cell, fabrication technology, quality, wear, ambient temperature and likely a bunch more factors. For QLC, data can degrade beyond recovery within months at 40C. Keep your offline backup SSD in the fridge!

Then what do you recommend for Ultra Long Cold Storage that can last for decades.

The only option I currently still use is Blu-Ray Discs which now use In-Organic Dyes.

 

[cyrusfox]

Then what do you recommend for Ultra Long Cold Storage that can last for decades.

M-disc they say, paper and microfilm for the rest You could always etch it in stone. Life is transient. The best backup strategy is multi-prong.