Intel, Micron QLC Flash Yields "Less than 50%"

Megalith

24-bit/48kHz
Staff member
Joined
Aug 20, 2006
Messages
13,000
QLC flash is expected to usher in affordable SSDs, but the technology has hit a major roadblock at Intel and manufacturing partner Micron: TweakTown claims the current yield for 64-layer QLC is “hovering right around 48%,” a depressingly stark contrast from TLC’s 90% yield. A source believes the yield may not even improve, as resources are shifted to more profitable alternatives.

A source close to the situation stated the current yield for 64-layer QLC is hovering right around 48%. Less than half of the die manufactured are reliable enough to use in an SSD. In contrast, current 64-layer TLC from IMFT has a yield right around 90% as of today. Usually companies work to increase yields over time but the push to finish more profitable 96-layer flash may divert engineering resources away from 64-layer QLC.
 
Well of course they want low yields, they want that money honey! It's all a scam.

God, please dear god let the Chinese steal all of the IP and mass produce cheap memory.

I've been using DDR4 for a while now. It can't be that complicated. I could give a rats ass if my memory uses 0.010 less electricity.
 
Well of course they want low yields, they want that money honey! It's all a scam.

God, please dear god let the Chinese steal all of the IP and mass produce cheap memory.

I've been using DDR4 for a while now. It can't be that complicated. I could give a rats ass if my memory uses 0.010 less electricity.

.... The higher the yield, the higher the profit margin. A wafer costs just as much to make whether or not any of the die survive the process.
 
No 'Q' bits for you!

Don't mind this so much. Pretty much over my TLC phobia. Almost.
 
Well of course they want low yields, they want that money honey! It's all a scam.

God, please dear god let the Chinese steal all of the IP and mass produce cheap memory.

I've been using DDR4 for a while now. It can't be that complicated. I could give a rats ass if my memory uses 0.010 less electricity.
It says right in the article, too. Avoid that cheap sweet spot and go for that high profit mainstay. "...the push to finish more profitable 96-layer flash may divert engineering resources away from 64-layer QLC."
 
.... The higher the yield, the higher the profit margin. A wafer costs just as much to make whether or not any of the die survive the process.
I kinda doubt that's the only factor of price in wafer fab world. How about competition yield as well? How about if you are in the position to hold rights on the new incumbant standard process for the next 4-7 years? The price is merely chosen, then.
 
Well of course they want low yields, they want that money honey! It's all a scam.

God, please dear god let the Chinese steal all of the IP and mass produce cheap memory.

I've been using DDR4 for a while now. It can't be that complicated. I could give a rats ass if my memory uses 0.010 less electricity.
Right, they developed a process they want to fail to cost themselves money just to charge the consumer more. Do pizzarias throw out half a pie so they can charge more for the remaining half? Try thinking things through before angrily spouting off.

96L yields higher bit growth (50%) than 64L QLC (33%). They will take lessons learned from 64L QLC and apply them to 96L QLC.
 
  • Like
Reactions: risc
like this
Good, don't want this shit. I'm glad the race to the bottom has hit a small roadblock.
 
Can someone explain to me please what the issue is with RAM pricing? I had a 2011 rig like three years ago...if not more than that...with a 4x4GB DDR4-2666 RAM kit that cost like $80 something dollars...now this shit is like twice that and it's been YEARS.

What is the problem? What were they doing then that isn't happening now?
 
Can someone explain to me please what the issue is with RAM pricing? I had a 2011 rig like three years ago...if not more than that...with a 4x4GB DDR4-2666 RAM kit that cost like $80 something dollars...now this shit is like twice that and it's been YEARS.

What is the problem? What were they doing then that isn't happening now?
Demand growth has far exceeded supply growth due to AI, AV, cloud infrastructure spending, increased memory content in mobile devices, IoT, etc. With only three companies manufacturing almost 95% of DRAM now there's no more undisciplined capex spending to try to take market share which results in pricing crashes every couple of years due to periods of overcapacity (and resulting bankruptcies and/or purchases by other companies for pennies on the dollar - how do you think we got to only three main DRAM companies?). The children will call this "price fixing" or "gouging" but they're ignorant of both terms.

The consumer chooses to view this period as "excessively priced" when they should have viewed the the previous periods as excessively cheap (i.e. when many of the DRAM companies were operating at a loss - not sustainable). Node shrinks yield less and less bit growth for more and more money, making increased capex less efficient, again reducing the desire for increased capex spend to increase capacity. Samsung, SK Hynix, and Micron are all focusing on raising capacity in line with demand growth as it's in their interests to keep pricing elevated while still supplying enough DRAM to the market. You'll note that the companies continually announce that any new fab space is to replace lost wafer starts due to more complex and time consuming nodes (i.e. it takes longer to complete a wafer).

For those claiming DRAM is a commodity, one need only look at the poor yields on previous generation nodes China is experience even with their state-sponsored billions of dollars in investments to see this commodity isn't so easy to just jump in and produce. China isn't expected to have a significant effect on the market until mid 2020s, maybe 2030.

TL;DR: DRAM will not become significantly cheaper any time soon.
 
Back
Top