SSDs Die, RAM Doesn't. Why?

Why does the NAND flash inside SSDs wear out, but typical RAM doesn't suffer from this problem?

FOLLOW US ELSEWHERE
---------------------------------------------------

[nathangoddard]

I understood the main differences, but this fills in the missing gaps in an easy to comprehend manner. Thanks for this outstanding video. Great work.

[ericwright]

The first ssd I ever bought back in 2010 is still functioning perfectly. I have another newer ssd with around 1PB of writes and it reports no problems. I've had one ssd die and it happened in the first week or two, got a free warranty replacement a week later. Overall ssd's are crazy robust.

Edited for clarity

[hjups]

So close... you almost covered the caveat. RAM can still fail due to wear-out like SSDs, however, this usually only occurs in datacenter settings where the DRAM modules are kept at a high utilization. In these cases, the failure rate is still quite low, but it is a problem that companies like Google and Meta are currently trying to overcome (and no, they are not letting their RAM modules get too hot or overvolting them). For the general consumer or enthusiast, the chance of them encountering such a failure is very low.
Also, you can wear out a SSD in read-only mode as well. There is a phenomena called read disturbance, where after continually reading from adjacent cells (floating gates), the voltage levels of that gate can diverge from the expected thresholds. To combat this, if a page is read too many times, the SSD controller will typically either relocate the page to another location in the flash or rewrite the current page (effectively refreshing it).

[dan_loup]

Just to add a thing, there's two basic types of RAM, the DRAM described on the video, and SRAM, that is a type of memory that use 6-8 transistors to logically store the bits.
SRAM is insanely fast as it's basically made out of logic, but it use 6-8 times the space of a DRAM per bit, which makes it 6-8 times more expensive.
Most modern computers use SRAM for caches like the L1/L2/L3 etc.. and DRAM for everything else.

[isaac]

I've written 30TB on my main SSD over like, 3-4 years. Doing pretty good so far, no need to replace it yet.

[santinojoshuatorre]

my very first SSD was a Samsung 840 EVO in 2013. I remember being so worried at that time about it wearing out (they were expensive AF that time) that I actually calculated its estimated life/cost vs. just getting another HDD.

Said SSD is still in use in my current PC as a cache drive for my game HDD and is the only thing i've carried forward from that old build. it's lasted three whole system rebuilds (Q8400, 3770k, 3700x) and i'll probably still bring it to my next one.

[RailfanSrikrishna]

I worked as laptop service engineer and one customer asked why all laptops parts cover replacement warrenty except RAM I told RAM is the last dying part in laptop

[brianfong]

SSDs have electricity "water dams" and the electricity breaks down the "water dams" over time.
Ram has no dams.

[arrowghost]

SSDs have been improved over time too since the early days, they were known to be a lot more fragile and easy to break down, and long before came M.2

[MotherGoose]

That was a great. Not only did you answer initial question of the video, but you somewhat described how SSDs and RAM work. Good stuff

[JohnDoe-kvkd]

I love when Professor James makes a visit. Funny as much as I feel like I’ve learned I still don’t really understand how it all works. But I get the basic concept SSDs die whereas RAM doesn’t for obvious reasons.

[Heavyharper]

I would find it quite interesting to see a simple CPU scaled up so you can show how i processes data, if that's something you could do a video on.

[gblargg]

Also, the capacitors in RAM lose data within milliseconds and have to be refreshed many times a second. They are like flash memory that's extremely degraded. Reading from (dynamic) RAM clears the data as well, so it has to rewrite it each time the CPU reads it. It's just that this is expected behavior, and the constant refreshing doesn't degrade it appreciably.

[jasper]

The very short version: storing data without power is a lot harder than with power.

[smo]

Hello guys, I love Techquickie and it was refreshing and fun to see James all studious for once.
Suggestion for a future TQ : could you go over positive/negative air pressure in a case? I believe you covered this topic several times in the past but I still get confused regularly. I just wish wasn't so puzzled when trying to figure out how many fans I need or how I should place them.
Great work and thanks!

[Aelfraed]

Weird that the video never uses the terms "NAND", "DRAM" and "endurance".

As far as I know, DRAM endurance is orders of magnitude higher than that of NAND, but then again, DRAM is constantly being written to when it's being used.

[rallisf]

You also neglected to mention that most SSD failures come from its controller (most times due to overheating) and not the flash chips themselves. That's got nothing to do with degrading but bad manufacturing.

[RaydeusMX]

Overclocking ram has to be the most pointless thing I've experienced, at the end of the day tightening the timings worked far better and didn't require me to increase the voltage at all.

[phiwolgast]

I had my boot SSD die on me about a year ago, and as I was recovering the data, I noticed it was read-only, but I had no idea why... until now! Thanks, James!

[chuckbotz]

James is getting ever more professional in these videos. It's a joy to see him progress. The glasses are a nice touch too!