web analytics

Home of my tech rants, free programs, and a story or two…

Samsung 840 EVO and the NAND flash 20nm wall

A new firmware has been released for the Samsung 840 EVO and it appears to have fixed the issue with speed degradation of older data but… and you knew that was coming, it has some side effects.

Samsung has stated that there will be periodic refreshes of the data in the background either by the firmware of the SSD itself or by a tool in the Samsung SSD Magician. To quote Anandtech…

Ultimately Samsung’s second fix is a bit of a brute force solution to the problem, but at this point there doesn’t seem to be anything Samsung can do about 19nm TLC cell charge decay other than to refresh the data, as the problem is intrinsic to the NAND itself. Pre-release versions of the firmware show that this fix works, and conceptually this is much more likely to work over the long run than Samsung’s initial fix. The tradeoff is that it does consume P/E cycles to refresh the data, but by our own calculations even 5 years of refreshes at 1/week would only be 26% of the drive’s rated 1000 cycle lifetime.

And it seems that TLC NAND isn’t the only type of NAND that’s having a problem at process nodes of 19nm or less. The Crucial MX200 SSD which uses 16nm MLC 128Gb NAND flash chips appears to be having the same very issue that the Samsung 840 EVO had, namely NAND flash voltage drifting which causes read speed performance issues on older data.

What does this mean? Well, it means that pretty much NAND flash memory becomes unreliable as the process node shrinks past 20nm. NAND flash manufacturers have hit the wall when it comes to planar NAND. From this point on the only way to make NAND flash reliable is to reverse the process node shrinkage and go to 3D-NAND where NAND cells are stacked on one another instead of laid out flat.

Samsung started this trend in which they released their 3D-VNAND and reversed the process shrinkage to 40nm with their 850 line of SSDs. Other manufacturers are also doing the same with their own approach to 3D-NAND. Ultimately it comes down to the fact that NAND flash memory manufacturers have hit the wall in terms of shrinking NAND and that as you shrink the process node the NAND has far more of a chance of voltage drifting that can effect read speed performance as the data ages.

Avoid Samsung SSDs at all costs! (Part 2)

Alright, some more information to go on.

TLC+NAND
As you can see in the above image, you can see how TLC NAND Flash works. TLC which stands for Triple Level Cell and is named that because it stores three bits or three binary states per cell. 000, 001, 101, 011, 100, 101, 110, and 111

Each binary state requires a different voltage state to represent the binary state in the NAND flash cell, specifically eight different states. This is much different from traditional MLC NAND in which it only has to hold four different binary states.

MLC+NAND
What is being theorized is that there is voltage drift inside the NAND flash cell. To show what’s going on, we have another image.

TLC+NAND+with+voltage+drift
Notice how you have two voltage states, the 010 and 011 states. They are closer together because the voltage state drifted. When we read the cell and we get a cell voltage back we have an issue. Does that cell contain a 010 or 011?

So now, we have to read the flash cell multiple times to come to a general consensus on what that cell is storing. If a majority of reads come back with a voltage state that represents 011 then it must be 011.

Therein lies the problem. Because of voltage drift in the NAND Flash Cell a Flash Cell must be read multiple times because we need to figure out what that cell is storing because of voltage drift.

Because both the 840 and 840 EVO use much smaller NAND Flash Cells, 21nm and 19nm respectively. Because the flash cell is so small there’s not a lot of room to store a lot of electrons thus any change in voltage state caused by an electron or two escaping the flash cell can cause the voltage drift to be very detrimental to reading the binary state back from the NAND flash cell.

Samsung has gone back to 40nm with their new 3D NAND technology which should in theory solve the issues with voltage drift but like the 840 EVO, the 850 EVO is also TLC or Triple Level Cell and thus open once again to voltage drift inside the cell effecting read speeds because of the many various forms of voltage states needed to represent eight voltage states for eight binary states.

With that being said, Samsung should be forced to do a massive world wide recall of the 840 and 840 EVO because the voltage drifts can become a bigger issue with reading data back as the NAND flash cells become older and more writes have been done to the flash cell. More voltage drifts means more time needed to read data and quite possibly lead to mass data corruption due to the inability to read data back from older flash cells.

But, Samsung won’t do this. This is why I call for a boycott on Samsung SSDs until they do right for the owners of the 840 and 840 EVO. They are playing with people’s data integrity here.

Avoid Samsung SSDs at all costs!

This includes the new 850 series!

To those that may not know about the issue that has been plaguing many users who have both the 840 and 840 EVO SSD products, the 840 series (non-Pro) has a serious hardware design flaw in which data that is older than 60 to 90 days reads far, far slower than data that is newer. Some people have seen read speeds as low as 50 MB/s, yes you read right… 50 Megabytes per Second! That’s slower than slow notebook hard drives!

The reason is that the Samsung 840 and 840 EVO uses TLC NAND Flash Memory, specifically 19nm NAND Flash. There is much speculation but the most logical reason why the data read speeds become lower as the data ages is because of severe electron leakage between the NAND flash cells which results in a much larger voltage drift inside the NAND flash cell.

Lets rewinds back a couple of months ago to when Samsung acknowledged the issue in the first place. They promised a firmware fix for the SSD along with a performance restoration tool. All the performance restoration tool did was rewrite the data to get the data’s read speed performance back. The firmware fix promised that this issue would never come back because they supposedly tweaked the voltage calculation algorithms and that data read speeds would be consistent from there on out. Fast forward to today and we know that that promise was something not to be believed.

So here we are, owners of the Samsung 840 and 840 EVO drives sit with SSDs that will forever experience these data read speed slow downs along with a very good possibility that data may become corrupted in the SSDs due to voltage drift in the NAND cell. Samsung said that something would be coming out in March and again, here we are in April and nothing but silence from Samsung.

Samsung lied to us when they told users of the 840 EVO that the firmware fix will solve the issue. And then when the problem came back, they lied to us again when they said that they would deliver another supposed fix. Samsung has no intention of fixing this issue. They can’t fix this issue. This is a hardware design flaw and one that would need a worldwide recall of the 840 EVO and by extension the 840 but Samsung won’t do that.

So at this point I call upon all users to avoid Samsung SSDs at all costs! Hell, I’d go so far as to avoid all Samsung products. If it has Samsung’s name on it, avoid it like the plague! This includes SSDs, hard drives, computers, televisions, tablets, smartphones, etc. Anything made by them, do not buy! Boycott this lying and cheating company!

The Samsung 840 EVO… An SSD with an obvious hardware design flaw.

The Samsung 840 EVO SSD has a very well known issue that results in reduced performance while reading older stored data. The issue manifests itself as reduced read speeds when you start to read data that is older than three months (90 days). This has been talked about in length on Overclock.net’s SSD forum as well as several reputable technology enthusiast news sites. Normally the SSD can read data from the drive at speeds around 500 MB/s but when reading older data read speeds can be as low as 50 to 100 MB/s which is drastically lower than that of the advertised speed of the SSD. This can effect the speed of loading programs as well as Windows itself.

Samsung admitted that there is indeed a problem and released a firmware patch back in October of 2014 that promised to fix the issue with the SSD. Now, several months later and we are back to where we started; reduced read performance. Samsung stated that they have plans to release some kind of software solution to improve performance of the SSD but that it would need to be done every couple of months. This is nothing but a band-aid, it is not a solution. It’s only purpose is to mask the issue and make it less noticeable to the users of the device.

The general consensus of users of the Samsung 840 EVO the world over is that this issue is a hardware design flaw and one that Samsung is simply trying to cover up.

I call on Samsung to refund or recall every single 840 EVO that has been sold.

The Samsung 840 EVO performance bug is back…

It seems that the firmware “fix” that Samsung put out nearly three months ago to correct the read speeds of old data on the 840 EVO did nothing.

A number of users over at the Overclock.net Forums, including myself, are experiencing the same very issues with the SSD that Samsung claimed the firmware “fix” would correct. You can see (at the link below) that a number of users have posted screen shots of HD Tune clearly showing that the issue still exists.

http://www.overclock.net/t/1507897/samsung-840-evo-read-speed-drops-on-old-written-data-in-the-drive/1360

I was one of many users who were seriously hoping that this issue would be laid to rest with the firmware “fix” but based upon the benchmark data that’s been posted, that clearly isn’t the case. At this point, we need to make Samsung know that users of the 840 EVO want compensation for the flawed hardware and that we expect nothing less than a recall of the flawed hardware and replacement with new SSDs.

Some people are experiencing average read speeds as low as 280 MB/s which doesn’t even come close to saturating a SATA 6 Gbps data channel. This is exactly what a lot of us started seeing when the data on the SSD became about 2 month old, from that point on it would drop lower as the age of the data increased.

I wrote an email to Tom’s Hardware and TweakTown.com and so far I’ve received an email from an editor at TweakTown.com stating that an article is going to be hitting the front page within the next five hours. I’ll be sending an email to someone at AnandTech as soon as I find some contact info.