OS SSD died. What to get now.

[OilSheikh]

+1 for Samsung Evo.

[Kovoet]

+2 for the Samsung evo

[CAT-THE-FIFTH]

-1 for the Samsung EVO.

[CAT-THE-FIFTH]

The durability of the MX100 isn't necessarily any better than TLC based drives, as the former uses Micron's newest 16nm NAND. As NAND cells get smaller, the number of program / erase cycles they support decreases. Of course all SSDs use wear-levelling & over-provisioning to mitigate cell failures, and the MX100 does still carry a 3 year warranty.

Actually from what I gather the 19NM TLC is still worse than 16NM MLC. The data redundancy mechanisms appear to be better on the MX100(RAIN),its cheaper than the 840 EVO and has higher capacity too.The bug in the drive which massively drops speed for older files is also not helping the 840 EVO.

This is the problem though - the 830 was a great drive especially at the low pricing it dropped to eventually. However,Samsung did a Corsair and birfurcated lines. Many people bought the 840 EVO thinking it was the successor to the 830 when it was not - that was the 840 PRO and that interestingly uses MLC NAND too. However,like with Corsair people thought they were the dogs bollocks even when they started cost cutting and it took years for it to sink in that there were better alternatives.

However,if Crucial can charge TLC prices for MLC NAND and so can virtually ALL the competing drives at a similar price,then TBH it just shows Samsung is just trying to maximise profits - just like with what Corsair did by cheapening out on more budget lines when companies like XFX and so on did not and just riding on the brand name. If they thought TLC was fine they would have used it in their 840 PRO and 850 PRO ranges and they have not. It shows they actually have little or no faith in using it for their "better" drives.

Ultimately the lower capacity and cheaper to make 840 EVO should be the cheapest of the 240GB/256GB SSDs out there - it isn't though.

[Agent]

I'm not a fan of the driver / RAM cache that's used to boost to scores on the EVO. One powercut without the data being flushed to disk and it's gone.

The Crucial drives (M500 up) have protection capacitors that'll flush the data from its internal cache to the NAND.

Add to that the real world use being basically no different and I tend to stick with Crucial. They also have a UK RMA based that is very fast.

[watercooled]

Wow that EVO issue is shockingly bad; I'm just scanning through the thread and it seems everyone who's attempted it has been able to reproduce the problem! Maybe it's worth its own Hexus thread to collate information?

I've just ran HD Tach on a Crucial M4 which has been in service for something like 18 months since the last erase, and the trace is pretty much flat as you'd expect.

Edit: Here's another thread discussing the issue: http://forums.overclockers.co.uk/sho...php?t=18615995

[Bonebreaker777]

Crucial MX100 or SanDisk Ultra Plus. Regularly purchasing both for "rejuvenation projects". At 120/128GB the SanDisk is sometimes even £5 cheaper which makes it a superb value.

[deejayburnout]

Some good news

Got an email today and I am getting a new SSD. Not only is it a newer model but also double the size.

http://www.newegg.com/Product/Produc...82E16820228105.

Worryingly it's the same 19NM TLC as detailed in the previous post.

[watercooled]

It's Toshiba MLC, the issue is with the Samsung drives using Samsung TLC however it's not clear if it's down to the NAND itself or a firmware bug.

[deejayburnout]

That's good to hear.

[watercooled]

I haven't seen a thread dedicated to the issue, but it's been mentioned on this one so possibly worth posting:

It seems Samsung have admitted there's a problem with the TLC drives and are supposedly working on a fix: http://www.anandtech.com/show/8550/s...-is-on-the-way

TBH though it still seems quite poor that it was left for end users to discover rather than Samsung pro-actively searching for problems like this. I mean it's not even like it's hard to recreate and doesn't seem uncommon at all.

[Agent]

It's taken over a year for any issues to really surface - it's always annoying when there is an issue, but this isn't as pronounced as some people are making out. A lot users will never notice it and due to the nature of the bug, I can see why this could have easily have been missed in testing.

[watercooled]

The sequential speed has dropped below that of an old hard drive in some cases, to the point it has prompted people to investigate.

I could understand it being missed in initial testing, but it should have been an easy thing to spot if they even have some drives in service in an office, and I thought they would keep testing drives over time to check for time-related bugs which aren't all that uncommon on SSDs.

[CAT-THE-FIFTH]

Looks like Samsung acknowledged the problem:

http://www.anandtech.com/show/8550/s...-is-on-the-way

[watercooled]

I guess the likelihood of a firmware patch lessens my concern that it could have been a more serious issue related to DSP/ECC struggling with the charge margins on TLC. It's just, we hadn't really seen any consumer TLC drives until the 840(evo) so we didn't have anything to compare against.

[Agent]

The sequential speed has dropped below that of an old hard drive in some cases, to the point it has prompted people to investigate.

I could understand it being missed in initial testing, but it should have been an easy thing to spot if they even have some drives in service in an office, and I thought they would keep testing drives over time to check for time-related bugs which aren't all that uncommon on SSDs.

Yeah, but sequential isn't often that noticeable compared to how fast accessing the data is, unless you're doing large transfers. I'm not trying to underplay the issue (I actually prefer Crucial), but I think given how long it took to surface and that a *lot* of people needed to actively test for it, as opposed to feeling the difference is quite telling.

Hopefully the firmware is a real fix and not some work around. Samsung haven't actually said what the issue is.