Samsung 850 EVO? (TLC V-NAND) Almost here!

[SSBrain]

Have a look at this brochure and save it to your drives before it disappears:

http://www.netonnet.se/FileRoot/Ite...SE/Samsung SSD 850 EVO Data Sheet Rev 1.0.pdf

I guess it is very close to get released.

I believe its NAND memory is likely to be rated for 2500 P/E cycles.

 

[KryssNova]

Yep, it's going to be released in December 2014 for sure (shops getting their supplies by now).
Some Norwegian shops already have confirmed available supply date, for people to buy at 15-21 December.
They also have already prepared online adverts etc....
Though I'm just saying this unofficially, and everything here is to be taken as just speculation.

To be honest I already had one in my hand :whiste:, planning to get one for myself.
They are just starting to stack up the supplies now.

Well if I'm wrong, which I strongly doubt that I am, about them being released in December 2014.
They surely going to be released in January 2015 and no later, if there is aren't any "problems".

Though I'm just saying this unofficially, and everything here is to be taken as just speculation.
(I love to add this sentence, hahaha, sorry for being a tease).

 

[SSBrain]

Well if I'm wrong, which I strongly doubt that I am, about them being released in December 2014.
They surely going to be released in January 2015 and no later, if there is aren't any "problems".

I think that too. It's likely not going to be released after January 2015.


By the way, here are its technical specifications:

Endurance values seem weird to me and a tad too much underestimated for the 250GB and 1TB drives.
75 TBW (under a standard JEDEC random workload) seem about right for the 120GB model, assuming 2000-2500 P/E Cycles.

For some reason the 1TB version used the 840 EVO controller (MEX), not the updated version (MGX). I wonder if perhaps this one was is actually not intended for big drives and if there will be more large-capacity SSDs from Samsung in the near future.

 

[Berryracer]

I'd never buy a TLC SSD again after having so many slow down issues with my previous 2x1TB 840 EVO SSDs. Even though there is a fix, it's just temporary and after a while the performance drops again. MLC SSDs FTW

 

[Phynaz]

** Sequential Write performance measurements based on TurboWrite technology, The sequential write performances after TurboWrite
region are 150MB/s(120GB), 300MB/s(250GB), 500MB/s(500GB) and 520MB/s(1TB).

Rather slow.

 

[SSBrain]

Rather slow.

The 840 EVO has lower sequential write speeds in case you end up filling the Turbowrite emulated-SLC buffer.

 

[sancco]

I'd never buy a TLC SSD again after having so many slow down issues with my previous 2x1TB 840 EVO SSDs. Even though there is a fix, it's just temporary and after a while the performance drops again. MLC SSDs FTW

Hi, registered just to ask - I thought the firmware fix they put out was a permanent solution? Did you update the firmware, or just run the temporary optimization tool?

I've been hanging out for the 850 EVO but if TLC really isn't refined enough yet, I might change my mind.

Also, OP, nice work on finding that info!

 

[hojnikb]

Anyone knows if this 3D TLC is actually the same as in 850PRO ?
i've read lots of speculation on the internets, that samsung is actually "simulating" MLC on the PRO, because NAND itself is actually made for TLC in mind.

Any truth in that ? It would certanly explain the odd die size.


Also, EVO does have poor sustained writes fora 3d nand drive. But 5 year warranty is insane tho.

 

[StinkyPinky]

I'd never buy a TLC SSD again after having so many slow down issues with my previous 2x1TB 840 EVO SSDs. Even though there is a fix, it's just temporary and after a while the performance drops again. MLC SSDs FTW

THere's a firmware update that made it permanent.

 

[SSBrain]

It looks like for some reason there's the Samsung 850 EVO on the dutch version of the Samsung website:

http://www.samsung.com/nl/consumer/memory/ssd-solid-state-drive/840-evo/MZ-75E500B/EU
http://www.samsung.com/nl/consumer/memory/ssd-solid-state-drive

The URL shows it as 840 EVO though.

 

[Berryracer]

Hi, registered just to ask - I thought the firmware fix they put out was a permanent solution? Did you update the firmware, or just run the temporary optimization tool?

I've been hanging out for the 850 EVO but if TLC really isn't refined enough yet, I might change my mind.

Also, OP, nice work on finding that info!

performing the fix involves 1 things:

1) firmware update then reboot

2) long process of re-writing the data so it would be new again and not suffer from the performance loss

so that means:

a) uneccessary writes to the SSD reducing its write limits

b) just temporary until those files that were re-written are old again

c) performance drop happens in about 1 to 2 months again

d) from my experience, TLC is great in benchmarks, but in real world usage, I feel MLC offers higher performance as the system overall is snappier and gives more real world good performance not just in benchmarks

 

[SSBrain]

Do you have actual testing data to corroborate what you're writing? It seems weird to me that Samsung would put out a tool which only delays the issue by an additional month or two, as if people aren't going to check if it the fix works afterwards.

 

[Topweasel]

Do you have actual testing data to corroborate what you're writing? It seems weird to me that Samsung would put out a tool which only delays the issue by an additional month or two, as if people aren't going to check if it the fix works afterwards.

It'll probably eventually become a maintenance tool. Moving old files around to refresh data integrity in the cells. It's sad if Samsung is continuing to use TLC as part of its new 3d cell structure. It's a tech that should make the data density worries disappear and make TLC redundant. Seems silly and cheap to be using TLC with 3d nand.

 

[SSBrain]

People used to say the same with MLC SSDs when they started becoming more popular compared to more expensive SLC ones.

 

[Hellhammer]

It'll probably eventually become a maintenance tool. Moving old files around to refresh data integrity in the cells.

The restoration process can only be run once, so it's definitely not a maintenance tool. There is absolutely zero evidence that suggests that the performance degradation issue exists after the firmware update. berryracer's hasn't provided any actual proof other than his subjective analysis, which to be honest has no value.

It's sad if Samsung is continuing to use TLC as part of its new 3d cell structure. It's a tech that should make the data density worries disappear and make TLC redundant. Seems silly and cheap to be using TLC with 3d nand.

3D actually makes TLC more viable than ever before thanks to the larger lithography. Most of the client market will move to TLC with 3D.

 

[Topweasel]

The restoration process can only be run once, so it's definitely not a maintenance tool. There is absolutely zero evidence that suggests that the performance degradation issue exists after the firmware update. berryracer's hasn't provided any actual proof other than his subjective analysis, which to be honest has no value.



3D actually makes TLC more viable than ever before thanks to the larger lithography. Most of the client market will move to TLC with 3D.

Lets hold out and see. TLC has two huge side affects. 1. Less write endurance. 2. Data integraty issues. MLC and SLC still have them but its exponetial from SLC to MLC to TLC. The fact that people a year and a half later are starting to see the issues with stale files doesn't surprise me. This one run once tool recondition the drive may give people another year to think before they have to worry it again. Maybe they rewrote the firmware to do random data moving for stale files along with normal trim and garbage collection work. So maybe it will only be a blip for most users.

That said is the 3d nand already gives them a seamless end of data growth per process and they rewound the process enough to cover nearly two and a half decades of data growth requirements. So why play fast and loose peoples data using TLC with its shoddy performance, poor write endurance (not much of an issue), and its ability to hold its data structure. In theory yes going back to 40nm should make the last two a smaller issue. But by marrying themselves to TLC 3nand now, it just means that problems creep up as they hit the limits on cell height and have to look at using a smaller process again. Then all of sudden we are back here and they won't want to drop TLC because it would defeat the purpose of using the smaller process.

 

[Topweasel]

People used to say the same with MLC SSDs when they started becoming more popular compared to more expensive SLC ones.

And the statements hold true for the most part. It just wasn't as large an affect. But the cell cost went from extra ordinary to consumer ability really quick, it was easier to maintain performance, and MLC can go a reasonable period of time unplugged without data loss.

Think of it this way. SLC is an enterprise Nand, performance and endurance to run circles around anything else in big business. MLC is the General consumer Nand. Not nearly as reliable, tricky to keep performance up, but is much more cost effective. So TLC should be the economy Nand right? Needing a billion and one tricks to maintain performance while being right on the edge of data integrity, should be at highest well below the cost of a comparable MLC drive right? No as it is the sole TLC drive options actually cost more to comparative MLC drives with similar capacities even though it requires less chips.

 

[meloz]

The baseline performance numbers are so good, it makes you wonder what will happen to 850 PRO?

Since Samsung are effectively bottlenecked by the SATA interface (until Skylake makes SATA Express mainstream), they cannot truly position the 850 PRO as a high performance alternative to 850 EVO. When the performance of 850 EVO is so similar to the 850 PRO, why would anyone buy the PRO for more money?

Sure sure, there is a small segment of users that do a lot of write intensive work (HD/4K video editing/grading) and they can benefit from the higher endurance of the 850 PRO, but for most users the 850 EVO should be good enough.

 

[sancco]

The baseline performance numbers are so good, it makes you wonder what will happen to 850 PRO?

Since Samsung are effectively bottlenecked by the SATA interface (until Skylake makes SATA Express mainstream), they cannot truly position the 850 PRO as a high performance alternative to 850 EVO. When the performance of 850 EVO is so similar to the 850 PRO, why would anyone buy the PRO for more money?

Sure sure, there is a small segment of users that do a lot of write intensive work (HD/4K video editing/grading) and they can benefit from the higher endurance of the 850 PRO, but for most users the 850 EVO should be good enough.

my thoughts too. at the very least, there's little reason they release an 850 Evo now as there is no competition pressure. but even when more TLC drives hit the market, why would they want to close the gap between their two SSD lines?

 

[BFG10K]

Even though there is a fix, it's just temporary and after a while the performance drops again.

Huh? If it's only temporary because "only the files get moved around", what is the purpose of the firmware update?

 

[aviator79]

The fix from Samsung is NOT temporary, it's permanent. The issue was located in the Firmware, not the NAND. And along with the Performance Restoration tool comes the new firmare EXT0B"C"6Q. (And yes, the new FW does NOT come with magician yet).
I checked this double and triple. No degradation on the "old" data since the release of the fix. And they are not just randomly moving data. This would be too obvious. Just check SMART an TBW. Nothing suspicious so far.

 

[rockfella79]

The 850 comes with 10 years warrantee. I was sold on that. Using the 128GB:

albeit on SATA II.

SATA Express doesnot look promising.

http://www.anandtech.com/show/7843/testing-sata-express-with-asus/5

The baseline performance numbers are so good, it makes you wonder what will happen to 850 PRO?

Since Samsung are effectively bottlenecked by the SATA interface (until Skylake makes SATA Express mainstream), they cannot truly position the 850 PRO as a high performance alternative to 850 EVO. When the performance of 850 EVO is so similar to the 850 PRO, why would anyone buy the PRO for more money?

Sure sure, there is a small segment of users that do a lot of write intensive work (HD/4K video editing/grading) and they can benefit from the higher endurance of the 850 PRO, but for most users the 850 EVO should be good enough.

 

[Cerb]

If upcoming SSDs support SATA Express, it will be for legacy compatibility only, assuming motherboard firmwares aren't capable of handling that.

Whether SATA Express holds promise or not, it has little bearing on anything flash-based.

 

[SSBrain]

The 850 EVO was tested here:
http://tweakers.net/reviews/3825/de-nieuwste-ssds-samsungs-850-evo-en-de-concurrentie.html

I wonder if the embargo ended today. If so, we should se plenty of reviews later.

 

[meloz]

The 850 EVO was tested here:
http://tweakers.net/reviews/3825/de-nieuwste-ssds-samsungs-850-evo-en-de-concurrentie.html

Thanks SSBrain!

Performance looks rather decent.

In most cases the 850 EVO keeps up with and even outperforms the 850 PRO at similar capacities.

The biggest area of difference between the EVO and PRO is in the 4K random write. The stark difference between 120 GB, 250 GB and 500 GB 850 EVOs makes you wonder if this is an attempt at artificial segmentation.

To be fair the 840 EVO also had this 'problem', it is strange Samsung should (or could?) not fix it with their new generation MGX controller. But 4K random read performance has improved tremendously over 840 EVO -especially in lower capacities- and for most desktop users that is far more relevant anyway.

Let's hope the high price is a temporary thing (to help clear the 840 EVO stock) and soon 850 EVO will be available at same or lower prices than the generation it is going to replace.