Is Evo Ssd fixed now?

[Coup27]

Because it takes about six months for the data to become stale and slower

Where have you gotten 6 months from? Many sources claim the data only has to be a month old. One source:

http://www.anandtech.com/show/8550/...40-evo-read-performance-bug-fix-is-on-the-way

 

[Wall Street]

Where have you gotten 6 months from? Many sources claim the data only has to be a month old. One source:

http://www.anandtech.com/show/8550/...40-evo-read-performance-bug-fix-is-on-the-way

The amount of time between when the drives first debuted and when threads about really slow performance started popping up was more than one month.

 

[shredz]

I've decided to put the new Evo in my son's computer and keep the seagate Pro in my computer. The little bit of speed gained (if any) doesn't seem worth the hassle and risk of any future issues.

 

[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. And @berryracer could not prove his statement.

And RAPID? Nope: http://techreport.com/review/25282/a-closer-look-at-rapid-dram-caching-on-the-samsung-840-evo-ssd/8

 

[AntiV6]

RAPID has been a nice boost ever since Magician 4.4 came out. If you have 8+ gigs of RAM imo.

And RAPID? Nope: http://techreport.com/review/25282/a-closer-look-at-rapid-dram-caching-on-the-samsung-840-evo-ssd/8

You do realize a lot can change in 15 months, right? The system tested only had 4 gigs of RAM, which severely limited RAPID back in 2013 with RAPID 1.0.

Nowadays (Magician 4.5 has RAPID 2.0) Samsung allows systems to use up to 4 gigs for the caching and is way more efficient

 

[corkyg]

Is there a date after which all EVO SSDs will have the fix factory installed? (So the user does not have to be concerned.)

 

[Wall Street]

I would like to post a big "I told you so" for predicting that we wouldn't know if the evo is fixed until many months after the new firmware.

Some 840 EVOs still vulnerable to read speed slowdowns - The Tech Report

 

[VirtualLarry]

I would like to post a big "I told you so" for predicting that we wouldn't know if the evo is fixed until many months after the new firmware.

Some 840 EVOs still vulnerable to read speed slowdowns - The Tech Report

I think that... this is just the "nature" of using TLC for SSDs. More ecc recovery is required over time to retrieve the original host data.

 

[SSBrain]

Why do you assume that? At this stage, given that the bit error rate / retention quality of fresh (with no few P/E cycles spent) TLC NAND memory isn't remotely supposed to be an issue after the short amount of time it takes to make problem arise, it's still more likely it's due to insurmountable firmware or architectural problems of the NAND controller they designed rather than inherently due to the memory.

If it was the "nature" of TLC NAND, it wouldn't be possible to use these SSDs at the rated P/E limit or above that, because the Bit Error Rate in all NAND memory increases exponentially with wear. At the very least, it would mean that at 1000 P/E cycles performance would start dropping badly almost as soon as data are written.

A test with a "well used" Samsung 840 / 840 EVO would help clarifying this.

 

[Quad5Ny]

Cell voltage. The voltage level between cell states is just too close.

Don't rewrite a cell for a few weeks and the known state (what the controller thinks a 0010, 1011, etc. should be) is now way off and it has to go through multiple reads and error correction to give you back your saved state. That extra time results in severely decreased transfer rates and eventually loss of data.

Source: http://www.anandtech.com/show/6337/samsung-ssd-840-250gb-review/3

"The difference between SLC, MLC, and TLC is pretty simple: SLC has the fewest voltage states and hence it can tolerate bigger changes in voltages. With TLC, there are eight different states and hence a lot less voltage room to play with. While the exact voltages used are unknown, you basically have to divide the same voltage into eight sections instead of four or two like the graphs above show, which means the voltages don't have room to change as much."

SLC

MLC

TLC

 

[Quad5Ny]

Also, users have been reporting this for weeks. Why does it take a Tech/News site to actually get anything done?

Heck even I made a post about it: http://forums.anandtech.com/showthread.php?t=2415738

I wonder if Samsung will swap 840 EVO's for 840 Pro's... yeah like that's going to happen.

 

[SSBrain]

I'll stress again that with any kind of NAND memory (SLC, MLC or TLC) the bit error rate increases (together with load on the controller's ECC algorithms) exponentially with P/E cycle depletion, and if the SSD is having issues when the NAND is new or almost new specifically because of this, then it should become very obvious when it's approached its rated P/E limit.

Have there been reports or tests showing that the performance loss problem on stale data becomes more severe (ie occurring earlier or leading to a larger performance loss) with NAND wear on Samsung 840/840 EVO SSDs? If there doesn't seem to be a clear correlation with P/E cycles, then it's most likely not inherently/directly due to the TLC NAND, but to different problems instead.

 

[WhoBeDaPlaya]

Looks like I made the right decision to dump my vanilla 500GB 840s (which Samsung didn't even bother to "fix").

 

[SSBrain]

Looks like I made the right decision to dump my vanilla 500GB 840s (which Samsung didn't even bother to "fix").

I've got two on different machines and I'm waiting for a fix too. Hopefully it will come at some point since Samsung hasn't hinted (yet) that no more firmware updates will be released for them, unlike with earlier 830 and 470 models (see here).

Anyway, if "all TLC NAND is bad" (although to me it looks more like it's actually "Samsung's current gen TLC NAND SSD controller is less sophisticated than we thought"), then the SanDisk Ultra II and the upcoming SSDs with Toshiba A19 TLC NAND should show the same issue as well. Has anybody checked out yet on the Ultra II?

 

[MustISO]

Saw this report this morning:
http://techreport.com/review/27727/some-840-evos-still-vulnerable-to-read-speed-slowdowns

Interested in seeing if Samsung responds

 

[redzo]

Saw this report this morning:
http://techreport.com/review/27727/some-840-evos-still-vulnerable-to-read-speed-slowdowns

Interested in seeing if Samsung responds

That's bad news. It could mean that speculation on how the fix works might become true: by rewriting old data while the drive is powered on. This should directly affect endurance big time. Bad news for people with endurance phobia.

http://www.anandtech.com/show/8617/...e-to-fix-the-ssd-840-evo-read-performance-bug

Kristian's speculation on how the fix works is more optimistic:

I suspect that the algorithm didn't take the change in cell voltage properly into account, which translated into corrupted read points and thus the read process had to be repeated multiple times before the cell would return the correct value. Obviously it takes more time if the read process has to be performed multiple times, so the user performance suffered as a result.

: no cell writes necessary, just buggy nand cell read algorithm.

TR's new findings could fit well with the theory that the drive may be doing additional nand writes of old data while being powered on and screwing up endurance during this process. Since their sample was powered off for a long time, there was no time for the so called "new C firmware fix" to do its job. This seems like a plausible explanation regarding TR's new findings. Still, we need more tests and more reports.

 

[SSBrain]

That's bad news. It could mean that speculation on how the fix works might become true: by rewriting old data while the drive is powered on. This should directly affect endurance big time. Bad news for people with endurance phobia.

If keeping performance on the bleeding edge on these drives meant rewriting the entire SSD content on itself monthly, using 1 P/E cycle every time, it wouldn't really be a big deal. After 3 years (warranty period), only 36 P/E cycles out of 1000 would have been spent just because of this, and it's assuming that the entire SSD contains static data.

But since people who have been using these drives regularly are having this issue as well, then perhaps the fix wasn't really intended to do this.

 

[SSBrain]

Apparently some users are reporting read performance improvements after executing HDTune's error scan a few times. Can anybody try?

 

[Hellhammer]

Also, users have been reporting this for weeks. Why does it take a Tech/News site to actually get anything done?

It takes more than a few user reports before it's worthwhile for us to contact the manufacturer and possibly post something. That said, I'm working with Samsung to get some answers.

 

[redzo]

OK. I did a few simple read tests on my home 250GB 840 evo drive sample.
First only format was on q4/2013, day of purchase.
At the time of samsung's fix launch the ssd was in the very low reads department. Fix was applied a few days after launch and did his job well.
Now the drive is doing very good in the read department.

 

[SSBrain]

Apparently some users are reporting read performance improvements after executing HDTune's error scan a few times. Can anybody try?

I have a Samsung 840 (non-EVO) on a different PC and it appears that performance recovered visibly after doing that.

Before (yesterday):

After doing several passes of the HDTune error check (which basically loads the SSD with sequential reads for a prolonged period of time):

Before: 130015 MB read in 1080.44 seconds, average data rate: 166MB/s
After: 128280 MB read in 387.92 seconds, average data rate: 337 MB/s

 

[GlacierFreeze]

Dang nice find. Weird that lots of Reads seem to alleviate this...

 

[SSBrain]

I would be more confident in those results if other people tested that as well, though.

Also, it would be useful knowing if HDTune uses a certain read pattern which could be helping the drive recalibrating (?) itself. Just performing simple large block sequential reads along the entire drive's LBA range (which I'm testing with a program called HDDScan) doesn't seem to be of much help. It could also be that there's a limit to the amount of performance that can be restored this way, though, especially on the older unfixed Samsung 840 SSDs.

 

[GlacierFreeze]

I would be more confident in those results if other people tested that as well, though.

If I remember, I was planning on testing it myself.

 

[bluwing]

Hi,

I really think that a FEW might have problems as with all electrical and mechanical Equipment..

As for me I have bought and using Eight Crucial drives (from 120 to 512) and one Mushkin 480gb SSD.. I have not got a bad one... When you read forums you can't take their answers on face value. A lot of what you posted did not say which model they had or if they had the latest firmware or controller..

My Daughter runs two 512gb 840 Samsung SSDs in her I buy Power Laptop... It is over six montes old with NO problems...

Nobody is forcing you to buy crucial or any other brand! So do MORE RESEARCH!!! One more thing that people do after they get a SSD is after a month or two Decide to Defrag the drive... DON'T DEFRAG A SSD!!!

bluwing