samsung 840 EVO new FW fix slow down

F

Flopper

[H]ard|Gawd
Joined
Nov 15, 2007
Messages
1,642
Last edited:
SRTie4k said:
An hour?? Damn, I feel uncomfortable flashing firmware that takes 1 minute.
Click to expand...

Yea bios flash is always for me like, ok cross my heart..
I dont have a 840 evo drive myself but those that updated have reported it works fine.
the time it takes are due to the software update the drive files on the disc.
 
I suspect the BIOS flash takes the same time and the rest of the time is spent reorganizing all files on the SSD doing something similar to a one time defrag.
 
Hmm, have one EVO SSD in a laptop. Bit apprehensive, but at least it has a battery in case of bad luck with power.

That's a loooong update...
 
SRTie4k said:
An hour?? Damn, I feel uncomfortable flashing firmware that takes 1 minute.
Click to expand...

Try reading the PDF before posting. The firmware update is right away, the rest of the time looks like it's just refreshing the contents.
 
According to my samsung magician I have the latest firmware. EXT0BB6Q is this correct? Purchased new about 5 months ago.
 
Flopper said:
scroll down on the link you have the restoration fw there.
Click to expand...

I am still drunk this morning, and don't see anything other than the EXT0BB6Q that was released on Dec '13. Am I suppose to install the Performance Restoration software?
 
ClutchAutomatic said:
I am still drunk this morning, and don't see anything other than the EXT0BB6Q that was released on Dec '13. Am I suppose to install the Performance Restoration software?
Click to expand...

if you have a samsung 840 EVO drive yes.
The software fixes the drive from the slow down happening to old files.
the FW update only does not as far I can tell.

Read the guide, back up data and dont use the computer while the software does its job.
 
Has anyone seen any performance boosts with it? My 840 EVO is quite new, so I'm not so sure what I'd get out of it.
 
Last edited:
Here are my benchmark results from Samsung Magician. Before my Sequential read was 518 and after was 552, sequential write went from 388 to 528, random read from 96768 to 96556, and random write from 69043 to 69458. Biggest gains was the sequential write speed.
 
I'd love to see a statement from Samsung on why this fix requires additional software beyond simply a firmware update. Is the software meant to be some sort of cell-by-cell refresh of the data to restore the voltage drift that's causing the performance degradation? If so, can they confirm running the "Performance Restoration Software" is a one-time thing, after which the new firmware will prevent the issue from happening again?
 
Pre-fix
plDYWmI.jpg


Post-fix
aOltjBK.png


Still a bit jittery IMO. Unless that's a HD Tach issue.
 
Still a bit jittery IMO. Unless that's a HD Tach issue.
Click to expand...

I believe this type of benchmark should be jittery. I mean if wear leveling is working sequential blocks (after wear leveling) need not be in sequential cells in the flash leading to lower performance on this kind of benchmark.
 
drescherjm said:
I believe this type of benchmark should be jittery. I mean if wear leveling is working sequential blocks (after wear leveling) need not be in sequential cells in the flash leading to lower performance on this kind of benchmark.
Click to expand...

Check this report--do the "before" and "after" images of a 500GB drive. See how smooth that benchmark is?
 
Its possible the size of the drive (that one is 2 times the size of yours) helped that test. Wear leveling is easier to balance for performance when the drive is larger. Also the frequency and nature of data may play a part on how your data is wear leveled.
 
This is what my laptop looks like post-fix. Don't have any pre-fix HD Tachs, unfortunately. Though it was terrible.

FmVVPSs.png


It is also extremely jittery. 750GB drive. I doubt SATA3Gbps has anything to do with it (XPS1640 doesn't do SATA6Gbps).
 
1TB Evo, 400GB used for Steam games and nothing else, just completed my nearly 3 hour journey to complete this.

Before:
2VgQX02.png


After:
ywIvdbd.png


Interesting that the area that appears to be empty still has faster reads. Also, access times had slowed greatly and have been restored.
 
This is not so unusual. Since the area was not written to, the controller does not even have to access actual NAND other than the mapping table to return empty sectors.
 
I wonder how this new firmware is going to effect drive lifespan. Has Samsung stated what exactly the problem was? I've assumed the issue is the TLC NAND itself not holding a charge properly and that data has to be moved around constantly to maintain the drive's rated speed.

That and WTF does this line in the installation guide mean? Vaccine?

15) Performance Restoration can’t communicate with SSD possibly due to security, Vaccine or DRM application.
Click to expand...

I guess Samsung can't afford a proper English translator after all the money they spent advertising these TLC NAND drives?
 
Last edited:
DeathFromBelow said:
I wonder how this new firmware is going to effect drive lifespan. Has Samsung stated what exactly the problem was? I've assumed the issue is the TLC NAND itself not holding a charge properly and that data has to be moved around constantly to maintain the drive's rated speed.

That and WTF does this line in the installation guide mean? Vaccine?



I guess Samsung can't afford a proper English translator after all the money they spent advertising these TLC NAND drives?
Click to expand...

Probably Virus/Malware protection. What, you're not a fan of engrish? Well no smorking for you buddy!
 
DeathFromBelow said:
I wonder how this new firmware is going to effect drive lifespan. Has Samsung stated what exactly the problem was? I've assumed the issue is the TLC NAND itself not holding a charge properly and that data has to be moved around constantly to maintain the drive's rated speed.
Click to expand...

According to Magician, no new data was written to the drive, so I don't think lifespan is going to be an issue.
 
Jeremy C said:
According to Magician, no new data was written to the drive, so I don't think lifespan is going to be an issue.
Click to expand...

Then how was the issue fixed? Magic?

Its obviously not sitting there doing nothing during the patching process. I would assume they could count internal data reshuffling seperately from new writes.

The lack of information from Samsung is really ticking me off. I have a lot of these drives in production machines.
 
Glad to see the new firmware/fix.

My case is a little different its two mSata 840Evo in raid 0, i kinda think i need to break the raid and flash them like that, but idk if there is a simple fimware bios flash? i just want to upgrade the firmware and then do hard erase on each and rebuild my raid array later and install everything again, not a big deal.

Anyone know if its flashable on DOS firmware or if there is a version to do it like that?
 
Here's what I have been reading about the fix.

There was a bug in the firmware that did not acount for voltage drift on old data blocks. It's normal on all types of NAND for cell voltage to drift over time, and when the controller detects errors due to the drift, it's supposed to make a correction that eliminates read errors. Sort of llike how you correct your steering wheel when the road tilts to one side, you counter the drift to keep the car straight. So the controller was in constant ECC mode, slowing read output having to correct the errors. TLC is supposed to be more sensative to drift as well, meaning less drift returns errors sooner than with MLC.

The restoration tool updates to the fixed firmaware, and refreshes each occupied data block at the NAND level. You only have to run it once, in fact you can ONLY run it once per drive.

The pain for me was the 250GB in py PS3, I had to make a partition in Windows and run the tool, then completely restore my PS3 data. The 250GB (90GB free) Application drive in my PC took all of 30 minutes. It took longer to make a back up with Ghost than to run the fix.

Currently the tool only works on NTFS partitions in Windows. Also they claim it only works in AHCI mode on standard MS drivers, Intel drivers, or the latest AMD drivers, though I read of one case where it worked on an old nVidia chipset in nonAHCI mode.

Hopefully the controller will now account for drift on old data blocks. I have noticed that my EVO runs about 5c hotter than before at idle, which seems unusual, unless the new FW has made GC more agressive.

http://techreport.com/review/27212/samsung-840-evo-update-fixes-slow-reads-with-old-data

http://www.samsung.com/global/busin...Samsung_SSD840EVO_Performance_Restoration.zip
 
Last edited:
iL8kOMa.png


Nope.

Secure-erased using Samsung's tool and my laptop STILL has the jitters.

Someone suggested it was just "the way I use the drive" but I find that explanation severely lacking considering every press outlet and every other user has reported a much-smoothed HD Tach measurement and mine's still as jittery as a five year old who drank too much soda.

EDIT: HD Tune disagrees.

dmUQ1Wr.png


I give up.
 
Last edited:
Thanks, I didn't think the fix was out yet, running now.
 
Check the power settings. Some settings severely impact performance. For the benchmark you can deactivate the power saving options and set them to maxium performance.
 
drescherjm said:
Hmm. I would have expected the reverse and that the benchmarks were somewhat like measuring 64 KB random io versus large sequential.
Click to expand...

Even when I match them (HD Tune will do up to 8MB, HD Tach can be told to use 8MB) HD Tach still shows jitters.

OC.net suggests that there are problems with HD Tach being jittery, but I haven't seen any graphs other than mine displaying the behavior.
 
You must log in or register to reply here.
Back
Top