Samsung Explains The Firmware Bug Causing The Failures of SSD 840/840 Pro
by Kristian Vättö on December 7, 2012 1:05 PM EST- Posted in
- Storage
- SSDs
- Samsung
- Samsung SSD 840
We've been covering the issues surrounding Samsung's SSD 840/840 Pro lately. The issue was first discovered when Anand's pre-production review sample died during testing and we also noted that in our initial review. Samsung quickly sent us another drive but it also failed after a couple of days of testing. My SSD 840 managed over a month but ironically enough, it died right after I had completed endurance testing.
Earlier Samsung told us that all review samples including our three shipped with a pre-production firmware that had a bug in it causing the failures (retail units were shipped with a newer firmware without the bug). At the time we didn't know what exactly was wrong in the firmware, but now we do. When the drive was issued a secure erase command, it would clear all table mapping information at the Address Translation Layer (ATL) but not at the Host Interface Layer (HIL). The data in both layers needs to be up-to-date for the drive operate properly, so when a write request came in, the controller wasn't able to map the data correctly, which caused the firmware to hang. An SSD obviously can't operate without a functioning firmware so from a user's standpoint, it looked like the drive had completely died even though only its firmware was broken.
All our three failures support this explanation. Our first 840 Pro sample died during a 128KB sequential write pass that we use to pre-condition our drives for enterprise tests, but the drive was secure erased just before beginning to fill the drive. The second 840 Pro died during power consumption testing but again it was secure erased right before starting the test. The regular 840 actually died when I tried to secure erase it. The secure erase command resulted in an error so I power cycled the drive but it was no longer detected by the system after reconnecting it.
| Comparison of Samsung SSD Firmware Versions | ||
| Pre-Production | Retail | |
| Samsung SSD 840 Pro | DXM02B0Q | DXM03B0Q |
| Samsung SSD 840 | DXT05B0Q | DXT06B0Q |
The good news is that all retail units have shipped with a newer firmware, only reviewers and others who have access to pre-production units were affected by this bug.
For users considering the SSD 840/840 Pro, this should be reassuring news. The 840 Pro is still the fastest SATA 6Gbps SSD we have tested and it's definitely one of the top choices where performance is concerned today. The TLC NAND based SSD 840 is more mainstream focused but from what I have seen, it seems to be fairly competitively priced. The SSD 830 spoiled many with low prices but that was only to clear stocks. If you can still find a bargain SSD 830, don't hesitate to pull the trigger as those won't be available much longer, but we're more comfortable recommending the 840/840 Pro now.
While long-term reliability is still unknown, if history is any indication the 840/840 Pro are in good company as the 830 was a solid drive. Our third 840 Pro with the new fixed firmware has been going strong for weeks now and we have even recreated the scenarios that killed the earlier samples. We are also waiting for more samples from Samsung to test all capacities of SSD 840 and 840 Pro, so stay tuned!
Facebook
Twitter
RSS
56 Comments
View All Comments
Death666Angel - Saturday, December 8, 2012 - link
In Anandtechs light workload benchmark, the 840 is faster than the 830, in the heavy benchmark it's the other way around. In the read/write benchmarks they trade blows with the 840 being a lot better in random r/w benchmarks.It also dropped price from ~195€ for 250GB at introduction to 155€ right now. The lowest the 830 256GB got was 145€ and it is now at 165€. I don't think there is anything wrong with these prices and I expect the 840 to drop even lower.
Now, the 840 Pro is really too expensive for normal desktop consumer users. But if the extra speed will increase your work productivity, it will probably be worth it.
I'll be looking forward to a 250 GB 840 for my system and down the road another 500 GB one for my games. :D
And the new headline is much better and less confusing, thanks for that.
Also, is there a chance that the spare area test will be done with a 840 non-pro? I'd really be interested in the IO latency @ stock and with 25% spare area. :)
Kristian Vättö - Saturday, December 8, 2012 - link
I don't have a working 840 sample at the moment but I'll be sure to test IO latency once I get a new sample. I'll probably include it in our review of all the capacities of 840/840 Pro.Beenthere - Friday, December 7, 2012 - link
I wonder how many tweaks the "Reviewer" firmware had to get good numbers, that the "Shipping" firmware doesn't have? If this is Samsung's idea of quality control and proper validation of an SSD before releasing it to the meida for review, I'd say they failed miserably.Death666Angel - Saturday, December 8, 2012 - link
If they had waited to release it with the shipping firmware to reviewers, thus delaying any credible reviews until after it was available in retail, I'm sure people would say that they were afraid of bad reviews and wanted to cash in before they started to appear. I don't see the big deal here, pre-retail samples die all the time in every category. That's why they are pre-retail.Impulses - Saturday, December 8, 2012 - link
That's kind of a cynical point of view, though a valid one I guess, I'm sure Anandtech will re-run tests on a different capacity 840/840 Pro at some point though. The bug, as explained in this news article, wouldn't really affect performance, so it doesn't support your theory...I'd say the better question would be whether they discovered this bug on their own and fixed it before drives shipped out, or whether they were only made aware of it once Anand/Kristian and others reported drive failures w/the review samples...
If it's the former, I'd say that speaks highly of Samsung's validation (they were testing it even after review samples were sent out), if it's the latter then I guess it's a bit of black eye on their part but they still deserve credit for a really quick turnaround on the fix.
Henk Poley - Saturday, December 8, 2012 - link
That's why even for a tiny embedded system inside a harddisk, it's smarter to put the boot disk on another storage device.Azethoth - Monday, December 10, 2012 - link
Huh what? How does doubling the number of storage systems used lead to a smaller failure rate?longtom - Saturday, December 8, 2012 - link
I just created an account to say that they haven't fixed it yet.I bought two of these last week (both with the newest firmware - verified) to test under RAID0 for benchmarking and as far as BIOS or Windows is concerned, they don't exist at all when paired in RAID on the AMD SB950 (with newest BIOS firmware).
I know this is a problem with the 840 because, to confirm my suspicions, I pulled several SSDs out of my pile to put on RAID, and they all worked flawlessly.
I am still very disgruntled with this product. "Buyer Beware" until they come out with yet another firmware revision.
frief - Saturday, December 8, 2012 - link
Hi longtom,did you do a secure erase at some time? (I'm a little shy as I've been burnt by Indilinx, Crucial M4 and Seagate Momentus and if I don't learn from that when will I ever learn?)
Kristian Vättö - Saturday, December 8, 2012 - link
Is the problem just when you're RAIDing them (i.e. do they show up normally when not in RAID)?