Request for Anand regarding Sandforce SSDs

[Elixer]

I also know for a fact that there is MASSIVE public assumption being made out there and there are VERY FEW actual damaged/dead drives and almost every single one sent back can be destructive flashed right back to life. For that reason alone I would love to get my hands on a case or 2 of those RMA'd drives along with the engineer's flash needed to revive them.

Assuming that was the case, and there is a way to flash the drives back to life, then why would OCZ (or whatever the brand is) not release said utility and save a crap load of $$$ by doing fedex shipments on new units?
I am not aware of any refurbished SSDs, so where are all these RMAs ending up?

 

[MobiusPizza]

Assuming that was the case, and there is a way to flash the drives back to life, then why would OCZ (or whatever the brand is) not release said utility and save a crap load of $$$ by doing fedex shipments on new units?
I am not aware of any refurbished SSDs, so where are all these RMAs ending up?

Sold as new

To me the problem seems to be just the mishandling of sandforce with regard to data blocks with checksum errors. Any attempts to write to or read those blocks, the drive becomes so sluggish and irresponsive that the OS died on it with BSOD. Whatever caused the checksum error in the first place, and the error handling, are the culprits

As to the reason why re-flashing or secure erase fix the drives, it's because the checksum CRC values gets deleted.

 

[groberts101]

Assuming that was the case, and there is a way to flash the drives back to life, then why would OCZ (or whatever the brand is) not release said utility and save a crap load of $$$ by doing fedex shipments on new units?
I am not aware of any refurbished SSDs, so where are all these RMAs ending up?

I can honestly only answer the first question.

Because Sandforce will not allow the engineer's flash to be released due to low level access to the firmware and potential reverse engineering/customizing of it.

Then add to that the fact that OCZ(or any other vendor using the same SF chip) will not allow opening the drive to flash it without voiding the warranty.

I completely agree though.. it would be far better to have a destructive flash made available as the Indilinx drives do. Since that particular d-flash was released.. the RMA of those drives has slowed to a trickle. Now everyone just requests that tool to revive their assumedly "dead" drive.

The very same assumptions are made with these SF drives as well and very few are actually irrecoverable. "Panic locked" is the correct terminology to use for the state that they enter when they go bye-bye from the bios.

 

[Elixer]

Sold as new

To me the problem seems to be just the mishandling of sandforce with regard to data blocks with checksum errors. Any attempts to write to or read those blocks, the drive becomes so sluggish and irresponsive that the OS died on it with BSOD. Whatever caused the checksum error in the first place, and the error handling, are the culprits

As to the reason why re-flashing or secure erase fix the drives, it's because the checksum CRC values gets deleted.

There is a difference between a checksum error, and the SSD going into 'panic mode' and locking the drive so you can't do anything to it.
Once the drive is in the panic state, there is nothing anyone can do to 'reset' it, and if there was, why wouldn't they release this utility instead of forcing people to RMA the units. There is more to it then just a reflash.
Panic mode on a vertex 2 is when the SSD shows both the green & red LED on at the same time.

 

[groberts101]

see the response above yours for the correct anwers to that question. Resetting these drives is as simple as opening the drive.. probing 2 pins inside it and running an engineering flash that accesses the entire firmware.

End user firmware is a watered down limited version which only hits certain parts of the drive.

Originally, Sandforce was going to let OCZ have a toolbox with the option to set the drives physically available space(increase/decrease OP).. but they backed out after realizing the potential issues that could arise from further reaching access into the controller's firmware.

In the end?.. they are just protecting their IP and could care less about the vendors they sell their chips to for having to RMA a drive to straighten them out in-house(at the vendor).

 

[MobiusPizza]

There is a difference between a checksum error, and the SSD going into 'panic mode' and locking the drive so you can't do anything to it.
Once the drive is in the panic state, there is nothing anyone can do to 'reset' it, and if there was, why wouldn't they release this utility instead of forcing people to RMA the units. There is more to it then just a reflash.
Panic mode on a vertex 2 is when the SSD shows both the green & red LED on at the same time.

Well what I meant was, checksum errors trigger the drive to go into panic mode. Sandforce can either prevent this from happening through more robust firmware which doesn't produce as much checksum errors, or through fixing how the drive reacts to these errors, i.e. not to panic.

 

[Elixer]

see the response above yours for the correct anwers to that question. Resetting these drives is as simple as opening the drive.. probing 2 pins inside it and running an engineering flash that accesses the entire firmware.

End user firmware is a watered down limited version which only hits certain parts of the drive.

Originally, Sandforce was going to let OCZ have a toolbox with the option to set the drives physically available space(increase/decrease OP).. but they backed out after realizing the potential issues that could arise from further reaching access into the controller's firmware.

In the end?.. they are just protecting their IP and could care less about the vendors they sell their chips to for having to RMA a drive to straighten them out in-house(at the vendor).

Heh, you replied when I was typing, and then I took a bit of a break, and didn't bother to see if there were any replies...

You wouldn't by chance know which pins you probe ? I am sure once the warranty on these things is up, people with bricked SSDs would like that info, and see if they can figure a way to save these things.

 

[groberts101]

sorry.. no I don't. The one's who did pull data could not decipher the jibberish and reverse engineering would surely be required unless the engineer's flash ever leaked out. Never hurts to dream though.

Doubt there will be much market for these things when the warranty expires and they'll be piled up like too small USB sticks with limited usage.

Plus.. the USB sticks will probably be far faster by then with better firmware anyways. lol

Here's proof that we're already almost there.
http://thessdreview.com/our-reviews...h-drive-review-the-possibilities-are-endless/

 

[MobiusPizza]

Here's proof that we're already almost there.
http://thessdreview.com/our-reviews...h-drive-review-the-possibilities-are-endless/

Super Talent RC8 USB 3.0 (SF-1200)
I would love to get a BSOD from a USB stick using sandforce controller

Guys, if you have problems with SF I suggest you email Anand and see if he can contact Sandforce on behalf of us. This is the most read topic apart from the stickies in this forum. I've emailed him but I think more people needs to say so before the issue gets noticed.

 

[groberts101]

While I hear that.. I guess I wouldn't really be too worried myself as I don't have any issues with 6 x V2 in R0 along with 1 single V3 added for good measure.

Never know though?(and you're certainly right to mistrust).. maybe the USB 3.0 controller will not play nice with the firmware on that drive too. Sure wouldn't be much of a stretch to think some could have issue.

And if they don't have issue?.. well.. maybe they need to consider using the USB sticks firmware revision for the SSD's.

 

[MobiusPizza]

While I hear that.. I guess I wouldn't really be too worried myself as I don't have any issues with 6 x V2 in R0 along with 1 single V3 added for good measure.

oO 6 Vertex 2s in RAID 0?
OMG

Back to topic, I myself suspect the problem may be caused by Trim. Some people reported IDE mode solved the issue, which does not support trim. If you never had any problem in RAID, it may be down to the same reason too since trim doesn't work in RAID.

 

[groberts101]

I've had them split up and running as single drives more than a few times for testing along the way(TRIM and no-TRIM).

My system is just stable and my bios gets along perfectly with these controllers is my main strongpoint.

But you are right.. some have found resolution by disabling TRIM commands from being sent to the drive. Of course.. the same thing could be said about the pagefile being eliminated form the SSD as well.

That's some of the biggest issue with these problems and makes it difficult to narrow down since not any 1 thing will work for all.

 

[wildcat293]

Yes it is silly for someone to ask the owner of the site to take action against OCZ however I would love to see an article on these problems. The SSD articles here are far above anything else I have read, and so far there has only been speculation and rumor on the issue (and nothing at all from OCZ). Hoping we can get some facts (and maybe a fix) from someone!

Until then I am avoiding OCZ...

 

[scajjr2]

It isn't just OCZ. I have a Kingston HyperX 120Gb, about 5 weeks old and have had it act up twice (about every 2-1/2 weeks it seems). First time was just a freeze up & had to hard reboot to get system to see drive. Last night it was a BSOD (had left computer on overnight to do a download) but for some reason Win7 didn't save the minidump file (and it is set to do so).

After my first incident I had a few emails back&forth w/Kingston. They thought it may be just a fluke, said they weren't having reports of issues. Now that their SF2281 based drives have been out for a bit I'm betting they start to get just as many reports as other SF2281 SSD makers.

Sam

TA870U3+
X3-720BE
8Gb DDR3-1333
MSI 6950 TwinFrozr II 2Gb
HyperX 120Gb
Hitachi 3TB
WD 1Tb

 

[MobiusPizza]

I've seen reports of problems from customers of big and small companies. OCZ, Corsair, mine is Renice.

 

[86waterpumper]

the 2281 is junk period. I don't think it matters what brand it is. My owc has been flashed to the latest firmware and it did not help one bit. Still freezes and bsod at random. It is super frustrating to spend that kind of money on a drive and it be such a turd. Oh well I learned my lesson I won't get burned by sandforce again

 

[MobiusPizza]

To keep everyone posted, it seems that the Renice SSD was from a bad batch and the support team assured me it was not a firmware issue, but isolated hardware problem with a small batch.

 

[zlejedi]

I have a simple rule that saves me all the trouble with SSD - nothings with letters OCZ and SAND FORCE on it goes inside my rig.

So far 100% problem free solution