SandForce Identifies Firmware Bug Causing BSOD Issue, Fix Available Today
by Anand Lal Shimpi on October 17, 2011 8:20 PM ESTAfter months of end user complaints, SandForce has finally duplicated, verified and provided a fix for the infamous BSOD/disconnect issue that affected SF-2200 based SSDs. The root cause is a bug in the firmware, although specifics are pretty slim. The typically sparse release notes just state a rare condition resulting in a blue screen error has now been addressed. OCZ has been testing the fix in-house for the past three weeks and now believes it is fit for public release. SandForce is simultaneously releasing the firmware to its partners, so if you have non-OCZ drives you should contact your drive manufacturer for an availability update.
The new firmware is version 2.15 for OCZ drives and 3.3.2 for drives that use SF's standard numbering system. It's important to note that the fix here is for the bug that SF discovered and may not solve all pending issues. Given the extremely long discovery and fix time for this issue I still believe the best policy towards SandForce drives is one of caution, at least until we start hearing from users as to whether or not this fix worked (and didn't create any new issues).
Update: OCZ sent along its official statement -
Source: OCZ
Facebook
Twitter
RSS
30 Comments
View All Comments
Aphelion02 - Monday, October 17, 2011 - link
" OCZ has been testing the fix in-house for the past three weeks and now believes it is fit for public release. "OCZ believes a lot of things are fit for public release that I would never buy. I'll trust this when I read all about the technical details in another one of Anand's excellent reviews.
ss284 - Monday, October 17, 2011 - link
BSOD issues since release(over a year ago) and there still isn't a fix out. Not surprising given sandforce's track record.Siorus - Wednesday, October 19, 2011 - link
I'm not disputing the veracity of your comment at all, but it doesn't match my experience, for whatever that's worth.i've got about 20 SF-1200 drives of various brands deployed at work and in my own personal machines, I've had exactly zero issues with them. I think I've got at least one or two of every brand that's available in the US, in fact, except for the OCZ and OWC drives.
I've also got a bunch of Intels (an X25-M G2, a couple 320s, a ton of 510s), a C300, some OCZ Summits (Samsung-based) and some Indilinx-based drives. Of the bunch, I've only had an issue with one-a 120GB Vertex. I wouldn't trust another OCZ drive, but I haven't personally had-or even really heard of-any large-scale issues with the SF-1200s.
Taft12 - Monday, October 17, 2011 - link
Lesson learned - Intel SSDs only.pc_void - Monday, October 17, 2011 - link
Intel SSDs haven't had major issues? That's news to me.ckryan - Monday, October 17, 2011 - link
The 8meg bug was pretty rare with the 320s. Some Marvell controlled drives (including the 510)had LPM problems, but that was/is easily fixed. There was a G2 firmware problem, but the bug was easily identified in all of those cases, and you knew that just getting the drive RMAd would fix the problem. With the SandForce problem, you could go though drive after drive... the problem was with all of them, but only manifested itself through certain hardware and usage patterns. I would say Intel controlled and Marvell controlled drives are still mostly the way to go, but Toggle NAND equipped SF drives are worth the problems (sometimes). Whether this new "magic" OCZ only-at-this-point FW fix works on drives as a whole is another matter entirely. There were a couple of ways to make a SF2281 unhappy, and it's unclear whether this addresses all of them. I can get two of them to occur, so when the 3.3.2 FW drops, I'll peep it out.icrf - Monday, October 17, 2011 - link
I have an Intel 320 at work and an OCZ Vertex 3 at home. I BSOD'd at home but the beta update from May/June fixed that just fine, no issues since.I haven't had any issues at work, but Intel recommends updating to their "fixed" firmware, anyway, as the bug can strike any time. I read a little more and people who had never seen the bug all the sudden got it right after apply the patch that should have prevented it.
I still haven't updated the patch on my Intel at work. While I don't have any statistical numbers to say for sure, the whole thing reads like a crap shoot, no matter whose drive you buy.
Play the odds, and hope you're not in the minority with issues, like everything else in life.
ckryan - Monday, October 17, 2011 - link
If your non-SF drive dies or is problematic, you can just swap it out with the vendor or retailer. If your SF22xx drive BSOD/Disconnects/etc, you can't do too much about it. The problem is with all the 22xxs. Luckily for everyone, it doesn't get exposed THAT much -- just enough to be a major problem for the SF vendors. I think a lot of the FW updates have basically pushed the problem around the plate for some, so that some get a perceived fix but some get the problem for the first time whereas they would not have normally seen the issue in their workload.seapeople - Tuesday, October 18, 2011 - link
Back up then apply the patch for the Intel. If you get the bug immediately after, then you get a new drive in the mail and lose no data.It probably won't matter though... I had an unpatched 320 for several months with a dying graphics card that lead to 50+ "unsafe power cycles" as stated in the Intel SSD toolbox and I never had a problem with the drive. Regardless, I have since patched my 320 and never had any problems before or since.
Movieman420 - Monday, October 17, 2011 - link
Well brace yourself...Intel will start selling SF drives next month...aka Cherryville 520. ;)