Results of one month of testing 13 Vertex 3's with 2.15 firmware

Interesting story.

Result Results of one month of testing 13, Vertex 3's with 2.15 firmware

 

Interesting link Phil. Thanks. Amazing that the V2's work but the V3's don't, and they (seem) to think it has something to do with other hardware.

 

The only OCZ drive that I have that still works is a rebadged Samsung 64Gb SLC drive.

 

Going over the Newegg ratings quick there seems that OCZ atleast have done something right. The last month, november until now have been almost flawless ratings. People reporting stable drives causing no problems with the newest firmware 2.15. Rewind 3 months ago or something and almost every rating was bad.
http://www.newegg.com/Product/Product.aspx?Item=N82E16820227706

Could it be that this guy who tested the 13 drives somehow got a bad batch?

The majority are saying the firmware fixed their BSODs here. But I should also mention that some still say they have problems. A bit of a mixed mess really...
General Discussion 2.15 is now live...use this thread for questions and all relevant discussion.

 

I suppose my reaction is "who cares?" They have proved themselves unreliable at best so are to be avoided.

I suppose if you own one you are very interested in new firmware, but if you are buying forget it.

 

Well the reason I posted it was that there were some claims that firmware 2.15 had completely solved the BSOD issue. This turns out to be not the case.

 

Thanks for the link, Phil

Guys, this shouldn't be a secret. There is a reason why so many of us are using M4's

And the vanilla Vertex 3 with the cheap 25nm nand memory coupled with a sandforce controller (that only works well with the 32nm memory found in more stable drives like the Vertex 3 Max IOPS version, Patriot Wildfire, Mushkin Deluxe)...will lead to a lot of BSODS.

The whole line of cheaper sandforce drives with cheap 25nm nand is completely flawed. Firmware can only do so much

It's also interesting to note that the new OCZ Octane drives use the latest Indulinx controller. OCZ went out and bought Indulinx perhaps in an effort to shed the bad reputation

 

Nothing wrong with the IMFT 25nm NAND though. It's found in the Vertex 3, Crucial M4 and Intel 320.

I think the reliability problems come from the controller and firmware.

 

Yes, you are very correct on this...its the combination imo of the 25nm nand, the current gen sandforce controller, plus a firmware that tries to tweak too much performance out of the drive. It completely makes things unstable

 

Interesting stuff, where did you get this from?

I had one Vertex 3 OEM fail on me, I think it had Toshiba 32nm NAND inside.

 

I can't quote one source. But i was very active on a few forums and we had very deep discussions about what changed that caused instability. That's why i said "imo"...but there is quite a bit of factual basis for my comments based on a correlation of evidence found in reviews, threads etc

Away from the internet, from a personal experience perspective, on my street, there are 3-4 tech geeks and we often get together and shoot the sh*t about stuff over beers. We all went out and got different SSD's. I have a patriot wildfire in my desktop with the sandforce 2281 controller but with the more expensive 32nm nand. Not one bsod. There are also 2 Vertex 3 Max IOPS among us and no bsods. My neighbor across the street has had 2 returns on a Corsair drive (sandforce 2281 + 25nm nand). He then went with a 240 gigs Vertex 3 hoping for better results, and it was great for a 99% of the time...but that 1% was fatal and required a reinstall.

he's since moved to a M4 and hasn't had any issues since.

I've also tried Vertex 2's with both hynix 32nm nand and intel 25nm nand. NO ISSUES what so ever. But the vertex 2's used a sandforce 12xx or 15xx controller on SATA 2 which was completely stable.

So we have to look at the variables and draw some conclusions. IMO, they tried to tweak too much speed out of the sandforce 2281 using fast 25nm nand. The 34nm nand technically isn't as fast so, the speed of this can be causing issues also

 

Kingston HyperX (32nm afaik) also seems to do quite good. 5 star rating on Newegg.

I think you were very lucky. Vertex 2 has the worst imaginable track record on Newegg.

 

i take the reviews on newegg with a grain of salt ... i've got the two vertex 2 60 gig drives in raid 0 on a machine in my basement. Still running fine

 

and Intel too... Reliability has much higher precedence than the performance for me (and for several others as well).


--

 

I too have a OCZ Vertex 2 60GB in my desktop and yet to come across any issues, though I was lucky to have the first generation (32nm) version and not their newer 25nm version which caused a stir amongst some of their users for being inferior performance wise.

 

A lot of negative reviews on newegg doesn't mean that every unit will fail or cause problems. It only means that there is a more widespread issue, there will always be bad units from every manufacturer. A bad track record on newegg, usually means higher probabilities of problems rather than assured problems.

 

I think there are different issues.

Microsoft drivers send a "Trim" that contains more sectors than the OCZ controller are designed to handle.

Intels old RST drivers put SSD and PCIe bus into too much snoozland. (or maybe they fix the issue in the drive by not sleeping the bus when you have one or something)

Maybe more issues, but you would have to read what OCZ tell you on the forum, not what people "suppose". Including me... but who do you trust when noone wants to claim responsibility.

Personally, I returned OCZ's first sata USB stick years ago and reported that it was failing and losing Sector Zero, but I doubt anyone cared at the time. All candy, lucky I had that experience or I would have done myself on those untested drives as well, lol.

 

Would you explain a little further please. AFAIK the ATA identify command should tell Windows just how many counts of LBA's it can handle, is windows ignoring this?

 

I would have to google it again, was a few weeks ago I read it somewhere. I used to word "sectors" cos I was lazy, but whatever it is, ms was telling the controller to trim more than it was specced to "per message" or "per operation" or whatever. I think it was the controller OCZ use. I don't follow it much as I don't have any recent SSD's.

The Intel drivers do send the correct amount of trim. Although Intel had some sort of bug or fixed some sort of problem to do with many different drives not getting enough power on the bus when the PCIe was in low power... hence all the blue screen and corruption posts on the internet atm.

This is just my guess, as I inferred it is really up to the companies to tell you what is going on and I may have read something that was wrong. Companies like Intel admit to mistakes and don't fuss on if it was their fault or not, which is what you want.

EDIT:never mind.. that was vertex 2
http://www.behardware.com/news/10962/sandforce-trim-listen-up.html
sorry

looks like the say the newer drives are fixed as well, although I would have a google alert set up for this if I had a drive, lol
http://hothardware.com/News/OCZ-Stomps-Out-SandForce-BSOD-Bug-with-Firmware-Update/

p.s. Also you need to update the motherboard firmware and the Intel RST if you have a newer MB that sleeps the BUS.

 

Thanks for the links fileant. While one of those articles is quite old it's interesting that they mention word 105 and how it is supposed tell the OS how many counts it can handle per request. IIRC unless specs have changed, I only have draft copies as the real ones require a fair bit of money to obtain, for my agility that word is 1. That means if the LBA's are contiguous then one trim command can release upto ~33MB of media. It's strange that the OCZ SMART data marks that word as reserved instead of showing what it really is.

IMHO if the drive can not handle excessive requests from windows then it should reject the requests from the driver so as to not end up freezing the drive. It's not a perfect world and the firmware should be able to handle bad/illegal packets by rejection rather than dropping out of service.

I have yet to have a problem with my own Agility3's but they are in RAID0 so do not receive TRIM commands. This may change with a later driver so I'm a little concerned whether TRIM will improve things or cause problems. Of course it should be possible to disable TRIM in the event it is problematic.