SSDs and Reliability

[Voo]

So it's really simple but you don't have any proof for your claims? Because their specs are pretty much similar and they have the same PE cycles.

I've yet to see anyone running out of PE cycles on their SF drives, from a performance point of view there seems nothing amiss, but it's totally obvious - yep. If that's really the case than I'm sure someone showed that the flash isn't good for its published 3k PE cycles and I'd love to read that.

You know it's also totally obvious that the earth is flat because otherwise we'd fall from it into the sky.

 

[bulanula]

Believe whatever you want to belive but for me one thing is clear :

Micron > Spectek
Tier 1 > Tier 2
Buying from NAND manufacturers that design the NAND, make their own controller, use their own RAM cache and make their own firmware > Buying from marketing companies like OCZ who just buy components on open market, assemble the SSDs and sell them with their own firmware

 

[Voo]

Well you can make assertions all day, but if your complete proof for them comes down to "But I believe it's so" than you should be at least honest enough and declare them as guesswork and not facts.

Because facts as we all know are verifiable statements about reality whereas guesswork clearly is not. So if you have proof that shows that SpecTek flash doesn't fulfill its published specs or is otherwise verifiably inferior to IMFT flash that's great - we're all ears, but otherwise that's nothing but FUD.

 

[bulanula]

Because facts as we all know are verifiable statements about reality whereas guesswork clearly is not. So if you have proof that shows that SpecTek flash doesn't fulfill its published specs or is otherwise verifiably inferior to IMFT flash that's great - we're all ears, but otherwise that's nothing but FUD.

Well then look at the more subtle things :

-OWC post did not get taken down by OCZ lawyers
-Spectek website excerpts : "Whether it's MP3 devices, USB drives, or cell phones...our NAND Flash components deliver exceptional performance for your applications." "With growth in today’s NAND market being driven largely by Flash memory cards, USB Flash drives, MP3 players, cellular phones, and other portable devices, SpecTek is ready to deliver."

Show me where Spectek recommends using their flash in SSD, please.

 

[Voo]

Umn, sorry but I think we have EXTREMELY different conceptions of "proof" (they didn't take some insubstantial claim on a blog down! Well now then everything's clear) also usually I try not to evaluate a product based on marketing claims - you know hard facts are "somewhat" more reliable (hey Intel controllers have a WA of <1.1! Yeah sure, every RL test refutes that, but it says so on the marketing slides! Must be true then)
Fact is SpekTec flash comes from exactly the same fab as Intel's flash and is rated for the same number of PE cycles.

So please show some independent tests that show that SpecTek flash isn't good for their 3k PE cycles or violates their official specs in some other way.

 

[Mark R]

The controller adjusts for all sorts of variables, including latencies. OCZ purchases NAND from different manufacturers, and likely batches that were binned differently. We are talking about the bottom of the barrel here. Flash procedures change based on NAND type. For constancy, buy in-house brands like Intel and McDonalds, not OCZ and the local diner. For the record, my Solid 2 (with Intel NAND) is going strong after almost 1.5 yrs.

That's not the whole story.

Intel's controller is very much a generic controller, and the algorithms are generic, albeit with parameters adjusted to suit the NAND.

Sandforce controllers have multiple different firmwares, which are tuned to the unique properties of different brands of NAND. Sandforce spent a considerable amount of effort analysing different brands of NAND, and making unique algorithms for the characteristic patterns of degradation that occur in different brands of NAND. This is a unique selling point of Sandforce, as no other controllers (except for Fusion IO) actually tune the programming and ECC algorithms for the brand of NAND used.

 

[Mark R]

Fact is SpekTec flash comes from exactly the same fab as Intel's flash and is rated for the same number of PE cycles.

Not only that, but SpekTec flash is more intensely tested than Intel's own name-brand stuff. That is spektec's business, sophisticated testing and binning of ICs. They are able to buy in dies and test them, when the OEMs don't want to spend the money on testing.

Intel selects its flash based on minimal testing and statistics. They choose a few dies to test thoroughly, and if they all pass, they don't bother fully testing the rest of the batch (the chips just get a quick check). They assume that if 10% of the chips are A1-prime quality, then the other 90% are.

SpekTec buys batches where the sampling has shown that some chips might be flaky. They then test every single die thoroughly, and carefully grade each individual die. The SpekTec chips that have been showing up on OCZ drives are those that have been tested to 'enhanced' specifications (i.e. tested to a higher level than the OEM used for testing).

 

[bulanula]

OK so you say that Spectek + Sandforce is a good combination and even better than Intel NAND + Intel controller ?

Quite interesting.

 

[Voo]

OK so you say that Spectek + Sandforce is a good combination and even better than Intel NAND + Intel controller ?

Quite interesting.

No we're saying that the flash isn't the problem - someone here claimed that in other forums you find lots of people whose flash died, but he also didn't follow up when asked for some links. So I'm just assuming (and really I've looked a bit in Intel/OCZs forums) that anywhere else is just like here and people using up their PE cycles (can't think of even one case) or having problems with the flash are an extreme minority.

The problem for SSDs has so far always been the controller and from a compatibility point of view that's where Intel shines, but that has nothing to do with the used flash.

 

[groberts101]

OK so you say that Spectek + Sandforce is a good combination and even better than Intel NAND + Intel controller ?

If you understood the strong points of the Sandforce controller you wouldn't even ask that question. Durawrite throttling and compression algorithms make it possible for that particular controller to outlive any other even when using nand with much less PE/c. Is the major benefit of compression and was the way it was designed from day 1.

In fact, the first gen SF drives were so underrated for the amount of logical(not physical) writes that they can endure compared to traditional controllers?.. they have actually been able to use lower PE/c(25nm nand), increase GC aggressiveness(this alone is a biggie for WA), AND reduce throttling quite substantially all at the same time. That speaks volumes when 3 year warranties don't get changed at all.

And in a nutshell.. I could care less if my drive burns all the nand in 7 years as it'll seem like a slow and tiny little USB stick by then. Although I'm pretty sure some of the Intel guys will be happy to mention that theirs lasted for 9. lol

 

[bulanula]

Voo, I completely agree with you that the NAND is not the problem ( so far) but that the controller is to blame, especially in the case of sandforce.

Groberts, that is why I originally chose Sandforce because of their compression algorithm that is THE ONLY one that has a WA factor of 0.56 but the reliability issues put me off.

All the other controllers have a stupidly high WA factor because they don't use compression. I indeed wonder if any other manufacturer will be smart enough to come up with compression in their controllers or maybe only SF has the needed IP and nobody else can use compression in their SSD controllers?

 

[thescreensavers]

my vertex 2 died in a month, and then the bs ocz did, I wont buy shit from them again

 

[bulanula]

my vertex 2 died in a month, and then the bs ocz did, I wont buy shit from them again

Well this is exactly the reason why I am reluctant to get SF SSD.

Did you use sleep or hibernation?
Did the red and green led appear?
Did it just stop being detected by the BIOS? Tried an external enclosure etc.
What did ocz say eg did you get replacement or refund etcetera ?

 

[Elixer]

my vertex 2 died in a month, and then the bs ocz did, I wont buy shit from them again

So, what did they do ?

Well this is exactly the reason why I am reluctant to get SF SSD.

Did you use sleep or hibernation?
Did the red and green led appear?
Did it just stop being detected by the BIOS? Tried an external enclosure etc.
What did ocz say eg did you get replacement or refund etcetera ?

For what it is worth, my drive just died in the middle of posting a post on this forum.
It was a OCZ Vertext 2, and the red+green LED just came on.
On reboot, BIOS didn't see the drive anymore. I tried multiple machines with no luck.
My drive was almost 2 months old, and the RMA is still in progress.

The question is, if the Sandforce firmware 'panics', what actually happens ? Does it enter some kind of special mode waiting for it to be reset, or did it just crash with no hope of recovery ?
If this is common, then why isn't there a jumper or a reset switch made available, that people can 'reboot' the firmware ?

 

[groberts101]

what firmware came shipped with the drive and system did you have it installed on?

 

[Elixer]

what firmware came shipped with the drive and system did you have it installed on?

I think it was 1.32.
System is a AMD based vista 64bit, AHCI drivers enabled.

 

[bulanula]

If I ever get a SF based drive I will just be praying that the same things does not happen to mine.
Really a pity because SF is the fastest 3gbps and 6gbps controller you can get with the lowest WA etc.
The implementation sucked because they did not bother to test it properly before it got released to the masses.