What's wrong with sandforce?

[groberts101]

I'm sure glad I came to this thread to learn about Sandforce. /sarcasm

In all seriousness I think there were 2 or 3 constructive points made and I thank those who tried to stay on topic.

Hey!.. at least I got 3 good Sandforce related sentence fragments squeezed in there before I started sparring with my boyfriends again.

 

[cytg111]

Hey!.. at least I got 3 good Sandforce related sentence fragments squeezed in there before I started sparring with my boyfriends again.

you are an OCZ beta tester

... wtfh, dude, best advice you're gonna get ill give you right here, get a new nick, new handle, and start over... or just, you know.

 

[josephjpeters]

Hey!.. at least I got 3 good Sandforce related sentence fragments squeezed in there before I started sparring with my boyfriends again.

I don't mind and full disclosure I tend to side with you. I just think it's funny how these threads always seem to turn into the same discussion. People need to get over it. Most of what is said involves products that are more than a year old. Anyone who follows the industry knows that a year is an eternity.

If you're curious about what OCZ's current reliability is, I suggest looking at the Vertex 4 and their current reviews. I'm sure in 6 months we'll see data from that infamous French site indicating it's as reliable as all the others but no one will be here to bring it up because it won't support their belief. Maybe I will set a reminder to start a thread here.

Back on topic: anyone have any clue what the new Sandforce controllers are going to look like next year? I remember hearing a Q2 release next year for their 3000 series. I wonder how it'll compare with OCZ's Barefoot 3 and Samsung's MDX. Will it be able to deal with TLC (something the 2200 series supposedly can't handle).

 

[jwilliams4200]

If you're curious about what OCZ's current reliability is, I suggest looking at the Vertex 4 and their current reviews.

A quick and dirty test is to look at the percentage of below average reviews (1- or 2-eggs) of all models on newegg.

 2.3% = 7/308 Intel 330 (Sandforce)
 6.4% = 91/1411 Samsung 830
 7.0% = 15/214 Samsung 840 + 840 Pro
 8.6% = 25/291 Intel 520 (Sandforce)
12.2% = 82/673 OCZ Vertex 4

It seems OCZ comes out on the bottom again. What a surprise.

By the way, those are NOT return rates. Just use them as a relative index for customer dissatisfaction.

 

[Ao1]

A year is a lifetime and it was over a year that the SF drives had bug issues. Not all the blame can be put at SF’s door however as an SSD can fail for a lot more reasons than just the controller. For example some vendors used the SF reference board, whilst others used custom boards. Vendors also used different quality NAND.

This thread could do with a time line of events. Release dates, bugs identified and fixes. Reference boards vs custom, NAND etc. That way the picture would be a lot clearer and discussion could be more fact based.

OCZ sold more SF drives than all the over vendors put together, so it’s hard to not involve OCZ in a SF discussion, however people that bought a SF drive knew it was bleeding edge technology. Would they have rather waited a year or would they have wanted to get the technology immediately? A lot of xtreme enthusiasts will put performance above stability and would rather have something now and take the risk on stability.

Thinking you can have bleeding edge technology and 100% stability is “optimistic” and the trouble with placing performance above stability when a storage system is involved is that you can lose data and it involves a lot of frustration and reinstalling etc.

 

[Coup27]

A year is a lifetime and it was over a year that the SF drives had bug issues. Not all the blame can be put at SF’s door however as an SSD can fail for a lot more reasons than just the controller. For example some vendors used the SF reference board, whilst others used custom boards. Vendors also used different quality NAND.

This thread could do with a time line of events. Release dates, bugs identified and fixes. Reference boards vs custom, NAND etc. That way the picture would be a lot clearer and discussion could be more fact based.

OCZ sold more SF drives than all the over vendors put together, so it’s hard to not involve OCZ in a SF discussion, however people that bought a SF drive knew it was bleeding edge technology. Would they have rather waited a year or would they have wanted to get the technology immediately? A lot of xtreme enthusiasts will put performance above stability and would rather have something now and take the risk on stability.

Thinking you can have bleeding edge technology and 100% stability is “optimistic” and the trouble with placing performance above stability when a storage system is involved is that you can lose data and it involves a lot of frustration and reinstalling etc.

Firstly it is not even a full year because in Anandtech's 520 review which is dated in February of this year Brian Klug replicated a BSOD scenario with the latest firmware at the time, which incidentally was also heralded as being finally bug free.

This discussion has also provided a number of facts, none of which seem to want to be accepted by the pro OCZ/SF camp. A fact is a fact. OCZ drives have the highest failure rates and the highest number of dissatisfied customers according to the leading US e-tailer. Maybe as they're not French, these numbers will be accepted this time?

Your final point about having bleeding edge technology at the expense of stability is a cop out IMO. Intel, Samsung, Marvell and Toshiba were all providing stable controllers during the period in question.

 

[Ao1]

I am not sticking up for OZC and if you knew me you would know I do not like them, but OCZ offer bleeding edge technology at low cost and if you buy their products you must know that and accept the risks and there are a lot of people out there prepared to do that. Good luck to them. Remember that at the time Intel and others were “behind” the big 0 fill numbers and that was all that mattered to a lot of people.

As I mentioned a time line would be very helpful in this thread. There were defiantly issues related to custom vs reference boards. There are a number of other variables that might have made failure rates higher for one SF based drive than another. If you separate out all the issues the picture will become a lot clearer.

 

[Mfusick]

I just think it still funny how this thread degenerated into a OCZ bash.. when it is supposed to be about Sandforce.

 

[josephjpeters]

I just think it still funny how this thread degenerated into a OCZ bash.. when it is supposed to be about Sandforce.

haha, I'll ask again.

"Back on topic: anyone have any clue what the new Sandforce controllers are going to look like next year? I remember hearing a Q2 release next year for their 3000 series. I wonder how it'll compare with OCZ's Barefoot 3 and Samsung's MDX. Will it be able to deal with TLC (something the 2200 series supposedly can't handle)."

This is actually something that's worth discussion, in my own opinion.

 

[Coup27]

I just think it still funny how this thread degenerated into a OCZ bash.. when it is supposed to be about Sandforce.

I hope it's ironic humour that you are experiencing considering the very first comments which began to derail this thread into other discussions were actually posted by you:

I forgot to mention it's popular and "trendy" to hate on Sandforce and OCZ for some beginner troubles a while back... But technology changes quickly and it's more a myth and legend at this point than a reality.

Even OCZ has moved most of it's line up to newer controllers but people still hate on them.

I think Sandforce takes some crap because OCZ was a big player originally in them- and the OCZ haters jumped on board with bashing Sanforce too.

Other users then began to counter the points you made which started the thread deviation.

 

[josephjpeters]

A quick and dirty test is to look at the percentage of below average reviews (1- or 2-eggs) of all models on newegg.

 2.3% = 7/308 Intel 330 (Sandforce)
 6.4% = 91/1411 Samsung 830
 7.0% = 15/214 Samsung 840 + 840 Pro
 8.6% = 25/291 Intel 520 (Sandforce)
12.2% = 82/673 OCZ Vertex 4

It seems OCZ comes out on the bottom again. What a surprise.

By the way, those are NOT return rates. Just use them as a relative index for customer dissatisfaction.

Amazon reviews show Vertex with a ~7% 'dissatisfaction rating' and a quick look at a random Samsung 830 model pegs them at ~6%. Is that "statistically significant"?

The fact that you choose Newegg over Amazon to prove a point just because it fits your belief demonstrates your bias.

I'm done with this topic. You really ruin these discussions because this was never the OP.

 

[Coup27]

lol honestly.

jwilliams said right at the top that it was a "quick and dirty" test. It was by far no means a comprehensive study into the subject. The fact that you have found reviews from another site doesn't mean anything. Newegg is the most popular US e-tailer so I hardly think he was being biased because he chose to use their reviews.

These discussions are ruined before they start. You're either allowed to like OCZ or SF, or you're accused of being blind and biased - regardless what independent information is provided.

I also find it ironic [again] that you have singled him out for "ruining" this topic. This thread has 111 posts and they certainly are not all his and as I pointed out above, he did not start the de-rail either.

 

[josephjpeters]

lol honestly.

jwilliams said right at the top that it was a "quick and dirty" test. It was by far no means a comprehensive study into the subject. The fact that you have found reviews from another site doesn't mean anything. Newegg is the most popular US e-tailer so I hardly think he was being biased because he chose to use their reviews.

These discussions are ruined before they start. You're either allowed to like OCZ or SF, or you're accused of being blind and biased - regardless what independent information is provided.

I also find it ironic [again] that you have singled him out for "ruining" this topic. This thread has 111 posts and they certainly are not all his and as I pointed out above, he did not start the de-rail either.

I don't mean to single him out so I'm sorry (and I mean it sincerely), but there are other variables at play when you're talking about satisfaction. I'm not going to go further down this rabbit hole so I'll just leave it at that.

 

[Ao1]

Firstly it is not even a full year because in Anandtech's 520 review which is dated in February of this year Brian Klug replicated a BSOD scenario with the latest firmware at the time, which incidentally was also heralded as being finally bug free.

It was a year between the first drive coming out with a SF2xxx controller and the date that Intel released a SF based drive. Issues have remained well after a year however.

10/7/2010 - SandForce Announces Next-Gen SSDs, SF-2000 Capable of 500MB/s and 60K IOPS
1/5/2011 - OCZ'z Vertex 3 Pro Demo: World's First SandForce SF-2000
2/17/2011 - OCZ Vertex 3 Pro Preview: The First SF-2500 SSD
2/24/2011- OCZ Vertex 3 Preview: Faster and Cheaper than the Vertex 3 Pro
4/6/2011 - The OCZ Vertex 3 Review (120GB)
10/17/2011 - SandForce Identifies Firmware Bug Causing BSOD Issue, Fix Available Today
10/26/2011 - LSI Announces Agreement to Acquire SandForce
11/2/2011 - The LSI SandForce Acquisition: Anand's Thoughts
2/6/2012 - Intel SSD 520 Review: Cherryville Brings Reliability to SandForce
Feb 2012 - Brian Klug replicated a BSOD scenario
6/11/2012 Intel Discovers SandForce SF-2281 Controller Can't Do AES-256 Encryption, Offers Return Program
11/22/2012 SandForce TRIM Issue & Corsair Force Series GS (240GB) Review

 

[JellyRoll]

I think it bears mention that through a good period of time there TRIM was not working.

Also, note that the BSOD problem went away before LSI and Intel were involved. Many speculate that LSI actually fixed it prior to the announcement of the purchase to clean up the SF image. It could have also been Intel, they were deep in validation at that point.
The BSOD was fixed days before the LSI acquisition, so it doesn't take a rocket surgeon to figure that one out

 

[jwilliams4200]

Amazon reviews show Vertex with a ~7% 'dissatisfaction rating' and a quick look at a random Samsung 830 model pegs them at ~6%. Is that "statistically significant"?

The fact that you choose Newegg over Amazon to prove a point just because it fits your belief demonstrates your bias.

I'm done with this topic. You really ruin these discussions because this was never the OP.

No need to get upset. You specifically wrote that we should look at reviews about the Vertex 4. I was responding to your request.

As for newegg vs. amazon, I used newegg because they usually have the largest collection of customer SSD reviews and I have used them before. But since you asked so nicely, I just applied the same quick and dirty method to amazon SSD reviews (I have not spent much time going through amazon reviews before, so I could have missed something, but I included my counts below and corrections are welcome):

 3.9% = 3/76 Samsung 840 + 840 Pro
 5.6% = 39/702 Samsung 830
 7.3% = 16/220 Intel 330 + 335 + 520
 9.5% = 39/411 OCZ Vertex 4

As you can see, there are not as many reviews on amazon as there are on newegg. But if we take the previous numbers from newegg (I merged the Intel models together and added the 335)

 5.2% = 32/620 Intel 330 + 335 + 520 (Sandforce)
 6.4% = 91/1411 Samsung 830
 7.0% = 15/214 Samsung 840 + 840 Pro
12.2% = 82/673 OCZ Vertex 4

And combine them with the amazon numbers, we have:

 5.7% (4.1, 7.3) = 48/840 Intel 330 + 335 + 520
 6.2% (5.1, 7.2) = 130/2113 Samsung 830
 6.2% (3.4, 9.0) = 18/290 Samsung 840 + 840 Pro
11.2% (9.2, 13.1)= 121/1084 OCZ Vertex 4

I've included the 95% confidence interval in parentheses.

I did not spend much time collecting and analyzing this data -- as I said before, this is quick and dirty. Someone with more time might go through the reviews and throw out any that are obviously not about the product (eg., review of the retailer rather than the product, or reviews of a different product). Perhaps add other refinements. I'm not interested in spending more time on it...I'll just wait for the hardware.fr return-rate numbers to come out next year.

 

[josephjpeters]

No need to get upset. You specifically wrote that we should look at reviews about the Vertex 4. I was responding to your request.

As for newegg vs. amazon, I used newegg because they usually have the largest collection of customer SSD reviews and I have used them before. But since you asked so nicely, I just applied the same quick and dirty method to amazon SSD reviews (I have not spent much time going through amazon reviews before, so I could have missed something, but I included my counts below and corrections are welcome):

 3.9% = 3/76 Samsung 840 + 840 Pro
 5.6% = 39/702 Samsung 830
 7.3% = 16/220 Intel 330 + 335 + 520
 9.5% = 39/411 OCZ Vertex 4

As you can see, there are not as many reviews on amazon as there are on newegg. But if we take the previous numbers from newegg (I merged the Intel models together and added the 335)

 5.2% = 32/620 Intel 330 + 335 + 520 (Sandforce)
 6.4% = 91/1411 Samsung 830
 7.0% = 15/214 Samsung 840 + 840 Pro
12.2% = 82/673 OCZ Vertex 4

And combine them with the amazon numbers, we have:

 5.7% (4.1, 7.3) = 48/840 Intel 330 + 335 + 520
 6.2% (5.1, 7.2) = 130/2113 Samsung 830
 6.2% (3.4, 9.0) = 18/290 Samsung 840 + 840 Pro
11.2% (9.2, 13.1)= 121/1084 OCZ Vertex 4

I've included the 95% confidence interval in parentheses.

I did not spend much time collecting and analyzing this data -- as I said before, this is quick and dirty. Someone with more time might go through the reviews and throw out any that are obviously not about the product (eg., review of the retailer rather than the product, or reviews of a different product). Perhaps add other refinements. I'm not interested in spending more time on it...I'll just wait for the hardware.fr return-rate numbers to come out next year.

Ironically, the "top rated" SSD on Amazon right now is the Vertex 4. http://www.amazon.com/gp/top-rated/e...zg_tr_tab_t_tr

Not making this up. You should question Amazon about their algorithm for what "top rated" means. Maybe present them with all this data?

 

[jwilliams4200]

You should question Amazon about their algorithm for what "top rated" means. Maybe present them with all this data?

I'm not especially interested in the method they are using. I did note that the Samsung 830 and Crucial m4 both have an average rating of 4.7, while the Vertex 4 has an average rating of 4.6, and yet amazon has the V4 ranked above those others. Given that strange ranking, and that amazon gives such a vague description of their methodology, I question the value of that list.

But if you are interested, feel free to look into it and report your findings.

 

[Mfusick]

I'm not especially interested in the method they are using. I did note that the Samsung 830 and Crucial m4 both have an average rating of 4.7, while the Vertex 4 has an average rating of 4.6, and yet amazon has the V4 ranked above those others. Given that strange ranking, and that amazon gives such a vague description of their methodology, I question the value of that list.

But if you are interested, feel free to look into it and report your findings.

You only question the value of that list because it does not represent what you want to see.

If the list said OCZ was the worst you would be standing on your soap box screaming about it's validity.

 

[Coup27]

You only question the value of that list because it does not represent what you want to see.

No he is questioning the list because either something is wrong or the list is bent. If product A has an average rating of 4.7 and product B has an average rating of 4.6, how can Amazon rank product B above product A?

 

[groberts101]

You only question the value of that list because it does not represent what you want to see.

If the list said OCZ was the worst you would be standing on your soap box screaming about it's validity.

Nope.. highly doubtful. Because he'll more than likely be screaming from the saddle of that.. big "damn OCZ" horse.. when he shows back up.

and .01 variance between the 2 sources??? roflmao.. so much for statistical engineering degrees and variable allowance rules. :$

IMHO.. the fact that this thread is so popular and also has high viewcounts is proof that this site has been allowed to become a personalized dumping ground for peoples daily frustrations. Not that I haven't added my fair share of steaming piles on ocassion.. but.. when in Rome.

 

[jwilliams4200]

and .01 variance between the 2 sources???

What are you talking about? What variance, what sources?

 

[Mfusick]

No he is questioning the list because either something is wrong or the list is bent. If product A has an average rating of 4.7 and product B has an average rating of 4.6, how can Amazon rank product B above product A?

Your argueing semantics is my point.

 

[Coup27]

Your argueing semantics is my point.

That doesn't even make sense.

 

[josephjpeters]

No he is questioning the list because either something is wrong or the list is bent. If product A has an average rating of 4.7 and product B has an average rating of 4.6, how can Amazon rank product B above product A?

Depends on your time interval.

Vertex 4 had less than perfect reviews in May-June because it was running a different FW. Now it seems those issues are fixed and the reviews are the same if not better than the competition.

If you're buying a Vertex 4 today does it matter how reliable the drive was with FW 1.4 if all drives are shipping with 1.5 which has proven to be really great?

Ask Crucial M4 owners about that 6 month bug. Now that that problem has been fixed should that previous FW bug influence your decision to purchase that product TODAY? No, it shouldn't.

The current state of reliability for OCZ is much better than it ever was and it should not be a surprise that Sandforce-based drives aren't being sold by them in quantity (trying to tie it back to the OP). The non-exact science of looking at the most recent reviews seem to support this.