What's wrong with sandforce?

[Hulk]

What I'm curious about is why Intel switched to Sandforce after developing their own SSD controller? It seems as though Intel must have taken a look at the cost of developing their own controller with more up-to-date performance than what they currently had on the shelf versus the cost of licensing the Sandforce controller. I'm sure they went through some brief testing of the SF controller to see how good or bad it was and decided it was good for the money. Once they signed on I would assume it went through the normal Intel validation test, which like Intel or not, is a pretty rigorous process as they do have a good record for reliability and stability.

This brings to my mind at least two interesting points. First, Intel has deep pockets and some of the best engineering and development in the business. Designing and producing a good SSD controller must be a pretty enormous project. And second, I don't think it is unreasonable that Intel was able to make some small proprietary changes to the SF controller either through firmware, source via SF, or some other as yet unknown means. We see many vendors using the "same" SF controller with varied performance and reliability records.

 

[groberts101]

IMO, aside from the needless posturing and proverbial lines being drawn in the sand.. there are some really valid overviews and points being made here.

Since I test these things for a company.. which some distrust and willingly go out of their way to discredit.. along with myself included.. any chance they get.. I will just make a few quick blanket statements and leave you guys to get back to it again.

Without giving too much info out and shooting myself in the "NDA foot" just for the sake of sharing facts/opinions or stroking my own ego just to prove how smart me is.. I will just say this.

IF.. Intel.. or OCZ.. or any other company for that matter.. could have fixed all issues related to Sandforce on their own?.. they would have done it long ago. Sadly.. the pieces(OROM's, drivers, power mgmt features in both the firmware.. AND the specific OS.. AND the bios) required to do that took many months of testing, revamp, and retesting to make them all play nice together. And as much as Intel likes to take the credit for doing it all by themselves through their "intense validation process".. it would not have been possible to do it in only 1 year if the other early adopters hadn't paved the way with all those stubbed toes and stock losses along the path to get them there.

So, in a nutshell.. Intel COULDN'T..not "shouldn't".. have released their drive's using this particular controller simply because all the pieces needed were not even close to being ready when the newer platforms bios, OS power mgmt default features, and latest sata3 spec's adherance were hot off the assembly lines. Then add a new largely invalidated "super smart" SF controller to that mix and it was inevitable to see many of the early issues some of us saw. Due to that above mentioned immaturity.. I also remember seeing and testing plenty of other power mgmt related issues with other non-SF controlled SSD's at the time too. For which SF and some other controller mfgrs were very grateful to find workarounds for as they helped flush out some of the underlying issues causing those growing pains.

 

[Ao1]

There was also a problem with AES-256 support that Intel missed
Although SandForce's SF-2281 controller has been shipping for well over a year at this point, it took Intel to discover a bug in the controller that prevents it from properly supporting AES-256 encryption. The bug is at the controller level and can't be fixed with a firmware update. AES-128 encryption works perfectly fine as does the drive's standard, un-encrypted operation mode.

http://www.anandtech.com/show/5971/intel-discovers-sandforce-sf2281-controller-cant-do-aes256-encryption-offers-return-program

Maybe minor but Intel introduced at least one unique firmware change as the SMART data set is different to the SF SMART data set.

 

[BFG10K]

unless of course ones own personal perspective includes having owned dozens of them without issue through the years. Then those "stats" are only useful for those who are looking to find ammo to use against them.

Jwilliams has addressed this point nicely. Not to mention that your sample size is rather miniscule compared to their data.

OCZ has clear issues compared to the other vendors, by a magnitude. It's not even close.

PS. I would think it safe to assume that an Anandtech mod's job would be to keep every single "Sandforce" thread from turning back into yet another OCZ bashfest.. but it appears that would be assuming too much. Pretty sad as it just brings the whole place's credibility down a notch or two.

I have just as much right to participate in this thread as you do.

 

[BFG10K]

And how many of those were based off the SF-2281 that caused so much trouble for early adopters? One. The topic at hand is about Sandforce, not OCZ. Your information is moot in this thread.

In case you're unfamiliar with their past tracking, here are some more figures for you: http://www.behardware.com/articles/843-7/components-returns-rates-5.html

9.14% OCZ Vertex 2 240 GB
8.61% OCZ Agility 2 120 GB
7.27% OCZ Agility 2 40 GB
6.20% OCZ Agility 2 60 GB
5.83% Corsair Force 80 GB
5.31% OCZ Agility 2 90 GB
5.31% OCZ Vertex 2 100 GB
5.04% OCZ Agility 2 3.5" 120 GB

Vertex 2 and Agility 2 drives are SF based.

Again, what I quoted earlier stated that OCZ and SandForce are somehow being treated unfairly, so I showed OCZ still has high return rates even today. Their SF drives also had high return rates back in the day, as shown above.

You're also not looking at the overall volume of sales.

Not really relevant when dealing with % returns. If 5 out of 10 drives fail, that's worse than 25 out of 100 failing. Even if OCZ sold more units, it still comes back to % figures.

Now, to get back on topic, all of the major firmware issues related to the SF-2281 controller have been resolved.

This has already been debunked repeatedly. In addition, Anand's 520 review: http://www.anandtech.com/show/5508/intel-ssd-520-review-cherryville-brings-reliability-to-sandforce

As luck would have it, our own Brian Klug happened to come across an unexpected crash with his 240GB non-Intel SF-2281 based SSD two weeks ago when he migrated it to another machine. The crash was an F4 BSOD, similar in nature to the infamous BSOD issue from last year. While two of the systems we reproduced the BSOD bug on were cured by last year's firmware update, Brian's system (an X58/Core i7 build) was BSODing regularly playing Battlefield 3.

 

[ShintaiDK]

What I'm curious about is why Intel switched to Sandforce after developing their own SSD controller? It seems as though Intel must have taken a look at the cost of developing their own controller with more up-to-date performance than what they currently had on the shelf versus the cost of licensing the Sandforce controller. I'm sure they went through some brief testing of the SF controller to see how good or bad it was and decided it was good for the money. Once they signed on I would assume it went through the normal Intel validation test, which like Intel or not, is a pretty rigorous process as they do have a good record for reliability and stability.

This brings to my mind at least two interesting points. First, Intel has deep pockets and some of the best engineering and development in the business. Designing and producing a good SSD controller must be a pretty enormous project. And second, I don't think it is unreasonable that Intel was able to make some small proprietary changes to the SF controller either through firmware, source via SF, or some other as yet unknown means. We see many vendors using the "same" SF controller with varied performance and reliability records.

You should ask yourself why Intel starting inhouse controller development again after it was shut down. New enterprise drives already got this new Intel controller.

Intel made the big mistake to assume the market would produce cheap reliable controllers and any issues could just be fixed in software. They payed the price twice, RMA rates and PR hit. Now they are back to fix the issue with their own again.

Intel employees are also very unhappy with SF based drives in their own centers. And it simply took time to go up in the management chain.

 

[bononos]

What I'm curious about is why Intel switched to Sandforce after developing their own SSD controller? It seems as though Intel must have taken a look at the cost of developing their own controller with more up-to-date performance than what they currently had on the shelf versus the cost of licensing the Sandforce controller. I'm sure they went through some brief testing of the SF controller to see how good or bad it was and decided it was good for the money. Once they signed on I would assume it went through the normal Intel validation test, which like Intel or not, is a pretty rigorous process as they do have a good record for reliability and stability.

This brings to my mind at least two interesting points. First, Intel has deep pockets and some of the best engineering and development in the business. Designing and producing a good SSD controller must be a pretty enormous project. And second, I don't think it is unreasonable that Intel was able to make some small proprietary changes to the SF controller either through firmware, source via SF, or some other as yet unknown means. We see many vendors using the "same" SF controller with varied performance and reliability records.

For the reasons in the post above mine and I think at that time, Intel was lagging behind in performance. Intel does have good track record but this sandforce thing was a fiasco for them. Heck even OCZ wants to rub away the sandforce reputation by going with Marvell.

 

[Hulk]

I know there is quite a bit of "siding up" in this thread. That wasn't the point of my post. I was simply looking into Intel's rationale for going with SF. As they usually don't make bad business decisions. Yeah I know Netburst was not a good decision. That's why I wrote usually.

I didn't know that the Intel 330's/335's have been so much of a problem for Intel. Just how high is the failure rate on those drives? And more importantly why would they still be selling them? Shouldn't they pull them if it's such a fiasco?

 

[ShintaiDK]

I didn't know that the Intel 330's/335's have been so much of a problem for Intel. Just how high is the failure rate on those drives? And more importantly why would they still be selling them? Shouldn't they pull them if it's such a fiasco?

Its a business. If it still makes money it sells. Else OCZ should have stopped selling anything long ago. And the RMA rate aint high enough to cancel it either. But its far from optimal compared to before. The AES256 issue wasnt good for PR either:
http://www.anandtech.com/show/5971/...nt-do-aes256-encryption-offers-return-program
http://www.xbitlabs.com/news/storag..._Bit_Encryption_by_SandForce_Controllers.html

Also it takes time to get an alternative solution back up. It makes no sense to withdraw from the market in the meanwhile.

And Intel completely rebranding its SSD line into a new series shows it dont want the new tainted.

 

[Hulk]

Its a business. If it still makes money it sells. Else OCZ should have stopped selling anything long ago. And the RMA rate aint high enough to cancel it either. But its far from optimal compared to before. The AES256 issue wasnt good for PR either:
http://www.anandtech.com/show/5971/...nt-do-aes256-encryption-offers-return-program
http://www.xbitlabs.com/news/storag..._Bit_Encryption_by_SandForce_Controllers.html

Also it takes time to get an alternative solution back up. It makes no sense to withdraw from the market in the meanwhile.

And Intel completely rebranding its SSD line into a new series shows it dont want the new tainted.

Besides the AES256 issue what other problems are killing all of the Intel SF drives?

 

[ShintaiDK]

Besides the AES256 issue what other problems are killing all of the Intel SF drives?

BSOD, bricking, powerstate failures etc. The usual SF issues.

 

[WhoBeDaPlaya]

BSOD, bricking, powerstate failures etc. The usual SF issues.

Honestly, I wasn't impressed by Intel's non-SF drive either. Both 80GB X-25M G1s I had died eventual deaths - they'd slow to a crawl (even secure erase didn't fix it) and all of a sudden became unreadable.

Oddly enough, I've been happier with the SF drives. Admittedly, not being an early adopter helped (I got the Vertex 2s pretty late in their life) and IINM, the Vertex 3s are the ones more afflicted with problems (I only have Agility 3s and SanDisk Extremes).

 

[Hulk]

BSOD, bricking, powerstate failures etc. The usual SF issues.

Hmm. This thread seems to indicate little is know about actual 330 failures. And a quick internet search doesn't turn up much either.

 

[groberts101]

Hmm. This thread seems to indicate little is know about actual 330 failures. And a quick internet search doesn't turn up much either.

you just need to use the correct search words to find them is all. Try "Intel 330 freezing".. and "Intel 330 Macbook freezing".

As per the typical SF rules of engagement.. there are workarounds and other factors at play here of course so we can't just blame Intel as whole. As I've already mentioned several times now.. power mgmt in both bios and OS along with drivers can have huge impact on those controllers.. regardless of who uses.. or validates them.

For that matter.. I'd LOVE to see some folks(and I still do on ocassion) with older hardware who still run outdated bios and non-updated OS's to prove the point that all is not what it seems as far as Intel's "extensive validation process" goes. In fact, many of the early growing pains with those particular controllers were caused by Intel themselves.. and IMHO.. they purposely dragged their feet on releasing those FINAL power mgmt related updates until shortly before they released their drives(there were large numbers of ASPM related updates along the way too). Which is why we used to troubleshoot some of the supposed "mfgr specific issues" by simply installing Linux on non-Intel based sloutions and checking drive integrity. Problems were very often miraculously solved and that spoke volumes to some of the underlying issues. Then there's the very fact that even those supposed "mfgr specific issues" suddenly became fixed through simple driver updates alone.. and WITHOUT a firmware patch even being needed. Coincidence? umm?.. I think not.

 

[zlejedi]

I wonder why would I want to buy Sandforce based drive when there are comparable products like Crucial M4 or Samsung 830 (840 pro needs to prove itself a bit after firmware problems) at similar prices.

 

[Mfusick]

Well the M4 is a slow old pig so I'd take a sandforce SSD like a MAX IOPS with toggle Toshiba NAND over that for same price every time.

The SAMSUNG is tougher choice because those drives are really nice.

 

[Hulk]

I wonder why would I want to buy Sandforce based drive when there are comparable products like Crucial M4 or Samsung 830 (840 pro needs to prove itself a bit after firmware problems) at similar prices.

Everybody has their own buying criteria.
I bought the 330 for the following reasons.
1. Positive Anandtech review, especially I would be using it as a boot drive and not for incompressible data.
2. Price. I got the 240GB version for $126 shipped. And that was a few months ago.
3. Intel reliability.
4. I'll be moving it to my daughter's PC once I upgrade my main rig.

 

[Zxian]

In case you're unfamiliar with their past tracking, here are some more figures for you: http://www.behardware.com/articles/843-7/components-returns-rates-5.html

Vertex 2 and Agility 2 drives are SF based.

Again, what I quoted earlier stated that OCZ and SandForce are somehow being treated unfairly, so I showed OCZ still has high return rates even today. Their SF drives also had high return rates back in the day, as shown above.

Yes, the SF-1200 controller. Can you/would you buy a SF-1200 based drive today? No.

This has already been debunked repeatedly. In addition, Anand's 520 review: http://www.anandtech.com/show/5508/intel-ssd-520-review-cherryville-brings-reliability-to-sandforce

And you can find several other scenarios where the M4 and the 830 also have incompatibilities with certain hardware.

The point that I'm trying to get at is that today, Dec 2012, SF-2281 are not as bad as internet history would have you believe. Are they my top pick for SSDs? Probably not, but they're not "the devil" and will probably still work just fine for the OP.

The fact that this thread has spun quite far from the original question of "What's wrong with sandforce?" is a little sad. So many people pointing the finger at OCZ in particular and not directly answering the OP.

 

[jwilliams4200]

The point that I'm trying to get at is that today, Dec 2012, SF-2281 are not as bad as internet history would have you believe.

How do you know? Do you have some statistics covering SF-2281 SSD problem rates for the past 3 or 6 months?

 

[blastingcap]

I have seen with my own eyes shills for OCZ on amazon.com reviews, and judging by the shrillness of support for OCZ and vague disparagement (as well as ludicrous disparagement of the M4, which has quite acceptable random 4k speeds, which are what really matter, not some stupid compressible sequential speeds) here I suspect there are less-than-unbiased people here as well. Also, at least one person in this thread is an OCZ SSD tester and has openly stated as much.

 

[BFG10K]

Well the M4 is a slow old pig so I'd take a sandforce SSD like a MAX IOPS with toggle Toshiba NAND over that for same price every time.

Slow old pig? I think not. Read performance is still extremely competitive with newer drives:

Write performance lags a lot, but for practical purposes you won't see a difference between it and newer SSDs unless you copy files to it all day.

 

[Mfusick]

Slow old pig? I think not. Read performance is still extremely competitive with newer drives:

Write performance lags a lot, but for practical purposes you won't see a difference between it and newer SSDs unless you copy files to it all day.

Ok. I'm going to try to remove your OCZ bias and hatred for sake of making my point. Lets assume OCZ doesn't exist anymore.

How can you reccomend an older Crucial that is generally outpaced by a Samsung and sells for same prices ?

I read everyday here at Anand and other sites about the 5000 hour blue screen of death in Crucial drives. The answer is always " firmware" but its spun in a most positive light suggesting its totally normal and acceptable for an SSD to experience BSOD errors and updating the firmware is no big deal. Reliability is in reality not any better or worse than other SSD drives and it loses out on other areas.

Just because the loyal and most vocal crucial supporters promote that drive does not make it the best choice. It's almost never the cheapest SSD, the fastest SSD, or the most reliable SSD. It's overrated and aged today. It's not 2011 anymore.

I really want to point out how the Vertex 4 uses the same controller as the M4 and should have as good reliability. Perhaps better since it doesn't have BSOD issues effecting 100% of drives that don't have firmware updated. Vertex4 has longer 5 year warranty, sells for less $, and is certainly higher performance. It's been out long enough to realize its an acceptable option so don't bury it under sandforce based examples because its unreasonable to assume it is in the same boat when it doesn't have sandforce controller.

Part of me regrets pointing out the Vertex 4 because I don't want this argument to turn into OCZ vs Crucial.

This argument is more Crucial = overrated.

Even without existence if OCZ most other good SSD drives are better choices over Crucial. I'm not sure you could convince a reasonable unbiased person a solid reason to choose an older slower Crucial over a Sansung or any other faster Marvel based SSD. A Toogle sandforce might be a better option too but I know you hate sandforce so I'll keep it to non sandforce drives if I must.

Old. Slow. Pig.

Most overrated in today's modern SSD landscape. SOS for crucial .. You drive needs updating to remain competitive with newest crop of SSDs.

 

[Mfusick]

And if coarse I could post repeatedly a myriad of benchmarks showing the newer drives outperforming the Crucial.

I just don't think that small real work difference is as important as the price/reliability/ performance ratio is overall.

Today there a very loud minority that owns Crucial drives and promotes the crap out of them in support of their own purchase decision.

There's nothing wrong at all with that. I just want to point out is a reality. It adds to a bit if overrated reputation on crucial reliability. Issues with Crucial get over looked and brushed away under carpet as no big deal.

In comparison OCZ is crucified for much smaller issues. OCZ firmware update and only a small percent if drives with BSOD is made out to be a huge deal in comparison.

I think it's a reasonable argument to point out as much as there is a negative trendy bias against OCZ there is a trendy over rated positive bias on Crucial.

 

[Ao1]

Slow old pig? I think not. Read performance is still extremely competitive with newer drives:

Write performance lags a lot, but for practical purposes you won't see a difference between it and newer SSDs unless you copy files to it all day.

Plus One. The M4 is one of the fastest SSD's I have for "real world" read performance.

Write speeds let it down, but as you imply that is a mostly useless metric for desktop users.

 

[Mfusick]

And how many do you have ?

Because I have one. I also have 17 others. My experience is different than yours.

Vertex 4 and 3 and also Samsung 830 " feel" faster. My Toggle NAND Sandforce also feels faster. I also have dual sandforce in RAID0. My next drive will be an 840 pro or vector.

The crucial "feels" faster than only the crappy agility 3 I have in my 20TB WHS Flexraid server.

Yes I've benched them.