SSD Reliability

Silencing hard drives, optical drives and other storage devices

Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee

dhanson865
Posts: 2198
Joined: Thu Feb 10, 2005 11:20 am
Location: TN, USA

Post by dhanson865 » Tue May 18, 2010 6:23 am

there is a reason I used the less than symbol. <2% is not the same as 2%.

and for the record even AMD fanboys trust Intel SSDs. I know I am/do.

madman2003
Posts: 182
Joined: Sat Jan 29, 2005 9:02 am
Location: Planet earth

Post by madman2003 » Tue May 18, 2010 6:29 am

I have an intel SSD in an AMD system, because it seemed like the best choice (price/performance/reliability). I'm not an intel hater when it comes to their cpu, it's just that the price/performance/energy consumption ratio is better for AMD often (So i'm not the worst of the fanboys).

dhanson865
Posts: 2198
Joined: Thu Feb 10, 2005 11:20 am
Location: TN, USA

Post by dhanson865 » Tue May 18, 2010 7:38 am

Cryoburner wrote:
dhanson865 wrote:See the near 50% failure rate for the Patriot Torqx drives or the vast complaints about OCZ drives
50% of reviews on Newegg mentioning a drive failing does not equal a 50% failure rate. People are far more likely to write a review if a product dies early than if everything goes smoothly. I'm sure the failure rates of these drives are far below what user reviews would indicate. That said, it is likely to indicate a higher failure rate relative to other SSDs, so I probably would avoid drives with above average reported failures.

Along the same lines, these specific drives with high numbers of reported problems shouldn't necessarily indicate a lack of confidence in SSDs in general. Just look at user reviews in the hard disk section at Newegg. Many disk-based drives, built on supposedly proven technology, have 20 to 40% of their reviews at 1 star. Most of these drives with higher reported failures are from Seagate, which is a brand a lot of people trust anyway.
If you go into the reviews for the patriot SSD in question (I'll put a URL up later when I get a chance) you'll see that while about half of the reviews mention a failed drive many of those reviews with failed drives mention more than one SSD failing. The reviews with positive experiences mostly apply to a single SSD. While it is true there are good reviews mentioning multiple SSDs and there are bad reviews mentioning a single SSD a solid count would show more failed SSDs per review. It's impressive to see a review where the 2nd replacement drive failed or where someone bought 3 drives and 3 of them failed. In those cases it implies higher than 50% failure rates.

I have no inside data and thus as you say I can't rule out the inaccuracy of a voluntary poll. It could be inaccurate by insanely high margins. It is however the worst of every single drive I looked at on newegg and I literally read thousands of reviews in preparation for the posts I've made in this thread.

http://www.newegg.com/Product/ProductRe ... 6820220389

It is worth noting as the worst SSD on newegg (in my own words) it still gets 3 eggs on a 1 to 5 scale.

reviews like these excerpts make you wonder what went wrong:
...I have had two drives fail since I bought one from Newegg in 11/2009. The first failed during the Newegg replacement period, and the replacement just recently failed. Now waiting on an RMA number from Patriot so I can get a second replacement...

...We purchased two of these and have RMA'd both of them. One got corrupted system files within two weeks, but the other one worked for about 6 months and then started getting errors...

...This is my second drive to die since purchasing. First drive died within a week, this next drive lasted a few month but has also died. On top of all that, they don't have advanced RMA...

...Installed two as a RAID pair 1 in a server for a client company. First drive went dead after 2 weeks! It also corrupted the 2nd drive on its way down. Newegg RMAed and replaced it quickly (thank you), and I thought it was a fluke, things happen. However, I did NOT enjoy totally rebuilding a client server a second time. Thinking I was done, but another drive died exactly 10 days later...

...Just had the 4th unit fail! We bought three originally; all three originally purchased at the end of August have failed; one replacement sent by Patriot failed too...

...Terrible reliability. We purchased (3) of these drives from Newegg.com.

First one failed within two weeks; 2nd one failed @90 days out. So User #2 failed drive gets replaced with the replacement drive received from Patriot from User #1 (who didn't get around to installing yet). It fails last week.

Replacement drives and customer service take way too long.

What good is 10 year warranty if you have to use it so much to begin with?! ...
You might notice on a previous page where OCZ recalled/stop sales of the ONYX ssd. That is the proper response. Somehow Patriot dropped the ball on this one. I can't say why because I have no idea but this gives me no confidence in Patriot.

I'm not saying OCZ is perfect (look back at the comments about the Vertex/Agility complaints) and at this point there are only 6 reviews for the OCZ ONYX ssd but if they pulled those drives back and fixed them BEFORE end users lost time/data then that is a good thing.

Newegg is only one retailer. How many Patriot Torqx SSDs were sold worldwide that failed like that. How many months did they continue to sell them?

frenchie
Friend of SPCR
Posts: 1346
Joined: Tue Jul 08, 2008 4:53 am
Location: CT

Post by frenchie » Tue May 18, 2010 8:33 am

Amazon has the same kind of reviews (though much fewer) :
http://www.amazon.com/Patriot-Torqx-Sol ... 101&sr=8-1

This drive seems to have a few issues...

dhanson865
Posts: 2198
Joined: Thu Feb 10, 2005 11:20 am
Location: TN, USA

Post by dhanson865 » Sun Jun 06, 2010 1:50 pm

June Newegg data dive.

http://www.newegg.com/Store/SubCategory ... 6&name=SSD and sort by most ratings will give you a quick look into the most reliable data though all of this is somewhat suspect.

I'm updating anything I notice but I'm paying special attention to drives that had low review numbers or low number of failures (new drives added, or drives that didn't sell much) as past checks haven't shown much changes in the major players. Let me know if I missed updating one and my numbers are out of line with reality.

quick summary:

Intel:
All Intel SSDs had low failure rates (maybe <5%?) easily the most reliable SSD from everything I've read though failures are higher now than the ~2% I saw earlier in their life cycle.

Indilinx: Note the M225 and Nova complaints are showing up now
Crucial M225 (Indilinx Barefoot) about 15% failure rate.

Corsair Nova (Indilinx Barefoot) about 7% failure rate.

OCZ Vertex and Agilility (Indilinx Barefoot) 30-60GB drives (chock full of so many negative reviews I find it hard to categorize them all).

OCZ Agility 120GB retail (Indilinx Barefoot) about 10% failure rate.

Patriot Torqx (Indilinx Barefoot) dead drives approaching the 50% mark.

OCZ Onyx (Indilinx Amigos). (This drive was taken off the market due to a firmware bug, makes it hard to get valid numbers for a post fix rerelease, besides the Amigos drives aren't priced appropriately to their performance so this drive has a double strike against it)

Sandforce (radical new controllers):
OCZ Vertex LE (sandforce 1200) about 30% failure rate.

OCZ Agility 2 (sandforce 1200) no failures reported yet.

OCZ Vertex 2 (sandforce 1200) about 10% failure rate.

Mushkin Enhanced Io (sandforce 1500) no failures reported yet.

Corsair Force (sandforce 1200) drive not found on newegg.

Sandforce drives have the most unconventional controller. It's unproven at this point and should only be bought by users with a strong understanding of the trade-offs this design uses. In some usage cases it's faster than the Marvell or Intel based drives in others it isn't. Long term reliability isn't proven yet.

Marvell:
Crucial C300 low failure rate but a big brouhaha over a firmware update procedure that could leave the drive unusable. New firmware may have fixed these issues but it's hard to tell from newegg ratings how this drive truly compares on reliability.

Plextor PX-xxM1S (Marvell Da-Vinci) no failures reported yet.


JMicron/Toshiba (avoid these drives):
WD SiliconEdge Blue no failures reported yet but this drive is so overpriced I'm not expecting it to get reliable data for a long time to come. I'll just quote Anand on that
I've repeated it enough times that you should get the point by now - Western Digital's SiliconEdge Blue is just a bit behind the performance of an Indilinx based SSD. ...

The SiliconEdge Blue just isn't worth it. ... It has to be cheaper than both Intel and Indilinx drives, which means cutting the MSRP in half.
Corsair Reactor (chock full of so many negative reviews I find it hard to categorize them all)


OK, I don't have the heart to dig through all the Jmicron drives and list them here. Nor do I want to dig up all the Samsung controller drives.

Note the Samsung controller is generally considered safe/reliable but doesn't offer the performance to make it worth tracking for an enthusiast.
Last edited by dhanson865 on Sun Jun 13, 2010 12:59 pm, edited 1 time in total.

dhanson865
Posts: 2198
Joined: Thu Feb 10, 2005 11:20 am
Location: TN, USA

Post by dhanson865 » Wed Jun 09, 2010 5:39 am

http://intelstudios.edgesuite.net/idf/2 ... S003/f.htm has a 30+ minute presentation. Only the first 15 minutes is worth listening to but it has solid data and concepts pertinent to reliability of consumer grade SSDs.

Reducing the partition size can increase the life of the SSD by over 3x as in over 300% or over 200% increase depending on how you like to think about the end result or the increase.

Unfortunately this is only true if you have never written to the drive's full capacity. A secure erase is the only way to return the drive to factory fresh and allow a smaller partition to increase the drives lifespan.

Eunos
Friend of SPCR
Posts: 378
Joined: Mon Dec 12, 2005 3:29 am
Location: Melbourne, Australia

Post by Eunos » Wed Jun 09, 2010 6:14 pm

Will definitely check that one out when I have time.

Is flash memory still rated for only 10 years? If so, no amount of partioning would make any difference except to users with a heavy workload. I worked out mine would last 30 years or more as is, and that's probably being conservative...

dhanson865
Posts: 2198
Joined: Thu Feb 10, 2005 11:20 am
Location: TN, USA

Post by dhanson865 » Thu Jun 10, 2010 8:23 am

Only 10 years? :lol:

I honestly don't care it'll last so long under most conditions

About 7 mins in there is a graph with yellow and blue regions. They talk about how long the drive will last in terms of TB and PB (thats Terrabyte aka 1024 GB and Pettabyte aka 1024 TB).

The X25M is good for 15TB worst case scenario (370TB if the writes are purely sequential). Assuming the 40GB value drive and the 80GB drive have the same basic endurance metric as the 160GB they put in the graph the flash itself will never wear out.

I've had the SSD in my work PC as the boot drive for about 6 months and I've written 700GB. At that rate I'm good for 10 years worst case or some insane number I don't care about in the best case.

Of course all of that assumes that their assumptions are correct. The concept to me is that for performance and reliability reasons there are strong suggestions from Intel that if I leave a significant portion of my SSD unformatted and unused by the OS the drive will maintain performance levels longer and the flash won't wear out as quickly.

Given that the performance affects me even if the endurance is a non issue I'll err on the side of smaller partitions.

The reality to me is that I only care if the drive lasts maybe 6 years at most if I'm using it as a day to day drive. If I'm using it as a backup media my concern about longevity is different but then again the lack of constant writes to a backup disk laying in a drawer means I won't be worrying about flash wear there either.

dhanson865
Posts: 2198
Joined: Thu Feb 10, 2005 11:20 am
Location: TN, USA

Post by dhanson865 » Thu Jun 10, 2010 8:30 am

Now take a look at the failure rates in my Newegg data dive and compare them to the years and years the flash should survive in normal use.

Something other than the flash is dying. And whatever it is (manufacturing defects?) the problem is basically an issue on every brand and model of SSD (some more than others).

Eunos
Friend of SPCR
Posts: 378
Joined: Mon Dec 12, 2005 3:29 am
Location: Melbourne, Australia

Post by Eunos » Thu Jun 10, 2010 7:13 pm

That applies to platter drives too. I've only had one platter drive die on me ever, and it was the controller board that failed.

This is why I still think the smart money is on investing in a reasonably reliable SSD as per the data gathered here, then just using it like an HDD without worrying about flash lifespan. Of course, a backup strategy is needed in any case.

dhanson865
Posts: 2198
Joined: Thu Feb 10, 2005 11:20 am
Location: TN, USA

Post by dhanson865 » Wed Aug 11, 2010 8:54 am

OK, its been a while.

August Newegg data dive.

http://www.newegg.com/Store/SubCategory ... er=REVIEWS (sorted by most ratings) will give you a quick look into the most reliable data though all of this is somewhat suspect.

I'm updating anything I notice but I'm paying special attention to drives that had low review numbers or low number of failures (new drives added, or drives that didn't sell much) as past checks haven't shown much changes in the major players. Let me know if I missed updating one and my numbers are out of line with reality.

quick summary:

Intel:
All Intel SSDs had low failure rates (maybe <5%?) easily the most reliable SSD from everything I've read.


Sandforce (radical new controllers):
OCZ Vertex LE (sandforce 1200) about 25% failure rate.

OCZ Agility 2 (sandforce 1200) about 10% failure rate.

OCZ Vertex 2 (sandforce 1200) about 10% failure rate.

Mushkin Enhanced Callisto (sandforce 1200) <5% failure rate.

Corsair Force (sandforce 1200) about 15% failure rate.

Sandforce drives have the most unconventional controller. It's unproven at this point and should only be bought by users with a strong understanding of the trade-offs this design uses. In some usage cases it's faster than the Marvell or Intel based drives in others it isn't. Long term reliability isn't proven yet.

Note: Sandforce drives perform very poorly with intensive writes of compressed data. Video or Audio editing or use of encrypted data on a sandforce drive can put the drive into a state where it remains slow even after the trigger event has ended.


Marvell August update:
Crucial C300 no failures reported on Newegg since June. It appears the new firmware fixed the big issues.

Indilinx:
Crucial M225 (Indilinx Barefoot) about 10% failure rate.

Corsair Nova (Indilinx Barefoot) about 7% failure rate.

OCZ Vertex and Agilility (Indilinx Barefoot) 30-60GB drives (chock full of so many negative reviews I find it hard to categorize them all).

OCZ Agility 120GB retail (Indilinx Barefoot) about 10% failure rate.

Patriot Torqx (Indilinx Barefoot) dead drives approaching the 50% mark.

OCZ Onyx (Indilinx Amigos). (This drive was taken off the market due to a firmware bug, makes it hard to get valid numbers for a post fix rerelease, besides the Amigos drives aren't priced appropriately to their performance so this drive has a double strike against it)

JMicron/Toshiba (avoid these drives):
WD SiliconEdge Blue <5% failures reported but this drive is so overpriced I'm not expecting many to be bought. I'll just quote Anand on that
I've repeated it enough times that you should get the point by now - Western Digital's SiliconEdge Blue is just a bit behind the performance of an Indilinx based SSD. ...

The SiliconEdge Blue just isn't worth it. ... It has to be cheaper than both Intel and Indilinx drives, which means cutting the MSRP in half.
Corsair Reactor (chock full of so many negative reviews I find it hard to categorize them all)


OK, I don't have the heart to dig through all the Jmicron drives and list them here. Nor do I want to dig up all the Samsung controller drives.

Note the Samsung controller is generally considered safe/reliable but doesn't offer the performance to make it worth tracking for an enthusiast.

Metaluna
Posts: 133
Joined: Wed Mar 12, 2003 2:47 pm
Location: USA

Post by Metaluna » Wed Aug 11, 2010 10:46 am

dhanson865,

Have you looked at the ratings on the enterprise class drives at all? It might be interesting to compare the failure rates, since many of those drives use very similar (if not identical) controller chips and firmware, but SLC flash instead of the less expensive (but possibly less reliable) MLC. Might give some insight on whether the most common SSD failure modes are related to the flash memory, the firmware, or whether the Newegg reviews are distorted by self-selecting user bias.

dhanson865
Posts: 2198
Joined: Thu Feb 10, 2005 11:20 am
Location: TN, USA

Post by dhanson865 » Thu Aug 12, 2010 6:18 am

Metaluna wrote:dhanson865,

Have you looked at the ratings on the enterprise class drives at all? It might be interesting to compare the failure rates, since many of those drives use very similar (if not identical) controller chips and firmware, but SLC flash instead of the less expensive (but possibly less reliable) MLC. Might give some insight on whether the most common SSD failure modes are related to the flash memory, the firmware, or whether the Newegg reviews are distorted by self-selecting user bias.
The only SLC based drives with ratings on newegg are

Intel X-25E 32GB which is an outdated G1 drive not worth looking at nowdays. FWIW it has no failures mentioned out of ~50 reviews suggesting a near perfect failure rate. By contrast the G2 MLC drives have had about 1000 reviews. But the drive is no longer relevant. It's like discussing the failure rate of 73GB raptors from 2006.

Some Transcend drives

Some RiData drives

None of the SLC drives are current controllers or of a capacity that would interest anybody.

the PCIe based drives only have 1 or 2 reviews each so there isn't enough there to be statistically relevant.

OCZ Colossus has just over 10 reviews and 1 failure which implies a ~10% failure rate in line with other MLC drives but is still too little reviews for me to give it much weight.

Seriously I don't see any drives more enterprise worthy than the Intel G2 MLC drives on newegg. Sandforce tried to go enterprise worthy with the SF1500 controller but I'm not willing to put those drives in my desktops/servers yet.

There is a Seagate Pulsar that is supposed to be enterprise grade with SLC in 50GB, 100GB, 200GB capacities but I have no idea how it performs and the only retailer I see that has it has no reviews. Until a drive like that makes it to newegg we won't know about reliability and until it gets reviewed on Anandtech or such we won't know about performance.

Eunos
Friend of SPCR
Posts: 378
Joined: Mon Dec 12, 2005 3:29 am
Location: Melbourne, Australia

Post by Eunos » Thu Aug 12, 2010 7:06 pm

Looks like there's a new SLC contender, not that it has much of a market that I can think of.

http://www.anandtech.com/show/3849/micr ... enterprise

dhanson865
Posts: 2198
Joined: Thu Feb 10, 2005 11:20 am
Location: TN, USA

Post by dhanson865 » Sat Aug 14, 2010 8:25 am

It'll be outrageously expensive but the Micron P300 SSD will be the fastest SSD on the market this side of 4 figures (less than $1000).

There will be plenty of businesses and gamers that will pay the premium for buying literally the fastest consumer grade product available.

Eunos
Friend of SPCR
Posts: 378
Joined: Mon Dec 12, 2005 3:29 am
Location: Melbourne, Australia

Post by Eunos » Sat Aug 14, 2010 6:01 pm

I think OCZ's RevoDrive would be the fastest, but I know what you mean. 8)

dhanson865
Posts: 2198
Joined: Thu Feb 10, 2005 11:20 am
Location: TN, USA

Post by dhanson865 » Sun Aug 15, 2010 9:23 am

Eunos wrote:I think OCZ's RevoDrive would be the fastest, but I know what you mean. 8)
You have to put a couple of asterisks next to the Revodrive.

1. It has sandforce controllers so it doesn't handle randomized data at full speed (encryption, MP* files, zip/7z/rar files, etcetera)

2. It requires a full height PCIe x4 slot so it isn't usable in a laptop or in small form factor cases

3. again that PCIe x4 slot. On cheaper desktop boards the only free slot might be x1. On gamer boards you might have to choose between your 2nd video card and the revodrive. On silencer systems you might have to choose between a fan on your Accelero (overhanging a PCIe slot) and the revodrive. On Servers you might be choosing between the revodrive and PCIe x4 network cards.

Though yes if you gave me a revodrive free or sold it to me at $1/GB I'd be happy to take one. I literally don't have a system I could plug it into here at home but I could use it as an excuse to buy a new motherboard with more PCIe slots.

ascl
Posts: 279
Joined: Tue May 05, 2009 1:15 am
Location: Sydney, Australia

Post by ascl » Sun Aug 15, 2010 1:01 pm

A slightly OT update:

-My OCZ Vertex which I talked about earlier in the thread got RMA'd about 2 months ago... but the replacement has started corrupting data in the same fashion as the original! Sending it back for RMA cost me about $40 -- not sure I want to do it again, am very tired of this whole thing.

So does everyone still agree that the Intel drives are most reliable? Ideally I'd like to hold out till the Q4 release (150/300/600 GB drives!~), but not sure I'll last.

Eunos
Friend of SPCR
Posts: 378
Joined: Mon Dec 12, 2005 3:29 am
Location: Melbourne, Australia

Post by Eunos » Sun Aug 15, 2010 7:13 pm

dhanson865 wrote:You have to put a couple of asterisks next to the Revodrive.
I was expecting you to say that internal RAID is cheating. :lol: It certainly has a small market despite representing good value for hardcore gamer systems, but my main criticism is that the energy consumption is more on par with a high-capacity 3.5" drive - I expect much better from an SSD.
ascl wrote:-My OCZ Vertex which I talked about earlier in the thread got RMA'd about 2 months ago... but the replacement has started corrupting data in the same fashion as the original! Sending it back for RMA cost me about $40 -- not sure I want to do it again, am very tired of this whole thing.
I presume you have made enquiries with the good folk at the OCZ forum? I would only write off the drive if they cannot help.
ascl wrote:So does everyone still agree that the Intel drives are most reliable? Ideally I'd like to hold out till the Q4 release (150/300/600 GB drives!~), but not sure I'll last.
I've had only 100% perfect results with my admittedly tiny sample of 3 Intel drives, with dhanson's data more substantial yet drawing similar conclusions.

A couple of points to consider though. The 60 GB Vertex 2 currently on special at PC Case Gear is a completely different drive to the original Vertex and seems to be holding up very well in the reliability stakes. It is also much faster than the original Vertex or the G2 Intel drives with the exception of compressed data that probably only makes up a tiny percentage of what is on an OS drive. Despite the Intel's solid reputation, the Vertex 2 is my personal recommendation at the moment.

Also, the G2's excellent reputation for ease of use and reliability does not mean that the forthcoming Intel G3, based on a totally new manufacturing process, should be fully trusted when it first comes out. Intel has had its fair share of brouhahas in the past to get to where it is. This is another reason why Sandforce may be the SSD of choice for the next several months.

ascl
Posts: 279
Joined: Tue May 05, 2009 1:15 am
Location: Sydney, Australia

Post by ascl » Sun Aug 15, 2010 7:21 pm

Thanks for the response. The symptoms are exactly the same as last time, which required a RMA... and I tried every suggestion and spent a long time trying various things, so I really don't expect any difference this time round (aside from the total lack of interest in spending lots of time on it this time).

PCCG has some pretty cheap (for Oz) Sandforce drives, and you are correct, there is nothing to say that intel's g3 drives will be good... but without real stats we are just guessing that the sandforce drives have a reasonable failure rate (based on forum noise, they *have* to be better than the indilinx drives tho).

orlandomike
Posts: 3
Joined: Tue Aug 24, 2010 6:44 am
Location: florida

Post by orlandomike » Tue Aug 24, 2010 6:57 am

I'm trying to think back on what kind of computer I had 10 years ago or even 5 years ago. 2 things here, most of us will have a different computer in 3-5 years anyway and 2, the price of these things by that time will be in the HD range right now. $50 a piece. So all your talking about here isnt warrantee, it's how long will this 1 drive last until they get cheaper. Will it go 3 years.

In addition to the other comments, I can say that these things really excel in the random read and writes of small files. Temp files, OS operations etc. Not so much sequential large files. So, for me, I bought a laptop with a 320 gig HD, (i7) that also btw, had USB 3.0 ports. So I pulled the 320 gig 7200 rpm drive out, put it in a usb 3.0 enclosure and use that as my file storage drive. USB 3.0 is the speed of Sata 2, so if I am not just cruising the net, I can plug that in and manipulate files straight from that drive. Meanwhile the boot is the Crucial c300 128. plenty of room to breath as after the OS, adobe, office , utilities etc, it probably uses up 30 gigs and Windows 7 is solid on Trim.

The nice part of that is I can use the portability of that 320 gig to grab peoples video files, plg it into the laptop and set them up in Premiere and Adobe Flash right off of the drive and leave them there and of course you get the pop of opening Premiere at the blink of en eye.

Anyway, I had read somewhere that the AHCI was necessary or an important component to make Windows TRIM function best but I really dont know.

My desktop is a Hackintosh and since I mostly am using that for the sake of OSX and not windows dual booting. I have an Intel 80 gig for OSX and hope that without trim, it will continue to stay fast. I need to boot up in AHCP mode anyway for OSX. Then the HD I have houses Windows and the split partition allows me to store large files onto one side of that for OSX stuff since the nice folks at apple wont let you write to an NTFS partition. Nice going Steve Jobs.

the good thing about that kind of setup is once you load your OS and programs your not writing alot to the extra space in the SSD. You should maintian open blocks for a long time.

Soon Gen 3 will come out knocking the prices down and upping storage capacities from 160 gb on the intels to 320 and even 600. At that time, I might consider running it a different way.

orlandomike
Posts: 3
Joined: Tue Aug 24, 2010 6:44 am
Location: florida

Post by orlandomike » Tue Aug 24, 2010 7:18 am

btw, before i bought the c300, i called them and the guy explained all yuor complaints on bricking should be in may/june. That they did fix the issue and any new drives have the necessary updates. My desktop has a couple sata 3 ports on it which the c300 should really fly with but OSX doesnt take advantage of that so I am going to keep it in my laptop for awhile.

From what I have read, these benchmarks look impresive but when you get down to it, can you really see the difference of 225 mb/sec compared to 285 mb/sec when your moving a couple of megs at a time? Seems hardly noticeable.

dhanson865
Posts: 2198
Joined: Thu Feb 10, 2005 11:20 am
Location: TN, USA

Post by dhanson865 » Sat Sep 25, 2010 2:06 pm

Intel Developer Forum is a bi-annual event. The most recent was September 13-15, 2010 - San Francisco, USA.

SSDs have a heavy presence at IDF (or at least that is the only thing at IDF that catches my attention). Luckily this is one of the few times you get developers, testers, fab guys, etc in front of mics when they are apparently willing to divulge hard numbers.

The newegg data says Intel SSDs fail in the low single digits as a percentage of reviews. IDF quote below says (minus the umms and one or two repeated words):
Scott Doyle wrote:So with regard to the, the catastrophic failures, The data that we have today with, within Intel within the drives that we have within the field, the field failure rate is coming in below 1% AFR and that's combined with everything. It's really sitting somewhere around, between 0.5% to 0.7% AFR
Scott is a SR. Technical Account Manager at Intel. His stammering isn't that bad I just wanted to try and make the quote reasonably accurate in case anyone wants to compare vs the audio at the source.

AFR is http://en.wikipedia.org/wiki/Annualized_failure_rate and I'll pull this comparison out:

Google's 2007 study found that actual AFRs for individual drives ranged from 1.7% for first year drives to over 8.6% for three-year old drives

I have a few more hours of audio to listen to but the big push right now is to have a new JEDEC standard and have specs on a per drive basis on how many TB written over the lifetime of a drive would wear out the flash.

In short the industry is prepared to prove over time that SSDs will remain more reliable than traditional hard drives.

I've said it before and it bears repeating. I'm a solid AMD fan but the SSD division at Intel has won my respect repeatedly since the G2 release (I still feel sorry for some G1 buyers that never got firmware updates to add TRIM support nor a version of the toolkit for G1 drives).

* SSD Toolkit for G2 drives giving manual trim and scheduled trim abilities for XP/Vista users
* Seeing low failure rates for the G2 drives
* SSD Toolkit 2.0 (adding a convenient way to secure erase among other thrings)
* Now adding smart commands/registers to real world test endurance in the short term so you can calculate long term endurance based on a spec they will back with their warranty.

If they were a little faster to drop prices when C300 prices dropped I would have little to complain about with Intel SSDs.

Sandforce and Marvell controllers have their pluses but they have a tough standard to compete with.

ascl
Posts: 279
Joined: Tue May 05, 2009 1:15 am
Location: Sydney, Australia

Post by ascl » Sat Sep 25, 2010 3:55 pm

Thats some nice information, thanks! Bring on the G3 drives.... (then I can ditch my Vertex -- yes my vertex is still limping along... seems using sleep was causing problems this time. Sigh).

dhanson865
Posts: 2198
Joined: Thu Feb 10, 2005 11:20 am
Location: TN, USA

Post by dhanson865 » Sun Sep 26, 2010 5:51 am

OK, found another nugget.

Intel is switching all their PCs over to G2 SSDs (80% mobile, 20% desktop). This is coming from Intel IT department (John Mahvi, Product / Service Line Manager)

>90,000 client PCs (used by Intel employees, don't let the use of the word client confuse you).

failures ~5% for hard drives and ~0.5% for SSDs (actual use by their employees)

33% Win XP
67% Win 7

75% 80GB Gen 2
25% 160GB Gen 2

Deployment of Gen 2 SSDs is on this schedule with the goal being that at the end of 2011 all of the mobile PCs will be on SSDs.
7,000 units in 2009
40,000 units in 2010
35,000 units in 2011

Das_Saunamies
*Lifetime Patron*
Posts: 2000
Joined: Tue May 15, 2007 1:39 am
Location: Finland

Post by Das_Saunamies » Sun Sep 26, 2010 6:45 am

Thought this would belong in this thread: http://www.xbitlabs.com/news/storage/di ... rives.html

Finally. We'll see if it leads to any sort of cohesion though, but it's a step in the right direction.

dhanson865
Posts: 2198
Joined: Thu Feb 10, 2005 11:20 am
Location: TN, USA

Post by dhanson865 » Sun Sep 26, 2010 8:04 am

http://forums.storagereview.com/index.p ... dule-2010/ has links to the presentations should anyone here feel like sitting through several hours of content.

dhanson865
Posts: 2198
Joined: Thu Feb 10, 2005 11:20 am
Location: TN, USA

Post by dhanson865 » Sun Sep 26, 2010 8:45 am

Das_Saunamies wrote:Thought this would belong in this thread: http://www.xbitlabs.com/news/storage/di ... rives.html

Finally. We'll see if it leads to any sort of cohesion though, but it's a step in the right direction.
That is discussed heavily in one of the IDF sessions.

SSDS001: Solid-State Drive (SSD) Endurance Verification
Neal Mielke, Intel Fellow Technology and Manufacturing Group Director, Reliability Methods, Intel Corporation

the guy that actually wrote the spec talks in the majority of that session and covers the spec specifically in the presentation and the Q&A session at the end so if you want to know anything about that spec that is the place you want to start.

Das_Saunamies
*Lifetime Patron*
Posts: 2000
Joined: Tue May 15, 2007 1:39 am
Location: Finland

Post by Das_Saunamies » Sun Sep 26, 2010 8:49 am

dhanson865 wrote:
Das_Saunamies wrote:Thought this would belong in this thread: http://www.xbitlabs.com/news/storage/di ... rives.html

Finally. We'll see if it leads to any sort of cohesion though, but it's a step in the right direction.
That is discussed heavily in one of the IDF sessions.

SSDS001: Solid-State Drive (SSD) Endurance Verification
Neal Mielke, Intel Fellow Technology and Manufacturing Group Director, Reliability Methods, Intel Corporation

the guy that actually wrote the spec talks in the majority of that session and covers the spec specifically in the presentation and the Q&A session at the end so if you want to know anything about that spec that is the place you want to start.
Thanks, might do. The earlier Intel presentation was very informative.

dhanson865
Posts: 2198
Joined: Thu Feb 10, 2005 11:20 am
Location: TN, USA

Post by dhanson865 » Sun Sep 26, 2010 9:02 am

old post wrote:switching to AHCI mode is more complex than changing BIOS settings.
revisiting this topic. Intel mentions in the presentations above switching from IDE to AHCI mode by changing two registry entries to disable IDE drivers in windows, changing two registry entries to enable AHCI drivers in windows, then rebooting to change the mode in the BIOS.

If you don't do the driver change by way of the registry you need to reinstall windows after the BIOS change to avoid the blue screen but if you are comfortable at all with advance registry editing then you can avoid the hassle and make the change on the fly back and forth and test performance without additional parameters mucking with your comparison.

http://intelstudios.edgesuite.net/idf/2 ... DS003.html about 13 minutes in gives you the screenshots and text needed to get there if you feel adventurous.

It's also worth mentioning that Intel's IT guy (from a different presentation) said they had a few systems where AHCI wasn't enabled (by accident) and they didn't consider it a big deal (even though it affects both performance and battery life), on top of that they weren't worried about TRIM because non trimmed G2 drives didn't slow down much and the Toolbox allowed end users to address it if it ever became an issue. He said repeatedly they would rather deploy drives as is out of the box than to tweak the OS, over provision the drive, or mess with firmware updates.

Hearing that makes me glad I don't work for Intel.

Post Reply