A Whole Bunch of Solid State Disk Testing

Status
Not open for further replies.
Thanks for all the interesting data and comments, I think my next system drive will be definitely be an SSD.
 
10k limit fix = SLC drive or buy an intel mlc and accept 5 years being long enough to make a drive economical

degraded performance fix = working trim command (os support= win7, firmware update=???, controller support?)
 
Pertaining to the 10K write limit issue and the slow down issue, is there any insight that the next generation's SSD will fix these 2 problems? If so, how long?

TRIM will "fix" this issue effectively, but it isn't an SSD issue it is an Operating System issue (need to be in the OS) to be standard. But....OCZ has a beta tool out already with the 1370 that will trim the drive and help elminate the slow down issue.

That said - in real usage the "slow down issue" is typically not something you can feel.

Also, the 10k write issue has yet to be an issue that I've seen. Only in extremely heavy benchmarking is it an issue. Actual usage is quite different.
 
alright, in regards to the 10K write issue, realistically, how many yr. is safe to use the SSD? Because I don't care about the warranty. I care about my data. If I lost data due to the limitation of 10K write, I won't switch to SSD even if it's free.

As to TRIM, since it doesn't support XP Pro SP2 or higher, am I correct that I have to upgrade to Vista or Win 7?

OCZ--I believe in R&D on REAL brand name. To me OCZ is a fly-by-night brand name. I never heard of it until I click this thread. So, is Intel or Kingston's hard drive has a fix equivalent to TRIM that I can use at XP Pro?
 
alright, in regards to the 10K write issue, realistically, how many yr. is safe to use the SSD? Because I don't care about the warranty. I care about my data. If I lost data due to the limitation of 10K write, I won't switch to SSD even if it's free.

As to TRIM, since it doesn't support XP Pro SP2 or higher, am I correct that I have to upgrade to Vista or Win 7?

OCZ--I believe in R&D on REAL brand name. To me OCZ is a fly-by-night brand name. I never heard of it until I click this thread. So, is Intel or Kingston's hard drive has a fix equivalent to TRIM that I can use at XP Pro?

The 10 K write issue is not a time limited problem. It is a volume of data issue. If you use the drive as a normal system drive, it will be long obsolete before you run into the 10K write issue. If you write 30 gig of data every day to a 30 gig drive, it will die in about 27 years. Can I make things any clearer for you. Who overwrites the data on a drive every day for the life of a drive. If you are using the drive in a some kind of server based situation, then buy an SLC drive for twice the price of a MLC drive.

Your statement about never using a SSD drive if there is a possibility of loosing data is IDIOTIC. All mass storage devices fail. That is a simple fact of life that cannot be gotten around. If you are stupid enough to store only 1 copy of important data on any mass storage device, you deserve to loose that data.

An SSD that is slowed 20% because of rewrite issues is still at %100 speed when reading data. Even a slow writing SSD will still be much faster than a magnetic drive in virtually any circumstances.

My 4 drive SSD setup in WinXP is far and away superior to ANY hard drive setup even without the mythical TRIM command. It's like saying my old car (call it hard drive) can do 30 MPH. That fancy new(call it SSD) car can do 90 MPH. But because it can't do 100 MPH, I will never switch from my old car going at 30 MPH.

I have used SSDs and hard drives on identical systems, and I will never willingly use a hard drive EVER again. Yes, they are new and evolving for the better, and a little pricey right now, but they are that good.

Don
 
I still use HDD on my WHS box, but yeah I've moved fully to SSD on my daily.

For the record OCZ has been around since 2000-08, and most people who work with enthusiast parts know them for their RAM (if nothing else).
 
alright, in regards to the 10K write issue, realistically, how many yr. is safe to use the SSD? Because I don't care about the warranty. I care about my data. If I lost data due to the limitation of 10K write, I won't switch to SSD even if it's free.

As to TRIM, since it doesn't support XP Pro SP2 or higher, am I correct that I have to upgrade to Vista or Win 7?

OCZ--I believe in R&D on REAL brand name. To me OCZ is a fly-by-night brand name. I never heard of it until I click this thread. So, is Intel or Kingston's hard drive has a fix equivalent to TRIM that I can use at XP Pro?

It is nice and simple: SSD will last longer than whatever disk you are using today.

Also, who cares? Don't you backup your data????!?!??

TRIM will be something available in Windows XP, but I'd bet mainly by OCZ and other manufactures that cater to people like are on this board. They have TRIM available RIGHT NOW with Windows XP.
 
OCZ--I believe in R&D on REAL brand name. To me OCZ is a fly-by-night brand name. I never heard of it until I click this thread. So, is Intel or Kingston's hard drive has a fix equivalent to TRIM that I can use at XP Pro?

If you've been following the SSD scene for the past year, or the memory scene in the past few years, you would know that OCZ is a very respectable brand name. After Intel released their X25-M SSD, OCZ was the driving force behind pushing that type of performance to the mainstream. If it weren't for them, the Intel would likely still be in the $500-600 range instead of $320 like it is today. They're very good at listening to what their customers want, and they have a strong community over at their forums.

I'm not affiliated with OCZ, I just wanted to make sure they had recognition for their accomplishments in the SSD scene.
 
If you've been following the SSD scene for the past year, or the memory scene in the past few years, you would know that OCZ is a very respectable brand name. After Intel released their X25-M SSD, OCZ was the driving force behind pushing that type of performance to the mainstream. If it weren't for them, the Intel would likely still be in the $500-600 range instead of $320 like it is today. They're very good at listening to what their customers want, and they have a strong community over at their forums.

I'm not affiliated with OCZ, I just wanted to make sure they had recognition for their accomplishments in the SSD scene.

I think we have OCZ to thank for the huge Indilinix hype machine and killing the Samsung gen 2. controller in the US market. I *do* believe however, that the emergence of Indilinix forced Intel's hand in rushing out their new firmware for their mainstream drives (MLC)--since they haven't bothered with the SLC variant, knowing that enterprise applications do not focus on sequential benchmarks (which is what the X25-Ms were struggling with) but IOPS, which is maintained regardless of fragmentation.

My personal opinion is Intel would have lowered their prices regardless. If there's one thing Intel does, it's maintain a very aggressive development roadmap (miniaturize their manufacturing process steadily every 2 years).

That said, OCZ's pricing is higher than the rest of its OEM partners, and in particular, it's new SLC Vertex EX line is a bit overpriced (95% of the Intel X-25E series but not within 95% of the performance).

I am appreciative of the OCZ support forum, although it's definitely a total organizational mess and you have to put your time to sift through the info, but it's definitely there.
 
My personal opinion is Intel would have lowered their prices regardless. If there's one thing Intel does, it's maintain a very aggressive development roadmap (miniaturize their manufacturing process steadily every 2 years).

OMG, he's drunk the coolaid.

The ONLY reason Intel, or ANY other manufacturer lowers prices is because of competition. They will NEVER lower prices just because a product has been around for a while. If they thought they could get away with charging a grand for every SSD, you bet they would.

You want to see what happens when there is no competition, just look at the high end desktop hard drive market. WD is the only OEM that wants to play in that market, so the enthusiast has been getting screwed for the last 7 years if they wanted a fast hard drive.

Don
 
The 10 K write issue is not a time limited problem. It is a volume of data issue. If you use the drive as a normal system drive, it will be long obsolete before you run into the 10K write issue. If you write 30 gig of data every day to a 30 gig drive, it will die in about 27 years. Can I make things any clearer for you. Who overwrites the data on a drive every day for the life of a drive.

An SSD that is slowed 20% because of rewrite issues is still at %100 speed when reading data. Even a slow writing SSD will still be much faster than a magnetic drive in virtually any circumstances.

I shouldn't say time. Let me re-phrase, I meant after the drive write the same spot 10K times, it's toasted. I don't have to worry about that problem w/ hard drive. And I don't see any reputable web site link to your story in page 2 saying:

no block is written to second time till all have been written to the first time.

From this link, they didn't say only slower by 20%.

http://anandtech.com/storage/showdoc.aspx?i=3531&p=8

Writing 12KB should have taken 12 seconds but since we had to read 12KB and then write 20KB the whole operation now took 26 seconds.

a job that takes 12 sec. now took 26 sec., by my math, it's a lot more than 20%

Also, any1 who read page 8 of the report would know that a design like that, is clearly idiotic. I don't know why any1 would come up w/ this kind of design.

as to your story about that 27 yr. thing, that's not what the report said neither. And do explain how you cook up this 27 yr. no.

http://anandtech.com/storage/showdoc.aspx?i=3531&p=6

It gets worse. Every time you erase a block, you reduce the lifespan of the flash. Standard MLC NAND flash can only be erased 10,000 times before it goes bad and stops storing data.
 
If you've been following the SSD scene for the past year, or the memory scene in the past few years, you would know that OCZ is a very respectable brand name.

In this industry, the most important thing is R&D. Intel spent billions on R&D. OCZ is nobody. They couldn't possibly spent billions on R&D. I don't care how nice their customer service is. And R&D takes time, it can't be done w/ a taiwanese fly-by-night w/ a few yr. under its belt, w/ very limited amt. of $$
 
In this industry, the most important thing is R&D. Intel spent billions on R&D. OCZ is nobody. They couldn't possibly spent billions on R&D. I don't care how nice their customer service is. And R&D takes time, it can't be done w/ a taiwanese fly-by-night w/ a few yr. under its belt, w/ very limited amt. of $$

Yes, Intel has deep pockets, so they spent millions to develop their own SSD controller chip, and that money must now be recouped in higher prices.

So fine, go overspend for your Intel SSDs. They are the best ones out there right now. I will happily spend a third less for a drive that is almost as good and we can both be happy.

Besides, if Intel's R&D was so awesome, why did it take a third party web site to expose the flaw in their firmware that they fixed last month? :p

Don
 
Intel hasn't spent billions on their SSD tech. Their entire market cap isn't even 90 billion. Those are the types of numbers spent on the entire life of something like the C2D. Did they spend a lot of money? Yeah. Is their product currently the "one to beat" in the consumer market place? Yeah. Will they end up spending billions on SSD R&D? Probably.

The main difference is that OCZ is a packager and reseller, like 95% of the companies in the tech industry. What gives them the edge is the ability to work closely with their customers (which they do), and their vendor (which they do) to make lightening quick changes to their product offering (which they did with the vertex). That's the beauty of a small company.

All of that being said, I love my X25-E.
 
I shouldn't say time. Let me re-phrase, I meant after the drive write the same spot 10K times, it's toasted. I don't have to worry about that problem w/ hard drive. And I don't see any reputable web site link to your story in page 2 saying:

Hard drives have a very high failure rate that tends to get higher with age. Google Report By 5 years, almost half of ALL hard drives will fail. Unless you are overwriting your MLC based drive several times a day, they will be much more reliable over time than a hard drive. I spend quite a bit of time here in the [H]ard forums. And I see daily threads about hard drives that have failed, and arrived brand new but DOA. Now granted SSDs are new to the scene and we don't have any long term usage on them yet, I have yet to see any DOA or failed SSD threads yet.

From this link, they didn't say only slower by 20%.

http://anandtech.com/storage/showdoc.aspx?i=3531&p=8

a job that takes 12 sec. now took 26 sec., by my math, it's a lot more than 20%

You are picking one specific example. If you look at this page of the same report, you will find and overall 22% drop in performance for a used SSD. So I rounded to 20%, whoop dee do.

Also, any1 who read page 8 of the report would know that a design like that, is clearly idiotic. I don't know why any1 would come up w/ this kind of design.

I agree 100%. What were they thinking. They prolly adapted the flash (thumb) drive controllers, and that's how it was being done there.

as to your story about that 27 yr. thing, that's not what the report said neither. And do explain how you cook up this 27 yr. no.

http://anandtech.com/storage/showdoc.aspx?i=3531&p=6

Just simple math. 1 rewrite a day, 27 years of life. Say for the sake of argument, there is a couple of maintenance writes for every normal write, that will still give you 9 years. MLC drives are not meant for that kind of use.

IF YOU ARE WRITING THAT MUCH DATA TO A DRIVE THEN GET A SLC BASED DRIVE.

In normal use, a MLC based drive will be long obsolete before you run into the 10K rewrite issue. And your magnetic drive will have died from seized bearings long before that.

If you read the rest of the recap in that article you are quoting from, he states,

It turns out that the benefits are more than worth the inconvenience of dealing with these pesky rules; so we work around them.

I can't make it any plainer than that! :)

Don
 
Last edited:
OMG, he's drunk the coolaid.

The ONLY reason Intel, or ANY other manufacturer lowers prices is because of competition. They will NEVER lower prices just because a product has been around for a while. If they thought they could get away with charging a grand for every SSD, you bet they would.

Why did Intel lower prices on the X25 series in December?

You want to see what happens when there is no competition, just look at the high end desktop hard drive market. WD is the only OEM that wants to play in that market, so the enthusiast has been getting screwed for the last 7 years if they wanted a fast hard drive.
Don

The 74GBRaptor drive costs $99 today. Back in 2003 the 74GB Raptor was $299. Why?

Besides, if Intel's R&D was so awesome, why did it take a third party web site to expose the flaw in their firmware that they fixed last month? :p

Even a degraded Intel drive is still faster than the second fastest drive out there. Fresh off the boat, 70MB/s 4k writes vs 12MB/s for the OCZ Vertex 120G. Even with 90% degredation it would have equal the steady state performance of the fastest competitor (8MB/s).
 
Last edited:
My personal opinion is Intel would have lowered their prices regardless. If there's one thing Intel does, it's maintain a very aggressive development roadmap (miniaturize their manufacturing process steadily every 2 years).

The ONLY reason Intel, or ANY other manufacturer lowers prices is because of competition. They will NEVER lower prices just because a product has been around for a while. If they thought they could get away with charging a grand for every SSD, you bet they would.

You both have valid points. Intel does do a large amount of its business like clockwork: Tick-Tock processor cycle, and even scheduled SSD price drops (such as this -- I can't find the link that shows that it was part of a roadmap, but I assure you it was). On the other hand, if you look at the high-end processor market, you will notice that the $1,000 Extreme processors take forever to come down in price because they have no competition--although they do come down eventually. So as you can see it's a little bit of both planning and succumbing to competition. I think if Intel didn't have any competitors that had a chance (Indilinx sure does), then the price of the 80GB X25-M would still be in the $400-500 range, but still not nearly as much as it was at launch.
 
Why did Intel lower prices on the X25 series in December?

I don't know. Maybe it was the economic downturn was hurting sales. I am fairly sure that it was not because they were feeling generous. A high launch price and rapid price drops are normal to screw the early adopter market. That is just normal marketing strategy.

The 74GBRaptor drive costs $99 today. Back in 2003 the 74GB Raptor was $299. Why?

Because it's performance no longer commands a premium. The competition was starting to catch up, so they finally came out with a newer and faster product to command the high premium they charge. SSDs can compete with the VRs though. It sucks to be WD.

Even a degraded Intel drive is still faster than the second fastest drive out there. Fresh off the boat, 70MB/s 4k writes vs 12MB/s for the OCZ Vertex 120G. Even with 90% degredation it would have equal the steady state performance of the fastest competitor (8MB/s).

What does that have to do with the price of Hoes in Vegas? I was just pointing out that Intels vaunted R&D aint perfect. Just look at all of their past miss steps in the CPU arena up until the C2D. AMD was kicking their ass, because Intel was not willing to invest enough money in R&D around the turn of the Century. Once they got serious, because of competition, they started kicking some ass, and continue to this day.

Intel's entry to the SSD market was good, maybe even great, but not perfect.

Don
 
Benchmark request>

Ubuntu 9.04 with an mdadm raid0 of 2 drives of each type. Can be on the on board controller or the Areaca, but I prefer onboard.

Do bonnie++ benches please! iozone would be great as well.
 
Give me the command lines to run and I can run against a mdadm RAID0 of 2x Vertex 30g (also 2x 60g Solid [removed the third so I could put the DVD back in]). Only have 1 intel drive though.
 
, I have yet to see any DOA or failed SSD threads yet.

Just simple math. 1 rewrite a day, 27 years of life. Say for the sake of argument, there is a couple of maintenance writes for every normal write, that will still give you 9 years. MLC drives are not meant for that kind of use.

IF YOU ARE WRITING THAT MUCH DATA TO A DRIVE THEN GET A SLC BASED DRIVE.

In normal use, a MLC based drive will be long obsolete before you run into the 10K rewrite issue. And your magnetic drive will have died from seized bearings long before that.

Of course there is no DOA or failed SSD, it's a box of memory, what's the chance of DOA. And the tech. is so new, most people doesn't have it yet, what's the chance to see a defective drive when most people who have it, got it for less than a yr.

Say I get a SLC drive, does the 10K write issue still occur? If not, is there a higher limit no.? I'm no ver in SLC drive.

As to MLC drive, why is there 1 write per day? An av. user write to his hard drive easily hundreds of time per day.
 
Of course there is no DOA or failed SSD, it's a box of memory, what's the chance of DOA. And the tech. is so new, most people doesn't have it yet, what's the chance to see a defective drive when most people who have it, got it for less than a yr.

In the last year I have bought 5 hard drives. 1 was DOA. I also had one of my backup drives in a USB enclosure die. In the last 5 months, I have bought 8 SSDs. 1 required a firmware update to increase performance. Otherwise they have all worked fine.

Say I get a SLC drive, does the 10K write issue still occur? If not, is there a higher limit no.? I'm no ver in SLC drive.

SLC memory has 10 X the theoretical lifespan of MLC memory. It is also faster. If you have an application requires extensive rewrites to the drive, like a database server, or want the fastest drive around, then buy an SLC based drive.

As to MLC drive, why is there 1 write per day? An av. user write to his hard drive easily hundreds of time per day.

Remember, I am just giving you a for instance to give you a real world reference as to how long MLC drives should last.

If you completely fill the drive, and then completely erase and refill it the next day, you can do that thousands of times. Remember, most hard drives are not used this way. Even if you erase and rewrite only a small portion of the data on the drive, the SSD controller will automatically spread the new writes evenly over the drive so you can't wear out one portion of the drive prematurely. Once data is written to the drive, you can read it an unlimited number of times.

Don
 
Actually, in the last year I've had about 2/3 of the spindle drives arrive DOA. Damn seagate 7200.12 drives.
 
The testing is good,but I would wait awhle to get one.I love mine and it was fast,but it was painful to get it working like it should,and then on top of that.I have not reached the full speed of this drive yet.Mabye its a issue with my board,but what ever the case may be.I would advise anyone to wait a few months yet on buying one.Mine has not slowed down and its about half full,but then again I cannot get to full specs either,but its fast enough.I will post back later as its gets about 90 percent full.
 
Last edited:
I'm personally waiting on win 7 official release to pick up a SSD, hopefully by then SLC prices aren't redic.
 
SLC memory has 10 X the theoretical lifespan of MLC memory. It is also faster. If you have an application requires extensive rewrites to the drive, like a database server, or want the fastest drive around, then buy an SLC based drive.

I have a stock streaming data software that writes data from streaming life feed on various stock in that 6.5 hr. of stock trading. So say I track 50 stocks, then these 50 files would be written every few sec. for all 50 files, daily for the duration of 6.5 hr.

Say they write the data every 5 sec., then in 390 min., or 23400 sec., a total write of 23400/5 = 4680 write PER DAY PER FILE. The market opens 5 days a week, so you do the math.

Now, my concern is w/ this kind of writing data on various data files, assuming 1 file per stock say, I can't see how can a SSD last?

Even if you erase and rewrite only a small portion of the data on the drive, the SSD controller will automatically spread the new writes evenly over the drive so you can't wear out one portion of the drive prematurely. Once data is written to the drive, you can read it an unlimited number of times.

where's the link from a few reputable web site article to confirm and read more of the above.
 
I have a stock streaming data software that writes data from streaming life feed on various stock in that 6.5 hr. of stock trading. So say I track 50 stocks, then these 50 files would be written every few sec. for all 50 files, daily for the duration of 6.5 hr.

Say they write the data every 5 sec., then in 390 min., or 23400 sec., a total write of 23400/5 = 4680 write PER DAY PER FILE. The market opens 5 days a week, so you do the math.

Now, my concern is w/ this kind of writing data on various data files, assuming 1 file per stock say, I can't see how can a SSD last?

where's the link from a few reputable web site article to confirm and read more of the above.

How big are these writes? If they are 4K or under, then 23,400 writes is nothing more than ~100MB a day. Windows Virtual Memory will write to disk 8-10MB/min, so the stock writes you mentioned in particular is nothing but a blip on the radar...assuming my math or premises aren't way off.
 
Snip! Already answered.

where's the link from a few reputable web site article to confirm and read more of the above.

There's a little company with a search site out there called Google. I'm sure you can find a little info about wear leveling there. Cmon, do a little basic research before you try to start an argument.

Don
 
I have a stock streaming data software that writes data from streaming life feed on various stock in that 6.5 hr. of stock trading. So say I track 50 stocks, then these 50 files would be written every few sec. for all 50 files, daily for the duration of 6.5 hr.

Say they write the data every 5 sec., then in 390 min., or 23400 sec., a total write of 23400/5 = 4680 write PER DAY PER FILE. The market opens 5 days a week, so you do the math.

Now, my concern is w/ this kind of writing data on various data files, assuming 1 file per stock say, I can't see how can a SSD last?

-The drives write to unused space before rewriting any cells.
- If you rewrite a file it is written to unused cells, not to the cells it was on.
- Intel, at least, not sure about the others, moves unchanged data periodically so that the cells that are being used least are available to be used.

If you buy a 60GB drive and write 60GB a day (though, who could?) you can rewrite to that drive for more than 27 years. (If MLC.) Don't worry too much about the 10,000 write "problem."

Now then, WHY ON EARTH ARE YOU WORRIED ABOUT A SINGLE POINT OF FAILURE? AREN'T YOU BACKING STUFF UP?

Data is only as good as it's last backup.
 
God I don't think Happy is gonna get it, just continue to live in his own world where bias controls all his purchasing decisions and progress halts. He needs to lay off the Haterade.
 
OCZ is a wannabe-Corsair, just like Energy and Paradigm are a wannabe-B&W companies..Zotac...i dunno, maybe a wannabe-MSI.
 
Hey everybody,

I am trying to decide between the OCZ Vertex and Intel. Can someone explain why Anand makes such a big issue about 4k RANDOM WRITES? I mean, I have used a FREE program called Process Explorer to determine that with 31 processes running, there are no write bytes happening when the system is not in use. Although Windows Explorer does READ 664 bytes about every 2 seconds, there are NO WRITES going on.

Further, a 4k write would happen in 1/250 of a second even if the WRITE speed at 4k was only 1 MB per second. The Vertex drive does 11.39 MB per second according to Jason on this thread. Therefore, one 4k write would take just 1/2848 of a second. Yes the Intel does 55 MB per second on 4K writes but is all that speed really needed? Can anyone explain? Are there a ton of 4k writes while browsing the internet or doing other things? Any insight anyone?
 
Last edited:
I've got a 60 gig Vertex running on my work laptop, and it is smooth as silk, even when replicating my Lotus Notes databases.

IMHO, save your money.

If this system is a mission critical database server, then spring for the Intel.

Don
 
I'm planning to get an OCZ vertex 120GB when windows 7 goes live.
Got 2 1TB seagate 500gb/platter drives coming. Those will be my storage drives in windows 7. And the SSD will be the system drive.

SSD is the best thing since sliced bread :)
 
Status
Not open for further replies.
Back
Top