The new SSD Thread (Benchmarks, Brands, News and Advice)

Intel released the specs of their SSD already and it is very slow, albeit also quite cheap. But you'd get the same level of performance from plugging a Compact Flash card into an adaptor and that into your laptop.

 

Intel will not release a dog. Specs change. You watch. Why do you think they are waiting so long? Dave

 

the released specs are for ultra low systems, namely the atom based MIDs.. for them, they're very fast (having an eee currently, i can see that even slow ssd's really give great responsable feeling => good performance)

 

I believe Intel will rock the world when they release a 2.5 inch fast SSD. I believe the only reason they have not already is that Seagate and WD would blow them up if they did Dave

 

Can anyone tell me if SLC is also slow for random writes ? How does an SLC drive compare to a regular 7200rpm HDD when it comes to random writes ?

I'm considering a SSD drive but these reports of slow random writes bother me as i have to compile alot and this generates alot of small files.

 

http://www.engadget.com/2008/07/06/new-report-says-ssds-are-in-fact-more-efficient/

SSDs more efficient than HDs. Les2win.

edit: also the OCZs are MLC as expected, are the lifespans that bad? hope they don't die like many people's Super Talents (even if they are lightning fast)... any chance of a review/sample from OCZ, Les?

 

I am really not sure aboput this MLC at all. I mean, is that what they have been using in Jet Fighters for the past decade or two? I THINK NOT! I want reliability as well as speed. I want affordable SLC. Come on Intel, let's get this party started! Dave

 

The past decade or two in Jet Fighters? You're talking then about F15 and F16s -- both of which are based on 30+ year old projects. That's older than Flash Memory itself. Also, I expect the requirements for a jet fighter to be dramatically different from a computer system (server, desktop or notebook).

I would guess you're probably looking at a dramatic difference in transistor count when discussing SLC vs MLC; but I'm not an expert on this topic.

 

Did you guys see this? I'm thinking of getting the 64GB version..ETA July 13th.

 

Yes I saw it. I'm eying it to. But first I'd like to see some benchmarks.

Either I'll get SSD or a 320GB 7200rpm.

 

Since the past few pages, the OCZ Core SSDs has been what we've been discussing It's MLC hence why it's so cheap.

 

But I don't think anybody has ETA yet. Even engadget didn't mention ship date. That's the first ETA I've seen in the US.

 

And interesting Directron's page does not state either MLC or SLC. The speed sure looks alot like SLC, but the price is MLC. VERY VERY interesting. Dave

 

In Europe is it expected in a couple of days. So it should not be long in US too.

 

Is it really MLC? I don't really care either way, as long as the speed is double or triple regular HDD. And close to if not around the SLC SSD speed.

That's why I'm very excited. I've seen those regular 40MB/s write speed SSD for a long long time. Been waiting for those 100MB/s read write to come down on pricing.

 

It has been confirmed MLC.

 

Do you have a source/link?

 

Had to retract what I wrote to add just a bit....

The new OCZ ssd has a mtbf of 1.5m hours.... This is common for slc drives and I wouldn't believe poss for mlc drives. Secondly, their largest size offering according to the article will be 128Gb which is what Memoright accomplished in their slc ssd.

I have looked everywhere to no avail and am going to suggest that thesizes offered and especially expected lifespan match that of a slc drive.

On the other hand, speeds reflect that of mlc with improved controller technology. It is possible though that this is OCZ flagship slc offering from their own workshops as we know their SATA II drive was the spitting image of the Samsung SATA II.

 

The game is afoot!

 

Sure it is. It's an aggregate failure rate, and doesn't apply to a single disk. If you have 10 of them, that's about 150,000 hours between failures. If you have 20, about 75,000 hours. Anyway, that's the way the math works.

I would like to see the MTTF (Mean Time To Failure) numbers which would be a more accurate arbiter of a single disk.

Aren't they all going to be powers of 2? I know the Super Talent's aren't sizing the drives that way (yet) but I wouldn't be surpised if you had more reported space than their claimed size of the drive.

It's possible. I do wonder if these might be produced on a smaller trace size plant, like say a 65 nm plant. This would give them greater yields from either technology which would lower the cost dramatically.

Best,

 

How well do SSDs perform in RAID 0? I'm seeing impressive Read/Write speeds, and I'm thinking of buying two 64GB drives at ~$279 each and put them in my desktop in RAID 0 for my boot drive. =D I'm imagining blazing fast speeds which aren't possible on laptops. =D

 

there are benchmarks of up to 9 mtrons in raid0, going close to one gb/s. so they scale very well.

I'm interested if mlc scale as well, espencially in write performance.

then again, lets just wait and see what OZC offers, and how well they scale, if they're very cheap for the performance.

 

http://blog.laptopmag.com/ocz-super-affordable-core-series-ssds-use-mlc-flash-will-reach-256mb

"So we spoke with OCZ CEO Ryan Petersen about these potentially game-changing SSDs. In our interview, Peterson confirmed that the drives are made Multi-Level Cell (MLC) flash, and that they will be available for purchase July 12th or shortly thereafter."

 

And if you lose 1 drive, you lose all your data. Raid-0 provides no redundancy.

If you're comfortable with that type of failure mode, that's fine. I'm not. Drives fail; even SSDs. They usually fail when you have no method of backup in place.

Cheers,

 

everyone's sorta neglecting the transcends LOL. check this out:
http://www.mydigitaldiscount.com/SP...olid-state-disk-sata--8000078E-1214298125.jsp
u reckon the intel ssd will have a better size to cost ratio? I suppose after the failure of transcend's first batch of SSDs......

 

64GB is too small.

 

I think it looks like a real good deal. I would like to see reviews and know more of reliability. The time has come. It is here now. SSD's will be available to those who care at reasonable prices starting... NOW! Dave

 

64GB on a notebook is just enough for me My old notebook only has 40GB and I'm doing fine...

 

Dave:

It's an ok deal, except that we know what's coming down the pike in a very short time window.

Cheers,

 

Actually, I considered RAID 0 because SSDs supposedly have better reliability than normal hard drives. So it would be a waste of its speed to set them up in RAID 1. As for backup, I'll have two 500gb hard drives that back up each other and also back up my boot drive. =) So with 4 hard drives, I have complete redundancy and RAID 0 speeds for my OS and applications.


Those transcend drives seem slower and more expensive than the OCZ SSDs. I'll pass on those.

 

RAID 0 is far less reliable than 2 normal hard drives. You would need at least 3 SSDs for redundancy using RAID 5 (I believe that uses the third drive to hold the parity info).

 

raid0 by itself is reliable (except the raidcontroller dies.. )

the trick is, it reduces the workload of the ssd to half, so it actually helps each ssd to life longer (espencially due to reduced write-workload and a possible cache on the raid, reducing scattered writes).

so actually, the individual disks in a raid0 live longer. on the other hand, the chance that one fails, taking down the whole raid, is bigger.

but a raid0 with ssd sounds quite reasonable with a regular backup. they don't have the random head crashes which let most of the raid0 setups die.

but the reliability of ssds itself is still another question. it can only get answered in the future.

 

I'm sorry, I'm having a tough time responding to this politely. From this statement it would seem that you don't have much experience.

This really isn't a good idea either. If two 500s are "backing each other up" and backing up the boot drive you wouldn't have much actual usable space.

128GB from each to back up the boot drive (2x64 GB).

That leaves ~325GB. To properly backup one another neither drive can use more than 1/2 of its capacity... that's 162.5 GB usable per drive.

Now maybe you worded this poorly, and you are talking about a RAID 1 for these drives. It doesn't change the math, but it is better than nothing.

I've had enough drives fail on desktops, laptops and servers to know that RAID 0 isn't an option for me.







Those transcend drives seem slower and more expensive than the OCZ SSDs. I'll pass on those.[/QUOTE]

 

raid0 is always a good option for the "system-disk", as long as you image it well.

using matrix-raid on the intels, one can nicely set a fast systemdisk for the os and the apps in raid0, and the resting space of the disks in raid1.

the actual backup of the system should never be stored in the same place anyways => unimportant for the raid setup.

this is, for home-systems. business is another thing.

i'm currently running on that setup here at home on my pc and it's very nice. backing up trough homeserver (and nothing else => no danger of the currently not yet fixed file deletion bug), raid0 for the first part of two 500gb disks, raid 1 on the rest. the important data moved onto the raid1, the rest on raid 0. the rest, that is: system, apps, games, data like movies or stuff that is synched with other places anyways (all my projects from the svn server))

most of my data is shared with my notebook anyways, and regular backups of my data happens as well, but this to an external disk.

thinking of getting a big raid5 nas sometimes in the future... so i can put all my data on there and "be done".

 

That makes no sense at all. If the individual disks last longer since there is less write-workload then how can the chance of failure be higher? I also haven't seen a RAID controller dying... While MLC SSD reliability still needs to be proven, SLC SSD is much more reliable than traditional hard drives and have been proven over the past few decades.

 

hmmm... i think what dave means is that by splitting writes over two ssds, each one will live longer. For example, 2 64gb used in raid 0, would last as long as a 128gb model (but with double the speed). SSDs being more reliable than regular hds, have little drawbacks wen used in raid 0. So even though the chance of a crash is doubled, that chance is still very small, small enough to be considered neglidgible.

 

The failure rate increases as the number of drives in the array increases. With RAID levels higher than RAID-0 you have a measure of redundancy which protects you from an outage caused by a single disk's failure.

With RAID-0, you get better performance, with the tradeoff being that your failure likelihood increases. Think of it as MTTF / N statistically speaking. In other words a single drive expects a failure within N. Two drives becomes MTTF / 2, 3 drives MTTF/3 etc. So if you have for argument's sake a RAID-0 grouping of 8 drives, the expected MTTF is 1/8th that of a single drive.

With other levels of RAID, the MTTF equation is the same, but a single drive loss doesn't lose data.

Cheers,

 

THis may or may not be the case. The SSD can fail for any number of reasons, not just by excessive write aging.

His failure is double that of a single drive.

It's his system, and he can do as he pleases -- my experience says it's a mistake.

Cheers,

 

Um, no, it's a bad idea. Lots of software licenses and activations are tied to the primary disk. So if you have to restore the image on it, you can have to relicense/reactivate stuff. In particular Windows Vista is inactivated every time you change the primary drive, and most times that you re-image it. In fact, Vista can be inactivated by simply changing the device driver for the primary drive (for example if you have to change into/out of AHCI mode).

Anything you can do, to reduce the number of times you have to do anything with the primary drive will make your life better. Unfortunately you can't go around to software vendors, find the idiots who make the decision to use the primary drive as a handle for licensing/activation, and thrash them to within an inch of their lives with a baseball bat.

 

I've had this question for years but nobody could give me a definitive answer on it:
If i partition my SSD, will the drive's wear leveling still apply to the whole drive or to individual partitions?

I would suppose that it would still apply, being at a lower level than MBR and the such. But i need a definitive answer. No guesswork, i can do that myself.
This is important cause i want to protect the OS partition with a write filter, but still have the ability to make changes to my data partition without having to commit change in commandline everytime.

 

Well I hope that's true, but the shortest lived drive I have ever had was my Super Talent 128GB SSD, which lasted about two weeks in service as a primary drive.

I agree that in time, the SSD will win, in just about every way, but it's early days for these things. I will try that new OCZ 128GB as soon as I can.

Sometimes the leading edge is the bleeding edge.

 

shouldn't happen with windows home server as a backup system. and even then, the activation doesn't hurt me at all. never did. i'd prefer to not have it, but i have no problem with it, as it's an understandable thing.

even phone-activation is no biggie. it's less work than searching for the newest version of the crack for the os because somehow they detected my key is evil again.

i'm only having valid licences today and i'm actually sort of proud to activate. no big hassle for me.

and still, a raid0 doesn't mean your system will fail daily or weekly. it just means it has a higher chance. i'm running my newer home-pc now since some months in this mode, and it hasn't failed so far. have had single harddrives which failed faster than this.

a failing disk always sucks. but a failing raid0 for me is just like a failing disk + you get a disk for free.

 

Have you tested a restore? If you haven't then how do you know it will work when you need it to work? Windows' backups have been notoriously flakey at best.

I don't have big issues with re-activation except that they've gone overboard with activations.

I'm sure you have, you were in the infant mortality phase.

When it fails (and it will fail) please let us know how the recovery went.

You do? I didn't realize it was a buy one get one free deal If that's the case, striped mirrors for me.

Cheers,

 

It's to each memory cell. Each memory cell has a finite lifetime. The typical failure time is X where X is the specified lifetime. It sounds tongue in cheek, but the failure of single cells will happen before the end of the lifetime of the device. These should be automatically remapped to spare blocks from the spare block pool.

Hmmm... have to ponder that a little more. It should work in theory, but in practice will there be enough gain to make it worthwhile?

 

A: I have Vista. Activation has been hell.

B: Phone activation has been unpleasant. As we speak, I am running on a system that I will have to phone activate since the Super Talent SSD died. I put off the activation as long as possible to get all the device drivers and other software set up to reduce the chance of deactivation after activation. I've been through this about five times in the past nine months.

C: I have valid licenses. Just today I renewed my Mathworks software even though that means going to activation as opposed to the previous licenses. It's like paying money to have another headache.

My experiences with activation almost make me physically ill. I've lost at least four or five full man days just rebuilding my system from things crapping out. I've been computing since 1969 and this recent activation/primary crapout fun is about the worst pain I've ever had in computing.

Raid 0 on the primary? Not in my lifetime. I'm the guy that started the thread about getting a Raid 1 primary notebook. At the moment, they don't make them small enough for me (15.4" too big to travel) but I am going to go there.

 

Wear leveling algorythm are more complexe than just having a pool of spare cells. What you're describing sounds like what a normal HDD would do if they found bad sectors, and that's only when it happens. But SSD wear leveling distributes writes to cells that have had less writes to them, and sometimes even moves around blocs of data that never gets modified.

 

nope, not yet. I'm currently too much reworking my whole network at home, so at current the system is not working. But there isn't much data alive currently that would get lost on a hd failure anyways. most is shared from pc to notebook, and to an external disk => a disk failure would at worst mean some days back lost.

I never had to reactivate any of my systems yet. Including tossing around quite some hardware. All i had to is activating notebooks by phone as i reinstalled them with an original vista disk after bought, as they had tons of stuff preinstalled that i didn't wanted to have. but using a cellphone, that's easy. set it to speaker/loud mode, and place it besides the notebook, and start typing. a 5min job at most, done once per notebook so far.

my english fails here, sorry. (i'm from switzerland)

yes, i'll do. but i'd guess, before it fails, i'll have one or two of the fast cheap ssds that come out soon to place in there

yep, one for free, namely the one that is together with the pair. if the c: wasn't that full (less than 50%), i can even restore the backup right onto the single working disk, and continue till i have a next second disk.[/QUOTE]

so two in raid0 can be even better than 1 for uptime, as you can't buy a fresh hd at saturday evening till monday morning (and you can sure bet it'll die then, just because it likes to)



and if the backup doesn't work, i just reinstall vista, isn't a big hazzle. i have a nice small list of what gets installed on it, it isn't that much. takes a while, but isn't something terribly bad. (but i do should get a vista disk with sp1 integrated someday..)

i know how bad it can be to have a disk failure if you're not prepared for it. most of my disks that died where external usb-disks, and i've lost quite some gb worth of data years ago. that's why i moved over to multiple systems => multiple places for my data. that way, if one dies, i still have it all. unimportant if the disk fails, the raidcontroller (been there), or the whole pc/nb.

oh, and as a dj, i do have to have 2 notebooks with me for gigging, if one fails, i can quickly turn on the 2nd. so i do have a raid1 all the time, in hardware, including the whole system

thats why i don't care much about the single disk. it's utterly unimportant for me.

 

anyways, discussion is actually offtopic in here.. (at least partially if we go too much into detail and too far away from ssds in raid, performance? savety?)

ontopic stuff:
http://www.dailytech.com/Samsung+128GB+MLC+SSDs+in+Mass+Production/article12319.htm

yay.. looks like that's the real OZC (as they till now allways just resold other companies products in general, and samsung ssds in particular..)

 

Sorry if I'm missing something here but I thought whether it has SLC or MLC, SSD were supposed to be dramatically faster than regular HD?
speeds of 90MB/s and write speeds of 70MB/s, I've read that the new hitachi Deskstar 7K1000 can pull 173.5MB/sec...

 

Duh... I know that. but didn't i say...

i didn't indicate that by 'crashes' i only meant maxing out the ssd's writes.