OCZ Vertex preview: Interesting results!

[geo]

4k random writes over 1MB/sec is extremely good as measured here:

http://www.xtremesystems.org/forums/...d.php?t=167857

You can see for yourself that 7200rpm SATA drives would struggle with that even on a Caching controller and 10,000rpm and 15,000rpm scsi/sas drives can manage 3MBps at those sizes.

If a Vertex can pull even 5MB/sec with those settings its nothing short of flat out amazing for that file size.

nice info there...

[DoubleZero]

Random writes drop the Vertex to 20MB/s, better than what we have but still not very good at all.

And still better then a velociraptor in 64k random writes. And if you compared the sequential writes from yapt to atto theres a huge diference, in yapt they never go above 100mb\s in atto they never go bellow 220mb\s from 64kb to 2048kb. So one can assume that yapt random writes are wrong to. It's in the review yapt "hasn't been updated or used in quite some time".
And probably every firmware used in the review won't be the final one.

[One_Hertz]

Why would the controller fail? The controller is basicly a CPU(ASIC) and nothing else. Plus the amount of data goign through the controller doesnt matter.

Controller to die first..lol...

And people talking about defragmenting SSDs. Great. Better run your memory defragmenters too!

What are you basing your completely off the line statements on? Do you do data recovery for a living like me? I think not. Memory chips don't die; well, they do in about 3-5% of the times, but that is usually where the controller fails first and takes the chips with it. I am basing this on the tons of USB sticks that I see. I haven't seen too many SSDs quite yet, just some Samsung MLCs they use in Mac laptops. PCBs on hard drives die too and those are much more polished than what they currently use for flash storage.

The controller chip is much more than just a 'CPU'. It is that chip that determines where all the data fed through the bus goes (which chip and where on the chip). It is a MUCH more complicated device than a NAND chip.

Back on topic: they really should measure writes that are smaller than 64K. If you take a look and extrapolate you will see all OCZ SSDs taking HUGE dumps if smaller writes are benchmarked (majority of the OS operations are dealing with small reads/writes BTW). The Intel drive does not drop anywhere as much. I suppose that was hidden in order to hide the flaws of OCZ SSDs (as the owner of the site is friends with OCZ). IOMeter is also rigged with OCZ supplied settings; notice how they don't even post the settings anymore (they only use 2million sectors and don't run it for long), how unprofessional. Wait for proper reviews people.

[FischOderAal]

Then read the whole article... why and I mean why in the world are you talking about something you don't understand first? In the first page there are two ATTO screenshots for lazy people, if you had seen it you'll have read the whole article

Allright. I read the article entirely and yes, I did understand the problem I've read about this issue several times all over the web but only alongside with the X25-M. My error was not to take wear leveling into account, what fragments the drive as well.

Yes, I was wrong as I stated other SSDs don't have that problem, I should've written: "not that extreme and quickly".

You can read in that article a few times that "write combination" appears to be the primary reason for fragmented files.

All we did here was write the OS files (~20GB) back to the drive. This was not a full blown installation process or other activity that would normally cause excessive fragmentation. What you see is the result of internal fragmentation caused primarily by write combining.

and

It is likely that other manufacturers will employ similar write combining techniques in the future, and with those new devices may come similar real world slowdowns.

But not every SSD uses the technique of "write combining" and not every SSD uses MLC-cells and therefore the effect is not the same on every SSD. So cursing every SSD is wrong imho. Furthermore, as I already stated, SSDs are not only about MB/s

[MF Traum]

This is a bit off topic question but is there any good tests to show differences with let's say four cheaper SSD's on a good RAID controller against a couple of more expensive ones on Intel's onboard controller. I admit that I haven't been following up pretty much any hardware development in four months now.

[kiikkuja]

Common sense says you don't. Yet I don't see your point because in general products live several times longer than the warranty period. Heck, we still got a Sony HiFi in the house that is older than I am and a Thinkpad that was build in 1998 and still runs fine.

Warranty is nothing but a bonus that depends on the decision of the company and therefore a shorter warranty period proves nothing.

I think in general products made nowadays tend to live a shorter life than products made in the past. My ex has a VCR ~20 years old functioning properly (maybe not as good as the day they were bought but still displaying picture properly and not eating tapes etc...). Todays consumer electronics are made one thing in mind and that's cheap price.

EDIT: I just read the Intel article and just have to say this ! Intel SSD is not obviously not for me. I don't know about other products but the fastest SSD in the market falling to it's knees is kinda disappointing. Have to wait a couple of months or more before this tech is usable imo.

[Zorlac]

My understanding is that the problems will be fixed with a new file system optimized for SSD (no more NTFS) as well as a new defrag tool (Intel supposedly is working on one).

[kiikkuja]

My understanding is that the problems will be fixed with a new file system optimized for SSD (no more NTFS) as well as a new defrag tool (Intel supposedly is working on one).

I believe so too but there isn't a solution NOW so i'm going to wait for new generations of software and disks before I do anything.

[Helmore]

My understanding is that the problems will be fixed with a new file system optimized for SSD (no more NTFS) as well as a new defrag tool (Intel supposedly is working on one).

You mean like this:

All is not lost, however, as the ATA spec is being updated to include special commands such as “TRIM”, “DISCARD”, and “UNMAP” (a SCSI command). The new protocol lingo will let the Operating System tell the SSD when areas are no longer in use, such as when files are deleted. This will speed up the process of writing data to flash blocks no longer containing valid data, as the wear leveling routine doesn’t have to play musical chairs with data that is no longer relevant.

Windows 7 will support some variation of these commands, and firmware flashable drives like the X25-M should have the ability to be brought up to speed as well. This will not completely solve the problem – It falls short on some RAID configurations (i.e. RAID 5), since all data must remain ‘valid’ for parity calculations to work properly in the case of a drive failure. Data recovery also becomes more complicated, since deleted files could be overwritten by the wear leveling routine even if the OS did not specifically write over the addresses where those files were originally stored.

Microsoft has stated that there will be optimizations for SSDs in Windows 7 at PDC 2008 (last November), where they talked about such commands and how they will implement them.

[STaRGaZeR]

Yes, I was wrong as I stated other SSDs don't have that problem, I should've written: "not that extreme and quickly".

They do have, how much is something we don't know as nobody has tested them with this kind of metodology yet. Point is, all current SSDs are full of quirks that I'm sure will keep you away from them after being some time with them. And this article is just the best so far: in X time your drive will become virtually unusable, and the only fix is to "reset it to factory defaults" after having to backup all your data in another place. This is 100% unacceptable. It doesn't matter how good the write methods are if the drive craps out every 2 months.

Speed is nice, but I just can't sacrifice security for it.

[Boogerlad]

will those changes be implemented to sata3?

[shabby]

Looks like the ocz summit drive is the real winner here, slightly slower in peak speeds but much faster than the vertex in random writes and iops, and even besting the intel ssd in some of those tests too.

[EniGmA1987]

The writes are not much worse thn the Random Writes on a Velociraptor 300GB... so it isnt THAT bad. Plus everyone knows that MLC has bad random write speed, the issue is whether the stuttering is gone. So even if it drops to 20MB/s, if the stuttering is gone then it is a good SSD for the money.

[alfaunits]

Hold on... if you pay for a 120GB SSD double/triple that of a 300GB VR, you DON'T expect it to shine in just about every aspect vs. VR instead of just a tad of shine in some?
LOL. You're the customer most companies would kill for then - too bad that market is small.

[mibo]

Hold on... if you pay for a 120GB SSD double/triple that of a 300GB VR, you DON'T expect it to shine in just about every aspect vs. VR instead of just a tad of shine in some?
LOL. You're the customer most companies would kill for then - too bad that market is small.

On a common workstation reads dominate writes - so it doesn't matter if the raptor writes faster if the ssd is just good enougth.
Furthermore there are notebooks, power consumption and silence recommending the ssd ;-)

[LowRun]

On a common workstation reads dominate writes - so it doesn't matter if the raptor writes faster if the ssd is just good enougth.
Furthermore there are notebooks, power consumption and silence recommending the ssd ;-)

Been watching SSDs to use in my laptop but at the price they sell i want them to be flawless else they have to be cheaper than HDDs considering the capacities.

[FischOderAal]

[...]

Problem is, we don't have a comparison yet. I really miss a SLC drive (without write combining) for comparison in that test.

For instance if you have a light usage and it takes two years before it comes to that point your drive is "virtually unusable" I don't see a big problem at all. But it indeed sucks if it already appears when all you did was installing the OS...

It's not all black and white. Just because one drive fails miserably in that aspect, I still don't see any reason not to go for a SLC drive without write combining...

It happens I made benchmarks on the day I installed my SSD.
http://www.forumdeluxx.de/forum/show...3&postcount=67

I've used it 10 days now and I'd say I wrote and deleted 15-20 GB after I cloned the OS. (Editing videos for my parents...)

time dd if=/dev/zero of=/Volumes/Mobi/testfile bs=1024k count=4096
4096+0 records in
4096+0 records out
4294967296 bytes transferred in 65.769871 secs (65302961 bytes/sec)

real 1m5.805s
user 0m0.041s
sys 0m16.106s

So, no big difference there (yet?). The difference could be caused by inaccurate readings.

[STaRGaZeR]

Oh, you got a SLC based drive, that explains everything :P

Since you're an usual poster keep us updated about this issue if you can. Would be interesting to see how the Mtrons are doing here.

I tried an OCZ Core series, and you know what? Never, and I mean never ever again until these drives are: a) at a sane €/GB ratio, b) bigger, and c) reliable, with all those lame issues resolved. People that thinked the Intel drive was some kind of god between SSDs (me included) should revise their opinion about them now. Tony said in the OCZ Vertex or Apex thread that the Intel drive was slow like the others when the drive is almost filled, and while that's not exactly the reason as proved in the article it's in fact true that it's no better than any other MLC SSD in the end.

[rge]

I bought the intel X25m when it first came out and am very happy with it. Mine has 50Gb used on the 80gb drive, and write speeds including random are still the same as when I bought it ~80mb/s or same speed as my perpendicular drive, and read speed 250mb/s with .1 random access times which blows away my hard drive. Basically you would never convince me to go back to a hard drive with or without raid versus this SSD.

I have played around with my WD perpendicular hard drive versus intel SSD. Despite supposed write speeds being same, I can install windows as well as games roughly 20% faster, games have installed more than 2x on both drives testing. Games load 2x as fast on SSD. Photoimpact which takes 6 seconds to load from WD7500AAKS (perpend drive) is instantly loaded from SSD.

Regarding having to "reset" the drive. Whoever wrote the review article obviously does not have one or spent a few months using one. The controller simply adapts to what you are doing automatically. Once you learn how it adapts you can "play" with the benchmarks while the disk is loaded, and get it to do pretty much what you want.
read my post here
http://www.xtremesystems.org/forums/...7&postcount=35

No question I will be happier when write speeds get to 250mb/s like read. But I cant tolerate using my HDD for anything other than storage at this point. Regarding longetivity, I could care less, the technology in my X25m drive is going to be obsolete in a year, I would upgrade by then anyways to something likely larger, faster, and cheaper.

[FischOderAal]

Oh, you got a SLC based drive, that explains everything :P

I didn't want to live with stutter/lags and I wanted something I could use for a long time. That's why I payed the quite hefty price. The 2,5-inch HDD was effing slow

Since you're an usual poster keep us updated about this issue if you can. Would be interesting to see how the Mtrons are doing here.

Sure will do (If I don't forget it )

I tried an OCZ Core series, and you know what? Never, and I mean never ever again until these drives are: a) at a sane €/GB ratio, b) bigger, and c) reliable, with all those lame issues resolved.

Yeah SLC drives are very expensive (about 3 times worse €/GB-ratio than MLC) but at least you don't have to mess without lags/stutter and so on. Interesting thing is, the biggest development takes part with the MLC-drives, yet they have more disadvantages than SLC-drives imho.

Both SLC and MLC are not ready for the mass market, they are either too expensive or have too many issues. Normally I'm not an early adopter, I used to shake my head when some bought the newest (and most expensive) hardware. But I was willing to pay the price because the HDD really annoyed me that much and it solved my issues. But I'm using a laptop for my everyday work and not a PC with relatively fast 3,5-inch drives. Using my PC the 3,5-inch-HDDs were sometimes a noticeable bottleneck but nowhere as annoying. If I didn't use a laptop I'm not sure if I would've bought a SSD already, seeing the issues and the not yet ready OSs.

Now I can understand your position towards SSDs much better and I hope you understand my enthusiasm. I guess I would've been as disappointed as well if I had bought a "cheap" MLC with jMicron-controller.

People that thinked the Intel drive was some kind of god between SSDs (me included) should revise their opinion about them now. Tony said in the OCZ Vertex or Apex thread that the Intel drive was slow like the others when the drive is almost filled, and while that's not exactly the reason as proved in the article it's in fact true that it's no better than any other MLC SSD in the end.

Even Intel can't take away the physical disadvantages of MLC-cells. And yes, filling the drive to the last bit is a problem as well. That's why I went for a 64 GB drive, even though a 32 GB drive would've been just enough after deleting my music, pictures and other stuff. Now I paid more, but I can keep my music on my Notebook xD Sometimes it sucks to be a music addict

[Levish]

Hold on... if you pay for a 120GB SSD double/triple that of a 300GB VR, you DON'T expect it to shine in just about every aspect vs. VR instead of just a tad of shine in some?
LOL. You're the customer most companies would kill for then - too bad that market is small.

some people just want something better even if its not 100-200% better in every respect, so long as it is orders of magnitude better in some key respects.

For example, I personally wouldn't hesitate to reccomend a RAID array made up of 4x X25-Es for enterprise use assuming I'm allowed to validate that they work as expected considering they would be 75-80% of the performance gain of some really exotic ramdisk setup, shifting the current IOPs limitations over to Processing / Coding limitations.

[STaRGaZeR]

Yeah SLC drives are very expensive (about 3 times worse €/GB-ratio than MLC) but at least you don't have to mess without lags/stutter and so on. Interesting thing is, the biggest development takes part with the MLC-drives, yet they have more disadvantages than SLC-drives imho.

Both SLC and MLC are not ready for the mass market, they are either too expensive or have too many issues. Normally I'm not an early adopter, I used to shake my head when some bought the newest (and most expensive) hardware. But I was willing to pay the price because the HDD really annoyed me that much and it solved my issues. But I'm using a laptop for my everyday work and not a PC with relatively fast 3,5-inch drives. Using my PC the 3,5-inch-HDDs were sometimes a noticeable bottleneck but nowhere as annoying. If I didn't use a laptop I'm not sure if I would've bought a SSD already, seeing the issues and the not yet ready OSs.

Now I can understand your position towards SSDs much better and I hope you understand my enthusiasm. I guess I would've been as disappointed as well if I had bought a "cheap" MLC with jMicron-controller.

Well I'm really excited like you about SSDs but all of them suck in something like you say, that something being basic. Ones because of perfomance and issues over time, others because of price. The technology is just too inmature right now. I could never justify the price of the Mobi, but if you want to pay the price and you're happy with it then

Even Intel can't take away the physical disadvantages of MLC-cells. And yes, filling the drive to the last bit is a problem as well. That's why I went for a 64 GB drive, even though a 32 GB drive would've been just enough after deleting my music, pictures and other stuff. Now I paid more, but I can keep my music on my Notebook xD Sometimes it sucks to be a music addict

When he said "filled" I understood filled by the user with useful data, but maybe he meant that when the lookup table is so complicated after many cycles of writing and rewriting and the controller spreading them between all the NAND chips that fragmentation reach the level of no return the perfomance just falls to the ground.

[BGVirtual]

Here are some benches of the Mtron Mobi MSD 3500 32GB SLC, scroll down to the bottom for some IOMetter results. Though it doesnt show crearly how much % are random writes, how much % sequential writes and the number of IOs, maybe the default settings are used.

Translated:
http://translate.google.de/translate...p%3Ft%3D185259

Original:
http://forum.coolaler.com/showthread.php?t=185259

[alfaunits]

some people just want something better even if its not 100-200% better in every respect, so long as it is orders of magnitude better in some key respects.

Better in some respects is one thing - but better in some yet worse in others is another thing

[3NZ0]

LOL, this thread has made me feel quite bad, I've lost my faith in ssd after reading it. I wasn't aware that ssd got slower the more you filled it up and the intel ones had a number of quite worrying problems.

Ah well, I have a pair of v raptors and a backup in case everything goes horribly wrong.

Was going to consider a pair of 250gb ssd's to replace the 320gb's in my xps and a single for my nc10 but I think I'll just stick with looking at the new seagate 500gb 7200rpm drives.