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

We released a product that at the moment of release was the best technology we could offer on that price point...now we have a new drive coming which will cope better with the random write issue windows throws up.

You guys need to think of this....the drive is working fine when you use steadystate...the reason why is random writes are forced to sequential writes. So if you want cheap SSD you either have to work around OS limitations with tweaks or apply steadystate. We had an end users come to the OCZ forum using an Intel SSD and even he was complaining about stuttering after a few days use on his near full SSD...so even they are not immune

SSD has caused M$ to rethink its code...that pretty much tells me they know present OS's are an issue

 

We have a feeling TrueCrypt also forces random writes to sequential...did you ever try the Core the encrypted?

 

hm.. how about the vista disk encription on business and ultimate? could that help, too?

 

Sounds like he did Tony. Very interested to see if this works though as Truecrypt would be a lot neater solution. Perfect time to buy a cheap SSD.

 

Do you mean Bitlocker?

 

yes

(filling the 10 character minimum again)

 

Just thinking about this some more, why hasn't someone at OCZ or JMicron or any of the SSD manufacturers written a driver which forces sequential disk writes? Why is it taking the discovery of a third party tool to do it if it increases performance not only of JMicron SSDs but all SSDs?

 

That's a good point, especially since most computers sold now days come with 3+GB of RAM making a huge dumping ground for the SSD to use as a temp cache so the writes can be made sequentially... and basically in the back ground so no one would notice.

With Win7 coming out and the next gen of laptops looking like they are going to be hitting the 8GB RAM levels soon it seems like a good time to jump on something like that.

 

btw, it's not a third party tool.. steadystate is from microsoft.. it's a kiosk software.

anyone with ocz disk around that could test bitlocker?


problem with temp-caches is the issue of dataloss. i'd prefer to have a small cache that is batterysave based on NAND memory (more expensive but no small write issues afaik?) and the rest cheap NOR mlc.. or something like that.

then again, I can't wait for memristor based ssd's. (hp is making big progresses with memristors, that's really cool to know)

 

Ok but you know what I mean - finding a piece of software which wasn't specifically written to fix this problem, and in this case SteadState doesn't fix 64-bit anyway. If it is possible to fix as SteadyState has proven why hasn't this been done already via a driver release for all the existing very frustrated OCZ users.

 

because steadystate is not fully hideable from the user. they have to know it runs and they have to know how it works to be aware of it's different behaviour.

but yes, an extended caching option which could result in data loss on powerdown should be made available as a driver.

still, all those fixes can not be made in silent. users have to know the risks, no matter how small.

they can only be made in silent if the disk(system) takes care of it itself. f.e. raid controllers with onboard ram that also have a battery to be able to flush their data on disk in the moment of a power loss. it can not be made silent in drivers.

then again, windows nowadays have options for enabling caches that may result in data loss on powerdown. they should/could plug in a driver there that then changes the write behaviour.

another problem of this write change the way steady state does it: it changes storage size behaviour. if the filesystem doesn't get changed, i think it's not possible to implemtent that write behaviour without requiring more storage space (steady state 'solves' is with a huge cache file at the start ). so not silent there, too.

and in the end, a simple plug and play solution is, again, to solve it on the drive level (or make an inbetween thing you can plug between drive and system then again, that wouldn't fit into the tiny notebook space..).

but to have a nice solution that really works, it has to be done on disk side, too. changing the os, changing the drivers, etc is all great to better fit the ssd's. but ssd's that don't fit into the current reality are still worse than those that do fit.

i prefer to pay, in case of the 30gb example, around 170$ more to get an mtron version that works without issues (and then can still gain from all the tweaks and tricks you'll find for the ocz ones as well as future win7 ssd development).

a bad disk can only get fixed so far. a good disk can use the same fixes, too

end result: if you want high end, you'll have to pay more.. that, uhm, is .. not new

 

They have very quick reads. They can be very useful. I'm using them for a RAID hosting company quickbooks files. They are several times faster than the conventional hard drives we were using before.

I think people expecting them to perform equivalently to drives that cost hundreds more are being unrealistic. ok, maybe they are not appropriate for system drives. Still, they have uses.

 

Expectations were high with Core, people forgot that the bulk of drives they were being compared to were SLC so when the pitfalls came to light they felt they had been let down.

At the controller level a few things can be done to work round the issues windows throws up...and yes we have a new drive near ready for release that should work just fine...it features a large onboard cache and faster controller than the JMicron found on the V2 core.

Im trying to get a good contact within M$ to talk about a driver, if you look seriously at what windows does with random writes and what a low level driver could do you would see its well worth M$ having a look at the issue as it would enhance an otherwise bashed OS (Vista).

Also remember MFT have a patent covering some of the driver technology, they are working on a version that will run on the OS partition, which again will bring blinding speed to cheap drives.

 

Interesting. Have EasyCo given any hints as to release date? Also I've noticed this fix (slipstreaming chipset drivers) on your forums Tony which seemed to work really well for a few users but I see it's not mentioned in any of the stickies. Has it been investigated any further?

 

This.


(blah blah blah 10 chars)

 

mainly for the vista tweaks we do recomend that the new drivers are either slipped into vista or installed F6 if possible.

Im going to be honest here, i installed vista64 to a v2 core 3 days ago...board is a test P6T with i7 965 that I have to test all our ram on. Before the tweaks all i had was slipped ICH10 drivers and Hibernation turned off...it took forever to install the ati vid card driver so i canceled and installed all the tweaks i outline on the OCZ forum. I then went back to the ATI driver which installed in under 1 min or so...very very different to the first attempt where i canceled 10mins in.

So as you see, if you are willing to take the extra 30 mins and install the tweaks you end up with a fast system even without Steadystate...but you have to be bothered.

We have started work on an automated script install which should apply most if not all the Vista64 tweaks...for those who may struggle or are a little lazy

I use an OCZ 15"DIY compal Laptop here with a 60gb V1 core. I have used Vista64 with the tweaks and 4GB ram and it was fast, no lags or stutter. I then moved to XP32 and did the same tweaks just to see what it would be like, again no issues once i had set the OS up. now im using vista 32bit, to start fully OS tweaked and again it was fast and not at all laggy...then i moved to steady state which took what was already fast to another level

Steadystate is a work around for this issue, we know this...but overall the gains outway the pitfalls....for me anyway.

 

Alright, so it writes to the cache as sequential, but when you commit the changes to disk, it's gonna make random writes by small changes to all the files that were cached?

 

in a bad situation yes (don't want to know how long the commit would take.. i'd guess copying everything to an ordinary disk, commiting there, and copying back would be faster ).

in a good situation it does sort of a live defragmentation pass to collect and write big junks (at best, non fragmented files, too.. which would help on ordinary disks then).

my personal experience from steadystate: it's quite slow. but that's on an ordinary disk, which has completely different behaviour (espencially the latency due to the "move to original data, move to cached data" is really feelable).

currently, i have updated my two mtron 64gb. the new firmware is on, which should make my slow-bios bug a thing of the past (my bios is still slow on the pc but that's hp's fault. with or without ssd sadly. but now it should not have a 1 min delay due to the two ssd's.. ).

i'm restoring the system from my home server right now and post results.

thanks for mtron for the quick support, and the detailed instructions including all required files for the update. now i know i've spent my money well. even if the company is no big name, for me, they get bigger everyday

 

Its actually quite fast as long as you reboot to renew the session once a day.

File copy in session as fast as MFT, which is up there with Intels drive if not faster, boot time with 10GB cache and 7 hrs use was 40secs today.

Those with mtron drives why don't you give it a try...you may be surprised just how much extra speed you gain.

 

I have considered using this solution however there are some major disadvantages. I rarely reboot my computer. it is mostly in sleep mode during transit, and my experience with windows has been to expect some sort of crash or other once a week on average. In between which I probably will not have rebooted or written out the cache.

This solution result in an increase of performance at the significant risk of data loss (atleast for my laptop usage).

It may be a good solution for desktop users or users who shut down or reboot their laptops before any significant crash events occur. However my opinion of the OCZ SSDs is very low. MLC drives with much better performance can be had for similar prices.

 

What MLC-drives are you referring to?

 

I run my laptop 24/7 and it very rarely crashes. I continually install and run a whole multitude of programs and games. Yours crashes once a week? Jees, I suggest you do a rebuild and update all your drivers. Or buy a new laptop.

Tell me what other 128GB MLC drives come anywhere near $195?

 

the super talents are (here in switzerland) at around the same pricing as the ocz. don't know the good/bad ratio between the two..

oh, and yes, my laptops and desktops never actually crash. _never_.

 

ok, i now updated my mtrons to the newest firmware, reloaded the raid0, and fiddled around on my system. now i get a full boot into msn, google talk, with anydvd, homeserverconnector and the audio8dj control in 33 seconds, from turning on the pc.

the bios itself is now much faster, too, as i took away a lot of the things like firewire, cardreaders and similar. so vista boots now in about 20sec by itself..

i'm happy with my mtrons.. next project: pimping winhomeserver (cheap mlc drives for storage for fast readback in an emergency ) and the lan.. 100mb/s still, the biggest bottleneck here now definitely.

 

hey guys, I'm completely new to this HDD business for laptops and I need some advice.

I'm planning on purchasing an OCZ Core V2 120GB SSD for around $520. Are they any good? Any issues? Should It be used as a second or primary drive?

Thanks.

 

The 128 GB V1 is available for $200 and is basically the same drive (I do not think it has a built in USB port though).

The Core drives have very high sequential read and write rates but very low random write rates. As long as you realize those implications and have a strategy for managing that, either through hardware RAID (not really an option for laptops) or using SteadyState, you can use it as a primary drive.

Check out the OCZ forum for details.

If you do not want to use SteadyState (or can't) then I would think twice before buying it.

 

if you have the choise for a secondary drive in your notebook, i'd guess you're better of with an mtron 32gb + an ocz version 1 for together around the same price. that way your system is very fast and you have quite some storage. else, i'd personally go for the 64gb mtron.. or the 128gb samsung, but they're more pricy.

 

I think that $520 is Australian so about $340 US - not bad. The v2's are selling for about $290 in the US after rebates.

If you can find a v1 cheaper though I would go for one of those. They are only very slightly slower on the reads and I believe some also have a USB port now. The only reason to avoid v1's is if you are going to be running them in RAID as some users have experienced issues.

 

You could debate if you really need 120GB since you also have a second hard drive.

Maybe it would be better to buy a OCZ 60GB and a Seagate 7200.3 320GB as your second hard drive (or a 500GB WD). It may work out faster and cheaper.

I agree with iaTa here, sounds like something's wrong. I'm running XP SP3 and it crashes once a year.

 

Just so you guys know the facts about V1 and V2 drives:

V2 has a die shrunk faster clock version of the JMicron controller, it also has a USB port and the ability to be firmware upgraded.

It also has 4GB more flash per drive over the V1 to account for wear leveling.

V1 =64GB, 4GB wear leveling shows 57GB in windows

V2 =68GB, 4GB wear leveling shows 60GB in windows

 

thanks for all the info. now we wait for V3 news

 

I'd like to get an SSD for my Alienware Laptop a 2007 model M5750, I currently have 2 Seagate 7200 Momentus drives installed, this 128GB drive looks perfect and the price $269 seems very reasonable. If I Ghost my current Bootable drive over to this SDD while installed as D: drive and then swap drives will the lappy boot up off this drive. I ask because a user on the NewEgg site bought the Intel SSD and installed it in his Lenovo T60 ThinkPad and it would boot. He had to install the drive in his desktop not his original plan.

Any thoughts would be appreciated - thanks

Here's the drive I like:
http://www.newegg.com/Product/Product.aspx?Item=N82E16820231221

Here's the Intel drive link - read first comment below link:
http://www.newegg.com/Product/Product.aspx?Item=N82E16820167005

Heres the review forum the foum on my Alienware laptop for any details:
http://www.notebookreview.com/default.asp?newsID=3206

 

Sweet benchmarks thank you!!

 

Well yes, admittedly Microsoft is pretty far behind the curve in filesystem technology. Log-based filesystems have been around for decades, and there's already a bunch to choose from on Linux.

This technology was important 20 years ago, before anyone had thought about flash-based SSDs, simply because it was already known that disk seek time was a performance killer. Think about it - you get a disk platter spinning at 7200rpm and you read full-bore off it, sequentially, and you're going to get pretty respectable performance. So, this whole turning-random-writes-into-sequential thing has been around for ages. As usual, Microsoft is at least 20 years behind the state of the art.

 

No, writing sequentially probably means reading randomly anyways.

 

Heres to hoping Win7 will be able to make mostly sequential writes, that would be a pretty big boost in performance and would alleviate the need for antiquated defragmentation software

 

The thing is, the V2 in AUS is only 20$ more than the V1 of similar capacity.

 

I was reading somewhere that because Windows 7 is due to be released so soon there is no hope for anything brilliant coming from Microsoft. They have announced the changes and improvements they are making for SSD drives and something as major as sequential writing is not on the agenda unfortunately. I'll try and find the article...

EDIT: Here we go - not a very impressive list.

 

No brainer then, go for the v2.

 

So i only have 2 questions. Did steady state solve the problem with the stuttering? If so, then all the MLC drives are pretty hot deals.

Secondly, if a SSD ever runs out of the writes in the drive, can your computer still read the data that's in it?

 

as long as the data was stored successfully one time, the drive can read it.

When the drive is about done, it can not save data anymore - errors would occure when it tries to write something.

 

A drive and a cell will not magically stop erasing at exactly 10,000 writes or whatever the figure is.

10,000 is the expected life time; not the actual.

Best,

 

http://www.nordichardware.com/news,8412.html

Intel/Micron -> 34nm, in early 2009... should make things cheaper.

 

does the new ocz solid series ssd still use the jmicron controller? are the any reviews of this drive yet? why isn't it recommended for older laptops?

 

I find the TrueCrypt info on the previous thread pretty interesting. If I'm understanding that correctly, TrueCrypting an entire drive forces sequential writes, which fixes the performance issues on the OCZSSD2-2C30G drive found here?

Or am I not understanding this.

 

I am not sure if you are referring to my post

http://forum.notebookreview.com/showpost.php?p=3993529&postcount=1222

In the Truecrypt forum (you have to be a registered member to access it), I found a post (see my previous post), where a user was claiming something to that effect.

Searching for Truecrypt in the OCZ forum reveals that it has been suggested to be tested as a possible solution but I could not find any results or recommendations as such.

FYI, Truecrypt did not increase random write performance on my Samsung SLC SSD in any notable way.

 

jketzetera: in comparison, did steady state improve it?

 

Still no reviews of the MOBI 3500?


Excellent review of the MOBI 3000 in RAID:
http://www.boards.ie/vbulletin/showthread.php?t=2055300504&page=1

 

i stated before how the mobi 3500 performs. the 2.5" is, based on that, perfectly predictable. i now run them in raid0 so i can't give you anything else without messing with my system.. :s

what was it again? 50-60iops in worst case (4k random write). else 100mb read/write. no stuttering or anything happened yet.

 

I have a few 3500's on the way (16GB for a Linux box and 32GB for a netbook.)

The 16GB SSD will be replacing a PATA Samsung 4GB SSD ... which was roughly the same price 18 months ago

Dave, thanks for the info on the 3500. This has been quite an informative thread.