backtop


Print 25 comment(s) - last by Fnoob.. on Feb 19 at 9:34 AM

Retailers may go above MSRP due to expected demand

OCZ has stated before that it delayed its Vertex drives due to firmware upgrades. DailyTech has now learned more details on why OCZ held back shipments of its flagship SSD drives.

The latest firmware for OCZ's Vertex series boosts sequential read and write performance, so much so that it can compete with Intel's X-25E Extreme series on performance. However, the Vertex boasts a much greater capacity and a much lower price.

The Vertex series, which uses an Indilinx Barefoot SSD controller, was originally specified at 200MB/s sequential read and 160MB/s sequential write. However, OCZ's internal tests show up to 250MB/s sequential read and 240MB/s sequential write speeds.

These tests were conducted on an empty drive, and will not officially be presented to consumers. However, it gives an indication of how fast the final drives will be and allows some results to be inferred.

While the firmware is responsible for a large part of the improved performance over the introductory specifications, it is only because of the hardware that it is able to work with. The 120GB and 250GB drives have 64MB of DRAM cache and more channels to access its MLC NAND flash, whereas the 30GB and 60GB drives only have 32MB of cache.

It should be noted that OCZ rounds down its capacities for several reasons. Most storage vendors show capacity in decimal format, whereas Windows show capacity in binary format. This means that the reported capacity is lower than what is on the label.

Most SSDs also reserve some capacity for redundancy in case of bad sectors, and also for wear leveling. These reserved areas may occupy up to five percent of an SSD's storage capacity.

Prices are also coming down, as more production of 16Gb and 32Gb NAND flash comes online at smaller process geometries.

OCZ will post official specifications and revised MSRPs when it ships the Vertex series later this month.
   

Model

Read

Write

List Price

Street Price

Intel X-25M 80GB

250

70

 $390

$369

Intel X-25M 160GB

250

70

 $765 

$779

Intel X-25E 32GB

250

170

 $415

$410

Intel X-25E 64GB

250

170

 $795

???

OCZ Apex 60GB

230

160

$199

$199

OCZ Apex 120GB

230

160

$370

$319

OCZ Apex 250GB

230

160

$830

$679

OCZ Vertex 30GB

200

160

$129

???

OCZ Vertex 60GB

200

160

$249

???

OCZ Vertex 120GB

200/240*

160/220*

$469

???

OCZ Vertex 250GB

200/240*

160/220*

$869

???

 *Unofficial specifications, subject to change



Comments     Threshold


This article is over a month old, voting and posting comments is disabled

Performance Degradation over time
By Fnoob on 2/16/2009 9:02:47 AM , Rating: 2
Can't find the article I read yesterday, but apparently the Intel X25-M drives suffer from serious drive fragmentation issues which essentially cut performance in half over time. Running a defrag utility does not help. Reghosting / reformatting the drives does. The issue would seem to affect any / all SSD drives due to the nature of the medium. Performance tests on a formatted drive are stellar, but once the drive has been in use for awhile, I/O drops significantly as it has to basically wipe each cell before it can write to it.

I hope this doesn't apply to these new OCZ drives, as they seem to have specs better than the Intel X25-E - and cheaper.




RE: Performance Degradation over time
By JohnMorconzol on 2/16/2009 9:57:48 AM , Rating: 3
The problem here is that when the files are being written on the SSD it must be written 0's before, the actual file systems just delete a file from the FAT but the data remains, the SSD's need a new file system optimized for it, one that for ex. write 0´s when a file is deleted, it will take a little longer to erase something but the writing after will be just as fast as new. We need a optimized SSD file system and OS too.


RE: Performance Degradation over time
By Fnoob on 2/16/2009 11:20:01 AM , Rating: 2
"We need a optimized SSD file system and OS too"

Agreed. So to summarize, this issue will affect all SSDs regardless of manufacturer. Benchmark the drive when clean - amazing performance. Benchmark after a few weeks of use, performance drops to average HDD specs or below. Not worth the investment at this point.

I was just about to pull the trigger on a pair of X25-E drives to setup in Raid0 for an OS partition. Given the latest report, I'll be going with a pair of 15K SAS drives instead. Cheaper too.


RE: Performance Degradation over time
By Whaaambulance on 2/16/2009 12:00:36 PM , Rating: 1
I was under the impression that windows 7 addresses this issue. Is this true? I am currently running win 7 beta with an Intel 80GB SSD and haven't noticed any particular performance drop.

Does Win 7 have a new file structure to benefit the use of SSD drives?


RE: Performance Degradation over time
By Kougar on 2/16/2009 1:51:05 PM , Rating: 2
Windows 7 is not SSD agnostic unlike Vista and XP, however, it does not address this issue.

The problem won't be solved until the ATA specification is updated to allow SSDs to wipe their LBA tables when a file is deleted. They stated it was in the works, but who knows when this will actually happen...


By Cypherdude1 on 2/18/2009 1:57:31 AM , Rating: 2
Personally, I wouldn't care about SSD except that Microsoft and hardware driver writers don't pay enough attention to Power Management.

I have 3 drives and only the last 2 ever shut down. It would be nice to have an SSD drive for the primary O/S partitions, then my system would become a true multimedia PC. For those of you who think the OCZ Apex 120GB wouldn't be enough for an O/S, my WinXP ProSP2 primary partition is only 8 GB's. Cutting down on the unused Virtual file and moving nonTSR applications to other logical partitions saves much space. At $320, the Apex is a decent price.


RE: Performance Degradation over time
By brentpresley on 2/16/2009 2:28:10 PM , Rating: 3
quote:
Agreed. So to summarize, this issue will affect all SSDs regardless of manufacturer. Benchmark the drive when clean - amazing performance. Benchmark after a few weeks of use, performance drops to average HDD specs or below. Not worth the investment at this point.

I was just about to pull the trigger on a pair of X25-E drives to setup in Raid0 for an OS partition. Given the latest report, I'll be going with a pair of 15K SAS drives instead. Cheaper too.


Sorry, those comments are just pure BS.

The X25-M has been in my main rig for 3 months now. It benches the same today as it did on day 1. It also benches the same empty as it does 90% full.

Very very early firmware revisions on the X25-M had problems similar to that described above, but the non-ES FW versions do not have these problems.


RE: Performance Degradation over time
By Fnoob on 2/16/2009 3:04:07 PM , Rating: 3
Pure BS that reiterates what was published here :

http://www.pcper.com/article.php?aid=669&type=expe...

Perhaps, as the article mentions, your particular usage patterns don't highlight the problem? I am not trying to spread FUD, just hoping to start a discussion on this relatively unpublicized issue. I do not currently use SSDs, but have been trying to do my homework before building a new rig - and this report dims my enthusiasm somewhat for spending insane $$$/GB on tech with potentially serious issues.

The same ~$800+ I would have spent on 2 X25-E 32G SSD drives will instead go towards 2 15K 300G SAS drives that will destroy the Intel drives in Raid0... over 4000MB/s burst!


RE: Performance Degradation over time
By freeagle on 2/16/2009 3:51:11 PM , Rating: 3
quote:
over 4000MB/s burst


Didn't you drop one 0 more into the figure?


RE: Performance Degradation over time
By Fnoob on 2/19/2009 9:34:29 AM , Rating: 2
No.

The actual benchmark spec was 4054.3MB/s - now that is burst, not sustained... but still damn quick.


RE: Performance Degradation over time
By michael67 on 2/17/2009 2:24:22 AM , Rating: 2
Unless you going to do some database were you going to have tons of random writes, noting beats a high performance SSD.

"We need a optimized SSD file system and OS too"

Yes, if this is implemented, it will give SSD's a even greater speed gain over HDD, a specially if this is done when the SSD is idle(*).
But SSD optimization will be implemented in Win7.
(*)idle optimization, data is written on other part of the SSD and later, the removed data blocks will be purged by writing zeros to them, there was something about that on the Win7 blog but properly wouldn't be ready before SP1

About 80-90% of you HD actions are read actions, properly even higher for most games!

I use a X25-E as my OS / C: drive.
2x Transcend 16GB in RAID0 on a *cheap* "Areca ARC-1200" whit 128MB cache HD controller for the games i am playing on the moment.
(The cache mem helps buffering writes, so unless the file is bigger then the controller and HDD cache combined i have no problems whit slow writes)
And a 300GB raptor for my swapfile and the rest of my files/games/programs.
No one beats my setup when it comes to HD speed/game load times.
(at least no one I know ;-) http://www.hardware.info/nl-NL/usersys/cmxxZ51hk2m...

By splitting up my main data streams i win a lot of extra speed, but even a single latest gen SSD will let a 300GB Raptor eat dust by a factor of 3 to 8 in most real-world benchmarks (except random write database benchmarks)

Boat Launch & Storage Mark 2008 Index imho the most importance bench as it resembles real life usages (Google translation from dutch)
http://64.233.183.101/translate_c?hl=en&langpair=a...

About 300 different HD Benchmarks in dutch but the Nr's talk for them self.
http://tweakers.net/benchdb/suite/13

For the gamers under us that wane use a SSD just for games here is all you need to know!
http://tweakers.net/benchdb/suite/33

Here you can see what a good controller can do (Intel ICH7R)
http://tweakers.net/benchdb/test/164

Here you can see what extra cash mem dose 250% gain whit 2GB over 256MB
http://tweakers.net/benchdb/test/153

Using the filter on the pages in the links, will make it easier to do A apples to apples comparison


PS. adding a SSD will have not mouths benefits if you don't have at least 4GB system mem, as you have to use the swap-file way to mouths if you only have 2 ore even less mem in your PC.


By michael67 on 2/17/2009 2:35:21 AM , Rating: 2
By Nihility on 2/17/2009 7:50:24 AM , Rating: 2
I hope they find a better sollution than writing zeros.

If you write zeroes there will be no way to undelete or recover data. Sure, it's a security freak's fantasy but it'll be hell on the regular folks who keep accidentally deleting their p...important files.


By Kougar on 2/16/2009 2:27:31 PM , Rating: 3
Here it is: http://www.pcper.com/article.php?aid=669&type=expe... Very excellent article.

Technically it applies to all SSD drives. The second part of the problem is that for an SSD even after a file is deleted the information remains stored in the LBA table and the drive treats it as valad data. Now combine that with what you were mentioning, having to move data out of a 512K sector before it is able to erase then write data to blocks within that sector... the analogy they used of musical chairs is very appropriate, except that most of the LBA data being juggled around pertains to already deleted files...

I'm surprised none of the major sites made reference to any of this. Only a few even hinted the drives would slow as they were filled, but not why nor with this severity.


sequential read/write performance
By irev210 on 2/16/2009 8:41:18 AM , Rating: 2
While sequential read/write performance is nifty, I think too many people are getting caught up into those #'s. How a controller performs overall is what matters. Random streaming of 4K blocks, for example, is much more demanding on the controller. I/O throughput + the ability for the controller to handle smaller blocks efficiently poses a much greater challenge to SSD makers.

This is where the Intel controller happens to shine. While it may not have the fastest sequential read/write performance, it is a very well-rounded controller that has best-in-class I/O performance, random block read performance, etc. Which means in real world usage, Intel still bests much of the competition, even SSD's sporting higher theoretical xfer limits.

I can't wait for a full review on the Vertex and how much additional power consumption the 32/64mb of cache will use, along with the rest of the performance aspects of this SSD. We can have a great alternative to the expensive X25-E.




RE: sequential read/write performance
By shabby on 2/16/2009 9:12:48 AM , Rating: 3
You should read this review that just came out, i was shocked how much the intel drive slows down over time.

http://www.pcper.com/article.php?aid=669&type=expe...


RE: sequential read/write performance
By Fnoob on 2/16/2009 9:23:53 AM , Rating: 2
Thats the one I was looking for, thanks.


RE: sequential read/write performance
By Khato on 2/16/2009 2:38:59 PM , Rating: 2
So, am I missing it, or did they not disclose % of disk space utilized anywhere in that 'report'?

It's well known that entirely filling up an SSD is a very bad idea, and I wouldn't be surprised if that's the sort of 'torture test' used in order to get the drives performing as reported in the first place. Now there's no question that ideally this wouldn't happen, it's hard to avoid all of the inherent flaws of MLC flash... By vastly increasing most all the important aspects, the Intel SSD gets itself into trouble when all of its storage capacity is used.

Anyway, it's not all that difficult to keep the drive performing like new, just leave it a decent buffer of free space and turn off the idiotic windows page file.


Samsung SSD?
By Mk4ever on 2/16/2009 9:14:09 AM , Rating: 2
What happened to the Samsung SSD? the one with 220MB/s sequential reading and 200Mb/s sequential writing? I have been waiting for news about it for months now.

On another topic, why don't I see any SSD reveiws on Anandtech? the last one was Intel's SSD. Still no reveiw for G.Skill Titan or the newly launched OCZ Apex. What's that all about?




RE: Samsung SSD?
By Jansen (blog) on 2/16/2009 11:50:34 AM , Rating: 3
Samsung SSD's are sold out to OEMS. I talked to them last month, they said don't expect any in retail for a while.

Anand has a bunch of SSDs and he is going to do an article soon. I know OCZ just sent him a Vertex with the new firmware. I expect that it will be comprehensive, as always.


RE: Samsung SSD?
By Mk4ever on 2/16/2009 4:22:00 PM , Rating: 2
Thank you very much for your response.

By the way, do you have any information about these? or if they will be included in anandtech's upcoming review?
http://www.techpowerup.com/index.php?85395


Evolving nicely
By psychobriggsy on 2/16/2009 8:37:33 AM , Rating: 1
"production of 16Mb and 32Mb NAND flash"

Do you mean Gb (gigabit)?

It looks like this technology is evolving nicely. Far nicer read/write times, price is getting acceptable, hopefully it doesn't use JMicron, although I believe they've tweaked their controller to reduce problems anyway even if it does.

Looking forward to seeing a review of these.




RE: Evolving nicely
By FishTankX on 2/16/2009 9:51:21 AM , Rating: 2
When you read the article, it actually says that it's using an indilinx barefoot controller. ;)


wow + vista
By Alphafox78 on 2/16/2009 12:40:17 PM , Rating: 2
As long as I can fit WoW and Vista on the same disk I will be all set. What, computers can be used for other tasks you say?!




Drools
By siberus on 2/16/2009 2:34:50 PM , Rating: 2
............I want one for my laptop :o




"Young lady, in this house we obey the laws of thermodynamics!" -- Homer Simpson

Related Articles













botimage
Copyright 2014 DailyTech LLC. - RSS Feed | Advertise | About Us | Ethics | FAQ | Terms, Conditions & Privacy Information | Kristopher Kubicki