SSD random write slowdowns & lockups: The finger finally pointed at the true culprit?

[Biker]

...and it's not a MLC issue!

It is looking like the way each SSD drives onboard JM602 controller interacts with the incumbent OS is responsible.

Full details embedded in Anandtech's X25-M review (http://www.anandtech.com/cpuchipsets/intel/showdoc.aspx?i=3403)

A very interesting read.

Can we get a chorus of "Jmicron controllers are crap", anyone? :D

Update:

The latest advice can be summarised thus:

1. Single or multiple drives using motherboard (onboard) controller for OS = dont buy.

2. Single or multiple drives for non OS drive using onboard controller for fast loads / sequential writes = buy.

3. Single or multiple drives with dedicated Raid controller with onboard cache used for any purpose = buy.

The key to making these drives sing is to pair them with a dedicated Raid controller with a decent amount of onboard cache.

A recommended controller would be PCI-E 4x with at least 128mb cache, with 256 mb being ideal.

The Raid controller cache acts as a buffer to smooth out the random writes and ease the JM602 stuttering many users have experienced.

[K404]

Can we get a chorus of "Jmicron controllers are crap", anyone? :D


LOL that aint really news!

From the IDE detect times on new mobos, to the friggin insanely bad performance and reliability on 965 and some P35, JM controllers= fail in my book

ICH chipsets are more widely used than JM though arent they- How many people are using SSD + JM?

[Biker]

Clue: all the Samsung based MLC SSD's are using JM onboard controllers....

Not that that was a hint or anything.... ;)

[speedfreak86]

LOL that aint really news!

From the IDE detect times on new mobos, to the friggin insanely bad performance and reliability on 965 and some P35, JM controllers= fail in my book

ICH chipsets are more widely used than JM though arent they- How many people are using SSD + JM?

The SSD's have the JM controller on board. The article actually suggests it is the controller's lack of cache.

[Biker]

The SSD's have the JM controller on board. The article actually suggests it is the controller's lack of cache.

Bah.

I was at least hoping that a few more would actually read the evidence..

:D

[Praz]

The SSD's have the JM controller on board. The article actually suggests it is the controller's lack of cache.
The Intel controller has a wider write pipeline and the minimum data that has to be written for every write command is decreased by approximately a factor of 8. Nothing to do with a lack of cache.

[Shintai]

Jmicron always sucked. Not a surprise.

[sonofander]

I've always hated jmicron. They're implementation of e-sata is just awful!

Anybody remember silicon image? Whatever happened to those guys?

[speedfreak86]

The Intel controller has a wider write pipeline and the minimum data that has to be written for every write command is decreased by approximately a factor of 8. Nothing to do with a lack of cache.

Could be true, but I did just say what the article suggested.

That being said, the root of the problem is still unknown. My first thought was that it was because the MLC drives had no DRAM buffer, and if you'll notice, Intel's MLC drive does have a DRAM buffer. I asked Intel about this and it turns out that the DRAM on the Intel drive isn't used for user data because of the risk of data loss, instead it is used as memory by the Intel SATA/flash controller for deciding exactly where to write data (I'm assuming for the wear leveling/reliability algorithms). Despite the presence of the external DRAM, both the Intel controller and the JMicron rely on internal buffers to cache accesses to the SSD.Finding good data on the JMicron JMF602 controller is nearly impossible, but from what I've heard it's got 16KB of on-chip memory for read/write requests. By comparison, Intel's controller has a 256KB SRAM on-die. And I'm going to go out on a limb and assume that given Intel's experience with CPU caches, that its SRAM implementation is probably very well done.

[tekjunkie]

Looking at some good results that people have got here with the OCZ core 2 series SSD(which uses jmicron controllers), I would say the caching does matter. Put these cheap MLC based SSDs on good raid controllers(Areca) you got yourself a screaming PC. I myself am planning to put4 of the core 2 60GB bad boyz in Raid 0 on an Areca and enjoy the silence and the performance.

[Buckeye]

I myself am planning to put4 of the core 2 60GB bad boyz in Raid 0 on an Areca and enjoy the silence and the performance.

Oh you will enjoy it ! :clap:

[Cooper]

8x 32GB MTRON Pro SSD in Raid 0 for OS

jesus...have ever waited for any file/app to open on this array :eek:

[Biker]

The Intel controller has a wider write pipeline and the minimum data that has to be written for every write command is decreased by approximately a factor of 8. Nothing to do with a lack of cache.

Are you saying that the differences in performance are purely down to these 2 factors and have nothing to do with the design of the JMF602?

Edit:

Interesting Anandtech review follow up thread here (http://www.ocztechnologyforum.com/forum/showthread.php?t=42769)on the OCZ forums, hopefully we'll all get some definitive answers soon!

[Buckeye]

jesus...have ever waited for any file/app to open on this array :eek:

HAHA acutally I do not wait for much with it.

I went back to my bench rig with 2x 74gb Raptors in Raid 0 and it feels so slow and makes so much noise. This is the same setup I used before I installed the SSD Raid, it's OMG so slow I can hardly stand it.

[Biker]

HAHA acutally I do not wait for much with it.

I went back to my bench rig with 2x 74gb Raptors in Raid 0 and it feels so slow and makes so much noise. This is the same setup I used before I installed the SSD Raid, it's OMG so slow I can hardly stand it.

lol

Show off.

I am not at all jealous.

Nope.

Not even a little.

Honest.

No, really.

:D

[dinos22]

Good article by Anad
this certainly opens my eyes about latencies and it's actually very very surprising to see this

[Praz]

Are you saying that the differences in performance are purely down to these 2 factors and have nothing to do with the design of the JMF602?
Those two factors are the design of the controller.

[Biker]

Those two factors are the design of the controller.

lol.

Forgive me i'm sleepy :D

So JM602 (relative to X25-M controller) = narrow pipeline + minimum write per write command x8 = rubbish random write performance.

In that case I sincerely hope they (Samsung) splice in a new controller soon, and preferably one with the ability to convert random writes into sequential ones on the fly which hopefully would go a long way to eliminate the problems many are experiencing ;)

[Praz]

lol.

Forgive me i'm sleepy :D

So JM602 (relative to X25-M controller) = narrow pipeline + minimum write per write command x8 = rubbish write performance.

In that case I sincerely hope they (Samsung) splice in a new controller soon, and preferably one with the ability to convert random writes into sequential ones on the fly which hopefully would go a long way to eliminate the problems many are experiencing ;)
Yeah, the random writes are the issue. There is software available that does exactly what you describe. However it won't be usable for a boot drive until probably next summer.

One thing most of us are forgetting is the cost of these drives. After rebate most of the non-Intel 64GB MLC SSDs can be had for about $200.00. The Intel looks to be at $600.00. In my mind for three times the cost we should be receiving nothing less then this type of performance.

[ak_47_boy]

Data should be striped across 8 NAND chips. That would be fast. Slackers need to start making controllers.

[BulldogPO]

Is there big performance difference with Areca and ICH10R when using two SSD's in RAID 0?

[tiro_uspsss]

Anybody remember silicon image? Whatever happened to those guys?

yes! i do! SI were all 'the rage' back in the s939 days... & I cant seem to remember no where near as many problems with them as with JM... I wants SI back! (or @ least JM out of the pic, & something decent in!)

[Biker]

Is there big performance difference with Areca and ICH10R when using two SSD's in RAID 0?
__________________


I'd say not a great deal as long as everything plays nice in both scenarios.

Raid 0 does not incur the huge workload overheads of say Raid 5 where the Areca would pull well past the onboard solution.

Having said that I do not have any Areca's or SSD's lying around right now to test with ;)

I do however have an Adaptec 5805 SAS, a LSI Megaraid 8708ELP SAS and an Adaptec 1230SA SATA host raid just sitting on the test bench just waiting for the right SSD deal to come along :)

[Biker]

yes! i do! SI were all 'the rage' back in the s939 days... & I cant seem to remember no where near as many problems with them as with JM... I wants SI back! (or @ least JM out of the pic, & something decent in!)

I've still got 2 AN8-SLI deluxes in service here - both have had Silicon Image arrays running (... for the last 4 years)...

Never had any SI problems ever

Where did SI disappear to (from every board I have owned since.....)?

Did they just price themselves out of the market or were their chips far more prevalent in AMD boards than Intel ones..?

[jcool]

Silicon Image just sucked as bad as jmicron does now. I lost countless raid 0 on SI controllers because I dared to update the firmware, or just because the controller shot itself while iding Windows. I don't miss them a bit.
Of course, jmicron isn't any better.

[bing]

Actually this is a really bad news, someone have to deal with those pile of Jmicron infected SSD in the warehouse ! :)

Beware of incoming cheap SSD in the market... hahaha.

[tekjunkie]

I just ordered an ARC-1231ML-2G and 4 OCZ core 2 60G SSDs. Will keep you guys posted when I get them.

[Nickel020]

You maybe should include in the first post that the problem is the JMicron controller inside the SSD, not the JMicron controller found on many motherboards.

Also, it seems that SI mainboard controllers have their share of problems too, see here:
http://forums.tweaktown.com/f69/ga-ep45-ds5-config-advise-please-27638/#post259261

And tekjunkie, have you read Anand's review? I don't know whether the Core 2 also uses a JMicron controller or another (which may or may not suffer the same problems) but if it does you won't exactly get great performance from that setup due to the terrible latencies:
http://www.anandtech.com/cpuchipsets/intel/showdoc.aspx?i=3403&p=8
I'd just go for two good SLC SSDs like the Mtron 7500 or a Memoright, performance is likely better due to the better controllers.

[m^2]

You maybe should include in the first post that the problem is the JMicron controller inside the SSD, not the JMicron controller found on many motherboards.

Also, it seems that SI mainboard controllers have their share of problems too, see here:
http://forums.tweaktown.com/f69/ga-ep45-ds5-config-advise-please-27638/#post259261

And tekjunkie, have you read Anand's review? I don't know whether the Core 2 also uses a JMicron controller or another (which may or may not suffer the same problems) but if it does you won't exactly get great performance from that setup due to the terrible latencies:
http://www.anandtech.com/cpuchipsets/intel/showdoc.aspx?i=3403&p=8
I'd just go for two good SLC SSDs like the Mtron 7500 or a Memoright, performance is likely better due to the better controllers.
As you noticed, mainboard controllers (IIRC Tony said that BIOSes actually) have their share in the problem too.
tekjunkie's gonna run external controller and it's likely that it won't have the issue. I'm looking forward to see how it turns out.:up:

[Levish]

Is there big performance difference with Areca and ICH10R when using two SSD's in RAID 0?

Not just areca, any controller that will cache writes and they can be had in flavors up to 2GB atm. Just be aware that sometimes the controllers disable write cache if there is no onboard (the controller) battery backup.

Of course the better ones will allow for better scaling of the drives

[XS Janus]

I hope someone (preferably Anand) would do a follow up and test the same SSD's on a hardware controller so we can compare his test and see if the situation is any better.

[RealTelstar]

HAHA acutally I do not wait for much with it.

I went back to my bench rig with 2x 74gb Raptors in Raid 0 and it feels so slow and makes so much noise. This is the same setup I used before I installed the SSD Raid, it's OMG so slow I can hardly stand it.

4x intel slc drives (which i'm getting soon) should attain similar performance, as speed is about double.

[Nickel020]

As you noticed, mainboard controllers (IIRC Tony said that BIOSes actually) have their share in the problem too.
tekjunkie's gonna run external controller and it's likely that it won't have the issue. I'm looking forward to see how it turns out.:up:

That's two different things though. The mainboard controllers usually have driver problems, which may result in data corruption. A good external controller lets you get around those problems.
The controller inside the SSD is responsible for bad random acces times/transfer rates and no external RAID controller will resolve that problem. So if the SSD as a bad controller itslef like the Core V1 there's nothing you can do except get a better SSD:

[m^2]

That's two different things though. The mainboard controllers usually have driver problems, which may result in data corruption. A good external controller lets you get around those problems.
The controller inside the SSD is responsible for bad random acces times/transfer rates and no external RAID controller will resolve that problem. So if the SSD as a bad controller itslef like the Core V1 there's nothing you can do except get a better SSD:

http://www.xtremesystems.org/forums/showpost.php?p=3219663&postcount=20

[Levish]

Tony's own tests confirm small random writes being a issue with the drive. The thing to bear in mind is if random writes are relevant to the intended market and how much so. There's also instances where latency can go up to slightly over one full second which is many times worse than a normal drive and should be the focus of attention of whatever Jmicron or whoever's developers are working on improvements of the firmware.

[@@@@]

If this is concrete that is way there are some speculations that a lot of the SSD manufacturer are changing to a wider memory controller in Q4 2008 or Q1 2009

[Shintai]

If this is concrete that is way there are some speculations that a lot of the SSD manufacturer are changing to a wider memory controller in Q4 2008 or Q1 2009

You mean changed supplier? :ROTF:

[Levish]

i'm sure Jmicron is capable of fixing this problem with a new controller revision, hopefully they can address it to some extent with a firmware update *crosses fingers*

[Biker]

You maybe should include in the first post that the problem is the JMicron controller inside the SSD, not the JMicron controller found on many motherboards..

I was hoping that by being mildly cryptic it would encourage people to actually read the Anandtech article and thereby garner an indepth understanding of the issue... but hey! ppl are lazy these days ;)

Post #1 updated.

And tekjunkie, have you read Anand's review? I don't know whether the Core 2 also uses a JMicron controller or another (which may or may not suffer the same problems) but if it does you won't exactly get great performance from that setup due to the terrible latencies:
http://www.anandtech.com/cpuchipsets/intel/showdoc.aspx?i=3403&p=8
I'd just go for two good SLC SSDs like the Mtron 7500 or a Memoright, performance is likely better due to the better controllers.

I have personally warned tekjunkie about this issue twice but he has chosen to go ahead anyway...


If this is concrete that is way there are some speculations that a lot of the SSD manufacturer are changing to a wider memory controller in Q4 2008 or Q1 2009

I hope they do!

That's all these products need to shine....




I hope someone (preferably Anand) would do a follow up and test the same SSD's on a hardware controller so we can compare his test and see if the situation is any better.

I doubt the situation will be improved much. The problem lies with the JM602 on the SSD.

I have no way to prove this right now so, like you, I await results :)

[Tony]

Guys

Remember SSD technology is speeding up at a furious pace, that said the tech behind core may be 3 to 6 months old so you have to remember that if newer faster controllers were available when we decided to sell SSD we would have used them.

You go with that is available and adopt new tech as it comes along, V3 is being planned already and V4 and V5 and V6 LOL...i hope you get what im saying here;)

MLC SSD is new technology, Core has limits but is priced pretty competitively and for some will be a good midrange mainstream quiet drive which is where we have tried to position it.

I also still feel SB controllers are tuned for HDD which is NOT SSD. I found raid 0 with 2 drives was faster setting UDMA4 on the drives prior to setting the controller to raid (ICH10) which was kinda weird...so code tweaks etc could make the slower MLC drives work a little better.

Also Vista is not to SSD friendly, once you get tweaking you can speed things up but ultimately we are waiting for the new windows to be released.

[Shintai]

Also Vista is not to SSD friendly, once you get tweaking you can speed things up but ultimately we are waiting for the new windows to be released.

*cough* jmicron *cough*

Nothing wrong with Vista.

http://www.anandtech.com/cpuchipsets/intel/showdoc.aspx?i=3403&p=9

[Biker]

Guys

Remember SSD technology is speeding up at a furious pace, that said the tech behind core may be 3 to 6 months old so you have to remember that if newer faster controllers were available when we decided to sell SSD we would have used them.

You go with that is available and adopt new tech as it comes along, V3 is being planned already and V4 and V5 and V6 LOL...i hope you get what im saying here;)


Exactly.

This particular tech is developing at such a furious pace I would find it impossible to know when to jump and with what parts if I were manufacturing a batch :)

Just wish it was possible to pop out the JM602 out and replace it a bit like the opamps on my soundcard ;)

[tekjunkie]

Core has limits but is priced pretty competitively and for some will be a good midrange mainstream quiet drive which is where we have tried to position it.


This is exactly why I bought it. I don't care if it does not set records in read/write performance. I am not there to set benchmark records.All I want is pretty good performance and the silence is golden. As long as the drives don't stutter because of the random writes(which I don't think it would being paired with the kick ass ARC) and as long I am able to enjoy it, I don't mind spending money. It did cost me 1100 for 4 drives though and yes I could have gotten a MTRON PRO or some other SLC based SSD for that amount of money, but the cost effectiveness(price/GB) also comes intothe picture.

[Tony]

*cough* jmicron *cough*

Nothing wrong with Vista.

http://www.anandtech.com/cpuchipsets/intel/showdoc.aspx?i=3403&p=9
Sandisk delayed SSD due mostly to Vista and the random write issue. MFT is a driver that forces the OS to sequentially write instead of random write.

So saying the OS does not carry part of the blame is not true.

And yes i know all about what Anand has said, i unlocked the OCZ support forum to allow end users to chat about the issue.

[Tony]

Exactly.

This particular tech is developing at such a furious pace I would find it impossible to know when to jump and with what parts if I were manufacturing a batch :)

Just wish it was possible to pop out the JM602 out and replace it a bit like the opamps on my soundcard ;)

I agree it is an issue, we know the flash is good its just the controller has limits.
I would love to have seen a socketed controller, just like I would love to see a socketed NB on a mobo so i could try a few to find a goodun ;) LOL

There may be a way to speed things up, we have not given up yet.

[Praz]

Thought this was old news but evidently not.

Microsoft's Windows Vista is to blame for slowing down the progress -- and, in turn, adoption-- of solid state drives (SSDs), according to Sandisk Chairman and CEO Eli Harari.

"As soon as you get into Vista applications in notebook and desktop, you start running into very demanding applications because Vista is not optimized for flash memory solid-state disk," Harari said, during SanDisk's second-quarter earnings call on Monday.

"Unfortunately, [SSD] performance in the Vista environment falls short of what the market really needs and that is why we need to develop the next generation, which we'll start sampling end of this year, early next year," he added.

The next generation of SSDs will use multilevel cell (MLC) technology, which is generally predicted to lower costs and widen the availability of the SSD -- a drive without any moving parts -- as an alternative to typically less rugged traditional hard drives (HDs).

But MLC SSD drives will also require a sophisticated controller -- and according to Harari, that is entirely Vista's fault.
http://www.betanews.com/article/Windows_Vista_hurting_SSD_development_claims_SanDi sk_CEO/1216749504

[@@@@]

The OS has something to do with the performance like Tony said, if you search for it somebody has already tuned their OS to almost double the read and right speed of their SSD well above the manufacturers specification

side note: i just put OS instead of a specific brand to avoid starting a confrontation

[WRC]

I wish microsoft could release a patch to increase SSD performance. Vista was obviously designed for HDD because it does write very few kb at a time, and SSD's like a lot of KB written at one time

[Zorlac]

Guys

Remember SSD technology is speeding up at a furious pace, that said the tech behind core may be 3 to 6 months old so you have to remember that if newer faster controllers were available when we decided to sell SSD we would have used them.

You go with that is available and adopt new tech as it comes along, V3 is being planned already and V4 and V5 and V6 LOL...i hope you get what im saying here;)

MLC SSD is new technology, Core has limits but is priced pretty competitively and for some will be a good midrange mainstream quiet drive which is where we have tried to position it.

I also still feel SB controllers are tuned for HDD which is NOT SSD. I found raid 0 with 2 drives was faster setting UDMA4 on the drives prior to setting the controller to raid (ICH10) which was kinda weird...so code tweaks etc could make the slower MLC drives work a little better.

Also Vista is not to SSD friendly, once you get tweaking you can speed things up but ultimately we are waiting for the new windows to be released.

So has Microsoft said they are going to optimize Vista for SSD anytime soon, or do we have to wait for Windows 7?

I have looked on the OCZ boards and have seen lists of things to turn off in Vista to make it more SSD friendly. I would hope that MS patches Vista to include some sort of SSD mode to run in if it detects the drive.

[rge]

Quote from http://www.pcworld.com/businesscenter/article/149518/samsung_microsoft_in_talks_to_speed_up_ssds_on_vis ta.html
"Samsung hopes to work with Microsoft to boost SSD performance on Windows by discovering optimal packet sizes for data transfers and the best ways to read and write files... Vista has been optimized to handle hard-drive data in smaller chunks. In contrast, the sector size -- also known as page size -- of SSDs are larger than hard drive sector sizes. That results in inefficient SSD performance when slotted into a disk drive bay, Wong said. "My guess is that [Samsung and Microsoft] are maybe working on the OS recognizing an SSD with a 4K-byte sector size instead of a hard disk drive with a 512-byte sector size," Wong said."

Looks like they may optimize for Vista, as for when....who knows, sp2?

[bingo13]

Thought this was old news but evidently not.


http://www.betanews.com/article/Windows_Vista_hurting_SSD_development_claims_SanDi sk_CEO/1216749504

Also, remember those comments came after SanDisk got caught with their pants down when Intel/Micron officially entered the SSD market. They knew they were behind on controller technology since it was outsourced to JMicron at the time along with potential MLC performance problems.

Regardless of the OS in question here (XP SP3 / OS X shows the same problems), Intel figured out how to make it work right for MLC, the fact that JMicron and others will have similar controller technology in the next three to six months just confirms there is a big pink elephant in the center of the room that nobody wants to openly acknowledge.

Microsoft has already confirmed there will be advancements in Windows 7 for SSD technology. However, until we get to that point and who really knows if the changes will work, the best possible solution lies with the new controller technologies and Intel happens to be first out of the gate with it along with a price to reflect it.

In the meantime, I am out about $700 on drives I will not even put in my kid's machines. Why, most of their usage centers on IM, browsing, and document/graphic creation, just the apps that are having all the issues. ;)

[m^2]

So has Microsoft said they are going to optimize Vista for SSD anytime soon, or do we have to wait for Windows 7?

I have looked on the OCZ boards and have seen lists of things to turn off in Vista to make it more SSD friendly. I would hope that MS patches Vista to include some sort of SSD mode to run in if it detects the drive.
Upgrade to XP.

[Shintai]

Sandisk delayed SSD due mostly to Vista and the random write issue. MFT is a driver that forces the OS to sequentially write instead of random write.

So saying the OS does not carry part of the blame is not true.

And yes i know all about what Anand has said, i unlocked the OCZ support forum to allow end users to chat about the issue.

You can always blame the OS. Its simply too easy.

Intel already showed its not an OS problem. Just lack of development or funding for an actual working controller. And JMicron isnt exactly the most quality known brand either. And here it was abit too cheap...

[Tony]

gentelmen

Intel worked around the issue, they did not fix the issue at all, infact i have some info that Intels drive running MFT within the OS can be upto 15X faster.

Any bench run within Vista test the drive how Vista wants to test it...you are NOT testing the drive you are testing how Vista tests the drive.
So the OS is still an issue, its just you can work around the issue and bring performance up...you do NOT fix the issue though.

We have a new drive planned with a work around fix, once released i am sure many will run the OS release of MFT also and see speed increases that make the drive incredibly fast. The issue is the controller on this drive is expensive and as such may push the drive to a cost where most will not jump...we will have to wait and see.


And Gary i openly challenge Anandtech to talk to Sam at MFT, learn what the real issues are here and test the software. show the world what can be done when bugs within an OS are ironed out.

[m^2]

gentelmen

Intel worked around the issue, they did not fix the issue at all, infact i have some info that Intels drive running MFT within the OS can be upto 15X faster.

Any bench run within Vista test the drive how Vista wants to test it...you are NOT testing the drive you are testing how Vista tests the drive.
So the OS is still an issue, its just you can work around the issue and bring performance up...you do NOT fix the issue though.

We have a new drive planned with a work around fix, once released i am sure many will run the OS release of MFT also and see speed increases that make the drive incredibly fast. The issue is the controller on this drive is expensive and as such may push the drive to a cost where most will not jump...we will have to wait and see.


And Gary i openly challenge Anandtech to talk to Sam at MFT, learn what the real issues are here and test the software. show the world what can be done when bugs within an OS are ironed out.

Maybe MFT fixes buggy flash drives handling in Vista, but it's not it's main purpose and I guess that not the biggest reason for speedup. And it's patented, so it's definitely not a bug that MS didn't implement the same.

Results with MFT are irrelevant to what users can expect when MS fixes the bugs unless MS decides to pay fees to Easy Co.

[Tony]

Or they fix the way the OS writes to its data bases within Windows 7 which I hear they have done. Im just reading about Readyboost, the gist of what i get here is they found a way to fix the disk slow downs vista causes, and called it readyboost...

let me reiterate we are working on a drive which will work around the issues Vista has presented. I would rather see Vista fixed with a hotfix though and cheap MLC drives for everyone from every company who sells them having no issues.

[Praz]

Maybe MFT fixes buggy flash drives handling in Vista, but it's not it's main purpose and I guess that not the biggest reason for speedup.
MFT has a twofold purpose. Windows writes to flash drives the same as conventional hard drives. To write 512k of data to a conventional hard drive is a simple matter of finding unused space, be it empty or previously deleted data, and writing the file to the disk. MLC drives cannot be written to in this manner. The existing block must first be read, changes to it made with the new data and then rewritten back to the drive.To compound this issue is block size. That same 512K of data may require manipulating 2MB - 4MB of data to complete the write. Operating systems developed specifically for SSD devices such as JFFS and YAFFS write the data in a linear fashion thus avoiding these issues.

If Windows distinguished between conventional and SSD drives in the same manner this would not be an issue. This is basically what MFT does. So to say these issues are not operating system based nor that this is not the purpose of MFT is not entirely correct.

The second benefit of MFT is extending the wear life of the SSD because of the way the data is written to the drive.

[bingo13]

gentelmen

Intel worked around the issue, they did not fix the issue at all, infact i have some info that Intels drive running MFT within the OS can be upto 15X faster.

Any bench run within Vista test the drive how Vista wants to test it...you are NOT testing the drive you are testing how Vista tests the drive.
So the OS is still an issue, its just you can work around the issue and bring performance up...you do NOT fix the issue though.

We have a new drive planned with a work around fix, once released i am sure many will run the OS release of MFT also and see speed increases that make the drive incredibly fast. The issue is the controller on this drive is expensive and as such may push the drive to a cost where most will not jump...we will have to wait and see.


And Gary i openly challenge Anandtech to talk to Sam at MFT, learn what the real issues are here and test the software. show the world what can be done when bugs within an OS are ironed out.

I already spoke with John last night about it who just had a conversation with Sam the new product developments. I know what MFT does and I am impressed with it. In the meantime, until he has a version released that allows you to use MFT on an OS drive, this subject matter is mute unless OCZ is going to include a copy of it with each Core drive. :D

The fact remains that Intel/Micron figured out a way to make their controller work with Vista/XP/OSX/Linux. The XM-25 drive and Intel's solution certainly is not perfect, but it is better than the current JMicron solution.

[Tony]

Elegant work around was the term I heard today, which is exactly what it is and exactly what we should not really have to do.

So will Anandtech add to the review that Intel have an elegant work around that gets around OS issues and the drives are actually fine as long as the OS has been coded to work with SSD?

Windows 7 is supposed to fix all this, SSD friendly from the start, bring it on is what I say

[saint-francis]

Yes, sure Windows 7 very well might be SSD friendly. What's the ETA for it? Is there even an ETA on a Beta we can tinker with? Vista was years overdue. By the time it comes out SSD's will be much cheaper and larger and will probably be free of these issues.

[rge]

First beta's by this Christmas, mainstream release they continue to quote late 2009 or by jan 1, 2010....assuming they do as planned.
http://gizmodo.com/5048501/windows-7-beta-1-coming-in-december-probably
http://blogs.zdnet.com/microsoft/?p=1582

Windows 7 may be the first time I try a beta OS....if I can get a later release of one, depending on response to it....and switch to SSD at same time.

[dinos22]

Doesn't Anand also write that he could replicate the problems on OCX so it may not all be Vista exclusive :confused:

[bmg]

Doesn't Anand also write that he could replicate the problems on OCX so it may not all be Vista exclusive :confused:
I suspect the problems could be replicated with any OS. It's fashionable to blame stuff on Vista though. Anand's article really lays out the issue in hard to refute detail, but some people seem to be in denial and looking for ways to say that the SSD drives using jmicron controllers aren't at fault. I applaud companies like OCZ introducing SSD products at more competitive prices, but current drives obviously do have problems. Hopefully the issues get fixed promptly and trouble free drives become more affordable than they currently are.

[Praz]

Doesn't Anand also write that he could replicate the problems on OCX so it may not all be Vista exclusive :confused:
MCL SSD drives will show similar issues on any non-log structured file system. The extend of the issues depend on the amount and dependencies of database type files that are part of the operating system. As it's really doubtful Microsoft is going to change the file structure in Window 7 the best we can hope for is a MFT type fix. Even that may be wishful thinking.

[Biker]

Doesn't Anand also write that he could replicate the problems on OCX so it may not all be Vista exclusive :confused:

Having a look around several forums reveals users reporting problems across the board.

A couple of helpful links:

Windows XP fix thread here (http://www.ocztechnologyforum.com/forum/showthread.php?t=42487)

Windows Vista X64 performance discussion and performance improving tips here (http://www.alternativerecursion.info/?p=106)

I do not own a SSD right now so I cannot verify if these actually work...

[dengyong]

Having a look around several forums reveals users reporting problems across the board.

A couple of helpful links:

Windows XP fix thread here (http://www.ocztechnologyforum.com/forum/showthread.php?t=42487)

Windows Vista X64 performance discussion and performance improving tips here (http://www.ocztechnologyforum.com/forum/showthread.php?t=42487)

I do not own a SSD right now so I cannot verify if these actually work...

You have the same link for vista64 as xp.

[Biker]

You have the same link for vista64 as xp.

No idea how that happened....

Anyway, thanks for the heads up :)

*fixed* :D

[dunn2953]

if you closely read this article you will realize that like tony is saying, the way Vista writes to the OS is at fault. This isn't solely a vista problem it will happen on any OS due to the fact they they were all coded to be run on a Hard Drive.

the Intel controller solves much of the problem with a great work around, but once Microsoft and other OS makers release operating systems designed to work on SSDs the speed of SSDs will explode. MFT lets you see drives jumping in speeds up to around 10x simply by changing the way the OS writes and reads from the disk and proves that once OS are coded in a similar fashion to MFT we will see huge speed increases and it will fix the problem caused my the JM controller not being able to well handle the way the OS writes and reads.

the Inter controller does much to solve it but in the end it cannot compete with the performance advantage that will be solved by a real file system solution.


WinFS ever?

[bmg]

if you closely read this article you will realize that like tony is saying, the way Vista writes to the OS is at fault. This isn't solely a vista problem it will happen on any OS due to the fact they they were all coded to be run on a Hard Drive.

the Intel controller solves much of the problem with a great work around, but once Microsoft and other OS makers release operating systems designed to work on SSDs the speed of SSDs will explode. MFT lets you see drives jumping in speeds up to around 10x simply by changing the way the OS writes and reads from the disk and proves that once OS are coded in a similar fashion to MFT we will see huge speed increases and it will fix the problem caused my the JM controller not being able to well handle the way the OS writes and reads.

the Inter controller does much to solve it but in the end it cannot compete with the performance advantage that will be solved by a real file system solution.


WinFS ever?

You make some good points. I just want to emphasize that the way Vista writes to the disk is *NOT* at fault. When you design and sell a new product for an existing OS environment it's your responsibility to make sure the product works correctly, or supply the proper drivers if it doesn't. I worked in the Laser printer business for years, and it was our responsibility to make sure customers had the drivers they needed to properly use our products, it wasn't Microsoft's, Apple's, etc. It's up to manufacturers like OCZ to ensure that their product works with Vista, not Microsoft's. Intel made sure their hardware solution works with Vista and other OS's, with no special drivers. If others aren't going to do that, they need to supply the proper drivers, or quit selling into markets where their products don't work correctly.

Sure, now that SSD's are becoming more common, Microsoft and others will try to ensure better built-in support when they release newer versions of their OS's. They'll do the same thing with integrating in drivers for a wide variety of newer products too, that's just the way things work. Right now the fact that many SSD products don't work well with Vista and likely other OS's isn't the OS maker's problem, that responsibility lies with the companies like OCZ.

Don't get me wrong, even though I mention OCZ by name and haven't looked up and listed all the other manufacturers that basically sell the same products, I do think it's great that OCZ has tried to take the lead in making these types of products affordable for us. It's just that many of the current products have issues that will hopefully be addressed with future products with improved SSD controllers, or possibly providing driver support to improve the performance of existing products.

[dunn2953]

You make some good points. I just want to emphasize that the way Vista writes to the disk is *NOT* at fault. When you design and sell a new product for an existing OS environment it's your responsibility to make sure the product works correctly, or supply the proper drivers if it doesn't.

you are right about this and intel seems to be the one with the money and massive development teams to be able to have achieved this so quickly

[Shintai]

WinFS ever?

WinFS is still the same NTFS you got today in Vista. Its just paired with an SQL aswell.

Windows 7 (And its Server 2008 R2 equal) will change nothing. Its a bad excuse....:down:

[Jack]

OCZ has lowered the price of the OCZ Core Series OCZSSD2-1C64G 2.5" 64GB to $99,- (after MIR)! :up:

http://www.newegg.com/Product/Product.aspx?Item=N82E16820227344&Tpk=N82E16820227344

Very nice price :D

Announcement: Dailytech.com (http://www.dailytech.com/article.aspx?newsid=12993)

[Shintai]

No wonder, JMicron based SSDs are basicly worthless.

[RyderOCZ]

They are definitely not worthless. At that price as a secondary drive with games installed on it. They are fantastic in this application. Game and Level load times are drastically improved off an SSD.

[Miwo]

I am guessing this is mostly a Writing issue, and reading is fairly unaffected? If so, I do agree with Ryder as the price is very nice (cheaper than a Raptor, which people get for gaming mostly) for a dedicated game drive. Anand does seem to back this up, though Raid HD configurations weren't tested http://www.anandtech.com/cpuchipsets/intel/showdoc.aspx?i=3403&p=14

I am using 3x 7200.10s in Matrix Raid with a 40GB slice over three drives on my OS/ Games partition. I am wondering if I'd be better off with a single or even two Core SSD in Raid0 , strictly for games. I don't care about Writing performance, just raw Reads and seek times. I should still be better off with a SSD for games though right, since its primarily read only? My main game is Warhammer online right now where there is definitely level load when you zone, or cross zone borders. I definitely still feel HD chugging on my Raid setup

[saint-francis]

I am guessing this is mostly a Writing issue, and reading is fairly unaffected? If so, I do agree with Ryder as the price is very nice (cheaper than a Raptor, which people get for gaming mostly) for a dedicated game drive. Anand does seem to back this up, though Raid HD configurations weren't tested http://www.anandtech.com/cpuchipsets/intel/showdoc.aspx?i=3403&p=14

I am using 3x 7200.10s in Matrix Raid with a 40GB slice over three drives on my OS/ Games partition. I am wondering if I'd be better off with a single or even two Core SSD in Raid0 , strictly for games. I don't care about Writing performance, just raw Reads and seek times. I should still be better off with a SSD for games though right, since its primarily read only? My main game is Warhammer online right now where there is definitely level load when you zone, or cross zone borders. I definitely still feel HD chugging on my Raid setup

If you don't write to it the Core will rock.

[saki_bomber]

lol

Show off.

I am not at all jealous.

Nope.

Not even a little.

Honest.

No, really.

:D


def not jealous of how much that cost:eek:

speed is important to me but it'll never be THAT important

[bmg]

They are definitely not worthless. At that price as a secondary drive with games installed on it. They are fantastic in this application. Game and Level load times are drastically improved off an SSD.
Exactly. Too bad the rebate doesn't cover 2 drives instead of just 1 (any chance of OCZ changing that?). It would be really tempting to raid a pair of them for a games only drive and stick with a regular raptor for the OS.

One thing about Core as a games drive...Anand's review showed poor minimum fps results in Crysis with Core (first time through the benchmark when getting data from the drive as the benchmark ran), however, that's undoubtedly because Crysis was running on the same Core drive as the OS. I'm sure that sort of thing's a non-issue if the drive's just being used for games and nothing else.

[dinos22]

OCZ has lowered the price of the OCZ Core Series OCZSSD2-1C64G 2.5" 64GB to $99,- (after MIR)! :up:

http://www.newegg.com/Product/Product.aspx?Item=N82E16820227344&Tpk=N82E16820227344

Very nice price :D

Announcement: Dailytech.com (http://www.dailytech.com/article.aspx?newsid=12993)

a lot of user feedback confirms what anand was talking about on egg

it would still be worth it though
far out $99 EEEEEK :D

why not just go out with a straight $99 price rather than do rebates cause overseas guy can't get them heheh

[bmg]

a lot of user feedback confirms what anand was talking about on egg

it would still be worth it though
far out $99 EEEEEK :D

why not just go out with a straight $99 price rather than do rebates cause overseas guy can't get them heheh
Newegg now shows it as a deactivated item, so looks like they sold out of their stock of the V1 64gb drives and likely won't be getting more. Killer deal if used as a non-OS drive.

Update. Looks like mwave has them for the same great deal: http://www.mwave.com/mwave/ViewProduct.asp?CartID=MWV@VXR9KF3RS7QT31PYGGY5EWB VDP7PWS5T9Y6B7IT499HWTE&PID=HDSSD-OCZ&DPDID=&Rebate=0&ST=&SortMethod=10

[Biker]

Sweeeet deal for a non-OS drive!!

I just wish we could get equivalent prices elsewhere.. eh Dino? ;)

The cheapest I can find it in the UK is £149.99....

[B.E.E.F.]

I'm curious.

Before the Core SSDs were mass produced and put for sale. Wasn't there any testing done to make sure they work? Properly?

[Zaskar]

Wow nice price, even with the Core V1's limitations, Id think 2 (you can almost get 3) for the cost of 1 V2 would still be better performance wise in a raid 0.

[Levish]

I'm curious.

Before the Core SSDs were mass produced and put for sale. Wasn't there any testing done to make sure they work? Properly?

They work, sustained 4k random writes just aren't as big a issue as they are being made out to be. If they are that big a issue the users should have chosen either a write caching controller or a SLC model drive.

[Praz]

They work, sustained 4k random writes just aren't as big a issue as they are being made out to be. If they are that big a issue the users should have chosen either a write caching controller or a SLC model drive.
True. But some users are having issues. So for them the Intel MLC might be an option. I think one thing we are losing sight of when comparing these drives is cost. The Intel is going to be 3x the price of the others so i think the performance difference is something that should be expected.

One thing I've hardly seen mentioned involves Intel's Adaptive Memory Controller Technology. Hopefully in real world usage the hype it has received will be justified. Evidently, because of AMCT, after running a program such as IOmeter the drive has to be reformatted with a secure erase procedure to get any meaningful results from other benchmarks. Guess we'll have to wait and see what constitutes a change of use for this type degradation to occur in the real world. Most reviews sites seem to forget to mention this.

[dinos22]

Newegg now shows it as a deactivated item, so looks like they sold out of their stock of the V1 64gb drives and likely won't be getting more. Killer deal if used as a non-OS drive.

Update. Looks like mwave has them for the same great deal: http://www.mwave.com/mwave/ViewProduct.asp?CartID=MWV@VXR9KF3RS7QT31PYGGY5EWB VDP7PWS5T9Y6B7IT499HWTE&PID=HDSSD-OCZ&DPDID=&Rebate=0&ST=&SortMethod=10

have a look at this NOT SO GREAT deal from Mwave in Australia :down::down:

http://www.mwave.com.au/newAU/mwaveAU/search.asp?CartID=mAU@WEXJ85WC4X7F33TK3JR1O1QTC32X B8O3KOLTV6LMSYVGS6&search=ocz%20ssd&cateID=2214

~US$290 for V1 64GB OCZ SSD

[Praz]

have a look at this NOT SO GREAT deal from Mwave in Australia :down::down:

http://www.mwave.com.au/newAU/mwaveAU/search.asp?CartID=mAU@WEXJ85WC4X7F33TK3JR1O1QTC32X B8O3KOLTV6LMSYVGS6&search=ocz%20ssd&cateID=2214
They're holding you guys at gunpoint. Saw something similar with a RAID controller recently. Knew people buying it in the U.K. for about $40.00 and here in the states still selling for $230.00

[dinos22]

They're holding you guys at gunpoint.

you think so :D
heheheh

[vol7ron]

Elegant work around was the term I heard today, which is exactly what it is and exactly what we should not really have to do.

So will Anandtech add to the review that Intel have an elegant work around that gets around OS issues and the drives are actually fine as long as the OS has been coded to work with SSD?

Windows 7 is supposed to fix all this, SSD friendly from the start, bring it on is what I say

I've read the whole thread from page 1 to 4. Some people seem to think it's a SSD Manufacturer problem, some people think it's an OS.

I'd like to address that it is both of their problem. The problem is that consumers don't like it. Where the issue is fixed does not matter to the consumer, so long as performance is restored. Whether the fix is a "workaround" or a legitimate redesign. The truth is an OS fix could be called a workaround as well.

The thing with hard drives, whether it's a disk or an SSD drive, is that they should be backward compatible. That means if I wanted to put Windows 98, 3.1 or whatever OS on the drive, it should work. I don't think that's guaranteed here. And you can be sure that Microsoft isn't going to update old OSs. While OCZ is right, the drive works if you use it how it's intended, and Software companies are right ~ their OS works so long as the hardware accepts it, this is like putting a Ferrari engine on a lawnmower. You have two things that by themselves do great at what they do, but one of them have to give.

The big issue with the OS changing in addition to what I say above, is how it would work with the combination of other hardware. If you increase the block size, what would happen to disk drives? Would that be a negative impact? Would it be able to change the write size depending on what hardware is being used? Or would you only be able to use one hardware technology at a time, that is you won't be able to use disk drives and SSDs concurrently.

Another issue that Intel has is the life expectancy, as we all know that SSDs get weaker over time with the writes. Is developing an algorithm to manage write locations also an OS problem?

I love OCZ and their memory business has put them in a prime position to distribute cost-affordable SSDs, but take some blame and some initiative. Sure, a better OS would solve the problem, but the OS is already there and works with other products. Yours is new - if you whine and do nothing, no one is going to buy.. they're just going to go with what works, no matter who fixed it.

I would also like to see some open-source software groups develop something similar to MFTs product as this shouldn't be too big of a project. A free version would make many users happy.

[Tony]

We are dealing with end users case by case, we have a new controller under test...there is little else we can do.

drop 3 cores on a nice card for the price of the 60GB Intel drive and get 2x+ the storage, more speed and no stutter...no brainer to me.

[Xoulz]

I am guessing this is mostly a Writing issue, and reading is fairly unaffected? If so, I do agree with Ryder as the price is very nice (cheaper than a Raptor, which people get for gaming mostly) for a dedicated game drive. Anand does seem to back this up, though Raid HD configurations weren't tested http://www.anandtech.com/cpuchipsets/intel/showdoc.aspx?i=3403&p=14

I am using 3x 7200.10s in Matrix Raid with a 40GB slice over three drives on my OS/ Games partition. I am wondering if I'd be better off with a single or even two Core SSD in Raid0 , strictly for games. I don't care about Writing performance, just raw Reads and seek times. I should still be better off with a SSD for games though right, since its primarily read only? My main game is Warhammer online right now where there is definitely level load when you zone, or cross zone borders. I definitely still feel HD chugging on my Raid setup



Aye, thinking the same thing. What server on Warhammer? (Dark Crag/Destruction)



.

[azza21]

We are dealing with end users case by case, we have a new controller under test...there is little else we can do.

drop 3 cores on a nice card for the price of the 60GB Intel drive and get 2x+ the storage, more speed and no stutter...no brainer to me.

Any advice on what card?

[Omastar]

Any advice on what card?

I like the Adaptec 5405, Highpoint RocketRAID 3510, and the Areca 1210 is always a winner, although the DDR-333 cache on that controller REALLY needs to be updated to DDR2-533 or 667 to match the other controllers in its range. 512MB cache would also be nice, as the 1210 is getting a tad long in the tooth. Or, if you have the money, the Areca 1680ix with 2GB cache is really nice. ;)

[Tony]

Anything with PCIE4x, a fast processor and at least 128MB ram, 256MB is better.

1210 is available at scan : http://www.scan.co.uk/Product.aspx?WebProductId=505261

[azza21]

Anything with PCIE4x, a fast processor and at least 128MB ram, 256MB is better.

1210 is available at scan : http://www.scan.co.uk/Product.aspx?WebProductId=505261

So if i brought 3/4 Ocz 30gb core 1 ssd's and raid0 them would if fix these ominous problems, or would i still get the problems people are experiencing with ssds?

[nox_uk]

i'm waiting for 60GB ~ £100, then will buy some. Reckon it will be around 6months at this rate.

I would of thought the cards would sort the stuttering, as it seems to be on the write buffering and if the raid card does that not the SSD... but unless someone is willing to buy me a load of SSD's i can't test, so don't quote me!! :D

Nox

[Jamesrt2004]

So if i brought 3/4 Ocz 30gb core 1 ssd's and raid0 them would if fix these ominous problems, or would i still get the problems people are experiencing with ssds?

If you brought a GOOD raid controller then what he is saying = Yes

[Biker]

Good new info here....

Post #1 updated...

Thanks guys!

[WhiteFireDragon]

just in case you guys didn't already know, intel's XM-25 finally available on newegg (http://www.newegg.com/Product/Product.aspx?Item=N82E16820167005), although you could have had it for about a month if you hunted for it elsewhere

[XS Janus]

Nice!

Anyone heard anything new about their "E" or enterprise version of SSDs?

[Helmore]

Nice!

Anyone heard anything new about their "E" or enterprise version of SSDs?

Only that they are very expensive, even compared to their mainstream 80GB drive.

[Biker]

Yup.

Really silly pricing (http://cgi.ebay.co.uk/INTEL-X25-E-EXTREME-32GB-SATA-SSD-SOLID-STATE-DRIVE-NEW_W0QQitemZ250316583296QQcmdZViewItem?hash=item2 50316583296&_trksid=p3286.m63.l1177)...

The X25-E varaint drastically inproves the write speed and theoretical longevity of the drive (it's SLC tech) but personally if I was rich enough to be getting one (or more) I would wait for the 64GB version just to get a bit of extra cushioning in terms of capacity obsolecence...

[m^2]

Here (http://www.digitalspyders.com/store/storage-devices/hard-drive-other/other/intel-ssdsa2sh032g1c5-p-240277.html) it's 25% cheaper.
Still costly...but would make a great portable drive .^^

[TimPrice]

overclockers UK (http://www.overclockers.co.uk/productlist.php?groupid=701&catid=14&subid=910) as always

[karl_eller]

I'm really hoping this problem gets fixed soon. I originally had plans on dropping a cheap 120-160 Gb SSD into my laptop to replace the slow-as-all-crap 5,400 RPM drive it's got now, but due to the stuttering issues with the cheaper MCL drives I put that on hold. And while I'd really love an X25-M, it's pretty hard to justify dropping an AU$900-1000 SSD into an AU$1100 laptop :p: And of course, with most laptops you don't have the option of secondary drives.

Here's hoping that the next round of cheap OCZ/whoever SSDs come around, they'll have a non-craptastic controller on them :p:

Eller

[Peakr]

Here's to hoping the OCZ V3 is up to snuff. Tired of waiting on an SSD with a decent price to performance ratio.

[Biker]

Here's to hoping the OCZ V3 is up to snuff. Tired of waiting on an SSD with a decent price to performance ratio.

Amen, brother.

[XS Janus]

Silly price!
No tests yet?

[Biker]

No tests / reviews yet.. well... no findable ones anyway ;)