OCZ Core series -- Affordable, high-performance SSDs

Silencing hard drives, optical drives and other storage devices

Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee

andyb
Patron of SPCR
Posts: 3307
Joined: Wed Dec 15, 2004 12:00 pm
Location: Essex, England

Post by andyb » Mon Sep 22, 2008 11:03 am

Thanks Aris, its good to know the details and extent of the problem(s).
That makes me weonder if it truly is that one part of if there is more to the story.
Well it seems that none of the manufacturers have admitted to there being any problems, so if they did bring out a fixed replacement problem that would be as good as an admission, and they would then be pursued by angry customers to replace tens of thousands of "faulty" SSD's - which they may or may not actually replace.

RE: Replacing the SSD's, strictly speaking they are not actually faulty. They function, they dont break or loose data, and they benchmark much faster. You would have a much better chance of having one replaced by describing it as "not fit for purpose" rather than faulty, but even then you would have a huge uphill struggle against a huge company that doesnt give a shyte about the end user - most large companies are like this.

BTW: I wouldnt expect any better support from Intel if their products end up having some mysterious fault, they are screwing me around with a faulty CPU, they want me to test it on another board - more on that later, as its way off topic.


Andy

dhanson865
Posts: 2198
Joined: Thu Feb 10, 2005 11:20 am
Location: TN, USA

Post by dhanson865 » Mon Sep 22, 2008 7:18 pm

5%? did you just pull that number out of your ass?
Yes, I didn't mean for that to be an accurate number. I did just pulled it out of thin air.
You didn't see any ATI/AMD or Nvidia chipsets mentioned on those threads? Then your blind my friend. Let me copy paste the exact same link you posted: http://www.ocztechnologyforum.com/forum ... 149&page=2 Did you even read it? Try clicking the "view poll results" button. Intel ICH6-10, Nvidia 680/780/790, AMD SB600/SB7xx, JMicon/other.
Sorry, I read the messages in the thread and forgot about the poll at the top. Also, when I looked at the poll I was looking for nforce 3/4 or 690g. I guess the SB600 is the one I should have paid attention to for that pairing. Unfortunately your responses are so full of anger I was distracted.

But eventually through all the vitrol, I am reading enough to actually learn, even if it is slow progression.

Aris
Posts: 2299
Joined: Mon Dec 15, 2003 10:29 am
Location: Bellevue, Nebraska
Contact:

Post by Aris » Mon Sep 22, 2008 9:13 pm

dhanson865 wrote:Unfortunately your responses are so full of anger I was distracted.
Sorry for my heated replies. This whole ordeal has just really gotten under my skin. Probably more-so because i have been waiting for the day of an affordable SSD for what seems like forever, and now that they are finally here they suffer from such a stupid oversight that should never have happened. And even though it has, it feels like no one is trying to correct it.

Bar81
Posts: 261
Joined: Thu Sep 04, 2003 4:19 pm
Location: Dubai

Post by Bar81 » Tue Sep 23, 2008 11:47 am

dhanson865 wrote:Sorry, I read the messages in the thread and forgot about the poll at the top. Also, when I looked at the poll I was looking for nforce 3/4 or 690g. I guess the SB600 is the one I should have paid attention to for that pairing. Unfortunately your responses are so full of anger I was distracted.
lol, I was hoping that you'd finally notice on your own but I guess that was too much to ask. Well, I guess the upside is that you're "learning", no matter how slowly.

Bar81
Posts: 261
Joined: Thu Sep 04, 2003 4:19 pm
Location: Dubai

Post by Bar81 » Tue Sep 23, 2008 11:53 am

andyb wrote:Can someone please do a little summary for me and everyone else who is not looking at buying an SSD any time soon, but is interested in what is going on here.

This is as far as I understand these issues.

MLC is not an issue in itself, the problems are all to do with the controller not having any on-board cache.

The controllers have issues with at least 50% of the chipsets on the market due to the fact that they have no cache.

The Intel controller does not have issues, even though it does not have cache.??? Please fill in the blanks or correct me on this one.


Andy
As has been stated, MLC is not the issue; rather, it appears to be the controller. If the controller is designed improperly it is likely to have a problem with any and all chipsets because the problem is inherent to the drive. The new Intel MLC SSD seems to be (but because no one actually has one but a few reviewers we can't be sure) the first MLC SSD that delivers SLC like speed without any of the random write issues we've seen from the latest batch of OCZ drives.

Bar81
Posts: 261
Joined: Thu Sep 04, 2003 4:19 pm
Location: Dubai

Post by Bar81 » Tue Sep 23, 2008 12:00 pm

Turas wrote:I just don't get something. If the controller on the SLC one does not have this problem then why don't thye just move that controller to the MLC drives. I have been with the understanding the the price difference between the SLC and MLC drives was due to the coast of the flash and not the controller itself. If that was the case though I would think this would have been done from all of these so far. At least would of though someone like OCZ would of brought out the v2 with the new controller if they knew in fact that the problem was solely due to the controller. That makes me weonder if it truly is that one part of if there is more to the story.
Because MLC and SLC don't work the same. In any case, OCZ doesn't really "make" the drives so I really think they're learning on the job - their support staff seems completely taken by surprise at the extent of the problems. On a final note, reviewers didn't do us consumers any favors by only benchmarking the drives without any real world use.

dhanson865
Posts: 2198
Joined: Thu Feb 10, 2005 11:20 am
Location: TN, USA

Post by dhanson865 » Tue Sep 23, 2008 12:56 pm

Bar81 wrote:
dhanson865 wrote:Sorry, I read the messages in the thread and forgot about the poll at the top. Also, when I looked at the poll I was looking for nforce 3/4 or 690g. I guess the SB600 is the one I should have paid attention to for that pairing. Unfortunately your responses are so full of anger I was distracted.
lol, I was hoping that you'd finally notice on your own but I guess that was too much to ask. Well, I guess the upside is that you're "learning", no matter how slowly.
Sorry, too busy rebuilding a work network and working on SAS70 certification issues to spend thorough time on this thread. I was guilty of skimming the OCZ thread.

In my defense I did ask that someone clue me in on the point more than once. I figured if no one would I'd get to it on a weekend that I wasn't working on servers...

highlandsun
Posts: 139
Joined: Thu Nov 10, 2005 2:04 am
Location: Los Angeles, CA
Contact:

Post by highlandsun » Tue Sep 23, 2008 2:15 pm

Something still doesn't make sense about all this. If the controller is really too stupid to reorder logically random writes into physically sequential writes, then caching alone wouldn't solve the problem. I.e., if you throw a long stream of random I/Os into a big cache, eventually it has to flush out and if the data span is big enough, it will still mean a lot of sparse writes. On the other hand, if the random I/Os all occur in a small region, it's possible that caching them is sufficient to make them all contiguous. I guess the question is how big was the RAID controllers' cache, relative to the data set in each test.

Anyway, the SSD's flash controller ought to be converting all writes into sequential ones, that's the only sane thing to do. With the current products you can't really do that on the PC side. Using a log-structured filesystem, or the MFT driver will probably work most of the time, but since there's no way for the host PC to know when the drive has remapped a block, there's really no way to ensure that what the host PC thinks is a long stream of sequential writes will actually be handled sequentially on the SSD.

dhanson865
Posts: 2198
Joined: Thu Feb 10, 2005 11:20 am
Location: TN, USA

Post by dhanson865 » Tue Sep 23, 2008 5:36 pm

highlandsun wrote:Something still doesn't make sense about all this. If the controller is really too stupid to reorder logically random writes into physically sequential writes, then caching alone wouldn't solve the problem. I.e., if you throw a long stream of random I/Os into a big cache, eventually it has to flush out and if the data span is big enough, it will still mean a lot of sparse writes. On the other hand, if the random I/Os all occur in a small region, it's possible that caching them is sufficient to make them all contiguous. I guess the question is how big was the RAID controllers' cache, relative to the data set in each test.
Typical flash memory block sizes are 64, 128, or 256 kiB. I may be translating size wrong but I'm assuming 256 kiB = 0.25 megabytes with with SLC SSDs using 64KB blocks and MLC SSDs using 128KB blocks

The Intel MLC drives have a controller with a 256KB cache again 0.25 megabytes.

The OCZ drives apparently have a controller with a cache that is 16KB which is not only 16 times smaller than the Intel controllers cache its also smaller than any common flash block size.

It's basically guaranteed that any random writes will cause the OCZ drive to flush the controllers cache (assuming I'm not just misunderstanding the concept).

ATTO benchmark runs from .5 KB to 1 MB meaning it runs from so small every random write will flush the cache to being bigger than the cache which is paradoxically good for SSD performance (in traditional hard drives the cache is way faster than the actual write mechanism so the curve works the opposite way).

The problem is not what the benchmark size is because the benchmarks allow you to vary the size and see the good and the bad.

The problem is what size writes are typical for your usage patterns?

For example there is this short list that I have no idea how old it is (possibly several years old, and I'll not promise there isn't missing context)

O/S Log Block Size
======= ==============
Solaris 512 bytes
HP-UX 1024 bytes
NT 512 bytes
OpenVMS 512 bytes
Digital UNIX 1024 bytes

That's 0.5KB and 1KB blocks.

I'm going to guess that programmers have been using low block sizes for a long time and it'll take time for software to be updated to start using larger block sizes.

Now assuming my math is still on the right part of the conversion factors we have this comparison:

An OCZ MLC SSD the controller has a 16KB cache and on a typical mainstream consumer hard drive there is a 8MB cache (about 512 times the cache)

An Intel MLC SSD the controller has a 256KB cache and a performance hard drive would typically have a 16MB cache (about 64 times larger)

I think a memoright SSD has a 16MB cache versus a really high end 7200 RPM drive having 32MB cache (about 2 times larger)

So hopefully when the flash gets cheap enough there will be enough money left on the bill of materials for SSDs to start to catch up with traditional hard drive cache sizes.

dhanson865
Posts: 2198
Joined: Thu Feb 10, 2005 11:20 am
Location: TN, USA

Post by dhanson865 » Fri Sep 26, 2008 7:49 am

dhanson865 wrote:32GB price comparison
Core v1 32GB - OCZSSD2-1C32G
Core v2 OCZSSD2-2C30G

64GB price comparison
Core v1 OCZ OCZSSD2-1C64G
Core v2 OCZSSD2-2C60G
The cheaper Core V1 supplies are drying up. Prices at the lowest end have risen about $10 to $20 since the post I made last week (8 days ago).

andyb
Patron of SPCR
Posts: 3307
Joined: Wed Dec 15, 2004 12:00 pm
Location: Essex, England

Post by andyb » Fri Sep 26, 2008 9:39 am

So hopefully when the flash gets cheap enough there will be enough money left on the bill of materials for SSDs to start to catch up with traditional hard drive cache sizes.
Cache is dirt cheap, and has been for years as it is simply standard RAM these days. I thought a couple of years ago that by now HDD's would have 256MB of cache onboard as its so very cheap - OK maybe only on high end models such as the current 1TB drives. I suspect it is not used because it doesnt need it, the cache is only really there to help the drive not run slower than it should do, which is why we have seen HDD's cache increa (more or less) inline with the drives mechanical performance.

As far as not even putting the same amount of cache onto a £200+ SSD as is used on a £30 HDD is so stupid I find it difficult to believe. Even if 8MB of cache cost £5, that is nothing compared to the cost of the SSD. Would you spend an extra £5 for an SSD that didnt stutter when you did loads of random writes, I would, but a the cache is integrated into the onbvoard processor and not as a seperate item on the PCB it doesnt need a simple firmware change to take use of the cache in whatever quantity is soldered to the PCB which is the case with HDD's.

I suspect that when Samsung (OCZ's OEM) was developing their MLC SSD's they didnt think this problem would occur, and by then they had already mass produced them and started shipping drives. I suspect that when they bring out their next model, it will have a lump of cache, and make Intel's new drive look slow - maybe thats me hoping what will happen rather than what will happen. Fingers crossed.


Andy

dhanson865
Posts: 2198
Joined: Thu Feb 10, 2005 11:20 am
Location: TN, USA

Post by dhanson865 » Fri Sep 26, 2008 12:54 pm

Cache is dirt cheap, and has been for years as it is simply standard RAM these days. I thought a couple of years ago that by now HDD's would have 256MB of cache onboard as its so very cheap - OK maybe only on high end models such as the current 1TB drives. I suspect it is not used because it doesn't need it, the cache is only really there to help the drive not run slower than it should do, which is why we have seen HDD's cache increase (more or less) inline with the drives mechanical performance.
Nah, they have increased cache more or less in line with capacity. As capacity goes up or common files get larger the need for more cache increases. You also **
As far as not even putting the same amount of cache onto a £200+ SSD as is used on a £30 HDD is so stupid I find it difficult to believe. Even if 8MB of cache cost £5, that is nothing compared to the cost of the SSD. Would you spend an extra £5 for an SSD that didn't stutter when you did loads of random writes, I would, but a the cache is integrated into the onboard processor and not as a separate item on the PCB it doesn't need a simple firmware change to take use of the cache in whatever quantity is soldered to the PCB which is the case with HDD's.
I think the thing is that it does take more than just slapping a chip of ram on there. You have to rewrite the firmware to optimize for the speed differences. You also **
I suspect that when Samsung (OCZ's OEM) was developing their MLC SSD's they didn't think this problem would occur, and by then they had already mass produced them and started shipping drives. I suspect that when they bring out their next model, it will have a lump of cache, and make Intel's new drive look slow - maybe thats me hoping what will happen rather than what will happen. Fingers crossed.
I'm guessing that some bean counter said screw the users and they knew about the performance issues and did it all based on cost.

** The You also's above are this: You also increase your amount/risk of lost data due to power loss the bigger the cache gets. While I'd love to have a 256MB cache on my hard drive I don't want to lose data due to an improper shutdown or power loss. I'd counter that if loss of data weren't an issue cache sizes would be considerably larger on traditional hard drives.

andyb
Patron of SPCR
Posts: 3307
Joined: Wed Dec 15, 2004 12:00 pm
Location: Essex, England

Post by andyb » Fri Sep 26, 2008 2:28 pm

** The You also's above are this: You also increase your amount/risk of lost data due to power loss the bigger the cache gets. While I'd love to have a 256MB cache on my hard drive I don't want to lose data due to an improper shutdown or power loss. I'd counter that if loss of data weren't an issue cache sizes would be considerably larger on traditional hard drives.
That is true, but I have lost data on PC's using HDD's with 256k of cache, and no data lost on PC's with HDD using 32MB of cache. Your milage might vary. Yes it increase the risk a great deal, but is also increases the performance a great deal, in reality a battery pack could be added to SSD's at a tiny cost due to the low power of the SSD, and the tiny amount of time it would take to write that data. It wouldnt take a lot to do the same with a HDD, and personally I would take the risk of data loss for the gain in performance and have HDD's with 256MB of cache.


Andy

dhanson865
Posts: 2198
Joined: Thu Feb 10, 2005 11:20 am
Location: TN, USA

Post by dhanson865 » Mon Dec 01, 2008 3:51 pm

Well, I take Bar81's warning to heart (I bought a traditional HD in 2008 and am waiting for 2009 to reconsider SSDs for my use) but I like to keep track of pricing trends so here we go:

32GB price comparison
These are still floating around for $150 to $200 but are not dropping in price right now. Moving on.

64GB/60GB price comparison
Core v1 OCZ OCZSSD2-1C64G about $170 with a rebate that takes it to about $100 ($70 Mail-In Manufacturer Rebate Available From 12/01/08 Till 12/02/08 )
Core v2 OCZSSD2-2C60G is about $195 and rebate takes it to $165. ($30 rebate Available From 12/01/08 Till 12/16/08 )

128GB/120GB price comparison
Core v1 OCZ OCZSSD2-1C128G about $290 with a rebate that takes it to about $200 ($90 Mail-In Manufacturer Rebate Available From 12/01/08 Till 12/02/08 )
Core v2 OCZSSD2-2C120G is about $330 and rebate takes it to $290 ($40 rebate from 12/1/08 till 12/16/08 )

250GB price comparison
Core v1 no such device
Core v2 OCZ OCZSSD2-2C250G 250gb still insanely expensive and/or not available.

hmsrolst
*Lifetime Patron*
Posts: 426
Joined: Sun Aug 11, 2002 3:26 pm
Location: Arlington, VA USA

Post by hmsrolst » Mon Dec 01, 2008 4:21 pm

Newegg has the 30GB on sale today for $50 after MIR.

http://www.newegg.com/Product/Product.a ... 6820227359

dhanson865
Posts: 2198
Joined: Thu Feb 10, 2005 11:20 am
Location: TN, USA

Post by dhanson865 » Tue Dec 02, 2008 5:39 am

hmsrolst wrote:Newegg has the 30GB on sale today for $50 after MIR.
Quick sale. It's now back up over $100.

Turas
Posts: 43
Joined: Thu Aug 02, 2007 9:29 am
Location: Alexandria, VA

Post by Turas » Tue Dec 02, 2008 6:00 am

They went on sale I believe on Thursday for the Black Friday sales which lasted until Monday.

Strid
Posts: 397
Joined: Tue Jan 08, 2008 5:09 am
Location: Copenhagen, Denmark

Post by Strid » Tue Dec 02, 2008 6:02 am

hmsrolst wrote:Newegg has the 30GB on sale today for $50 after MIR.

http://www.newegg.com/Product/Product.a ... 6820227359
But are they worth $50?

hmsrolst
*Lifetime Patron*
Posts: 426
Joined: Sun Aug 11, 2002 3:26 pm
Location: Arlington, VA USA

Post by hmsrolst » Tue Dec 02, 2008 6:36 am

Strid wrote:
hmsrolst wrote:Newegg has the 30GB on sale today for $50 after MIR.

http://www.newegg.com/Product/Product.a ... 6820227359
But are they worth $50?
It wasn't to me.

Aris
Posts: 2299
Joined: Mon Dec 15, 2003 10:29 am
Location: Bellevue, Nebraska
Contact:

Post by Aris » Tue Dec 02, 2008 8:21 am

Strid wrote:
hmsrolst wrote:Newegg has the 30GB on sale today for $50 after MIR.

http://www.newegg.com/Product/Product.a ... 6820227359
But are they worth $50?
Nope. Id rather use a 5400rpm 2.5" notebook drive than that crap, no matter how cheap it gets. Idunno if i'd even take one for free.

Bar81
Posts: 261
Joined: Thu Sep 04, 2003 4:19 pm
Location: Dubai

Post by Bar81 » Tue Dec 02, 2008 9:25 am

Concur. I wouldn't take one if it was free.

m^2
Posts: 146
Joined: Mon Jan 29, 2007 2:12 am
Location: Poland
Contact:

Post by m^2 » Tue Dec 02, 2008 3:00 pm

OCZ posted some beta quality solution to the issue for 32 bit OSes.
So they might be worth 50$ for some.

Tzupy
*Lifetime Patron*
Posts: 1561
Joined: Wed Jan 12, 2005 10:47 am
Location: Bucharest, Romania

Post by Tzupy » Wed Dec 10, 2008 7:39 am

OCZ and SuperTalent will launch new high-performance SSDs soon:
http://vr-zone.com/articles/ocz--super- ... l?doc=6262
The OCZ specifies upto 64 MB of cache, which should sort out the stuttering issue.

frostedflakes
Posts: 1608
Joined: Tue Jan 04, 2005 4:02 pm
Location: United States

Post by frostedflakes » Wed Dec 10, 2008 4:52 pm

Interesting, hopefully this next iteration will resolve the random write issues of previous drives. Intel needs some competition. :P

hmsrolst
*Lifetime Patron*
Posts: 426
Joined: Sun Aug 11, 2002 3:26 pm
Location: Arlington, VA USA

Post by hmsrolst » Wed Dec 10, 2008 5:39 pm

frostedflakes wrote:Intel needs some competition. :P
If these are in the class of the X25-M, it does seem like Intel will be forced to lower their prices substantially.

bgiddins
Posts: 175
Joined: Sun Sep 14, 2008 1:04 am
Location: Australia

Post by bgiddins » Wed Dec 10, 2008 6:53 pm

$129 USD for 30GB, $249 for 60GB... you'd just get two 30GB drives and stick them in RAID 0 for the $9 difference.

Glad to see there's smaller sizes being made available cheaply - great for a boot OS drive.

AZBrandon
Friend of SPCR
Posts: 867
Joined: Sun Mar 21, 2004 5:47 pm
Location: Phoenix, AZ

Post by AZBrandon » Wed Dec 10, 2008 7:28 pm

Tzupy wrote:OCZ and SuperTalent will launch new high-performance SSDs soon:
http://vr-zone.com/articles/ocz--super- ... l?doc=6262
The OCZ specifies upto 64 MB of cache, which should sort out the stuttering issue.
Looking at the posts up above, there's one saying the Intel MLC drive has 256kb of cache versus 16kb cache on the OCZ Core series. Now this article says the new OCZ has 64mb of cache - which isn't 4 times as big, it's 4000 times as big. This makes me wonder if it's a mis-print in the article, but for comparison, Western Digital's 640GB "black" hard drive sells for only $85 and has a 32mb cache, so it's entirely possible the new OCZ drive does indeed have a 64mb cache, and yeah, a 4000x increase in cache size probably goes a long way to fixing buffer problems.

bgiddins
Posts: 175
Joined: Sun Sep 14, 2008 1:04 am
Location: Australia

Post by bgiddins » Wed Dec 10, 2008 7:50 pm

...so are these available yet? Not sure if the announcement means they can be purchased today.

AZBrandon
Friend of SPCR
Posts: 867
Joined: Sun Mar 21, 2004 5:47 pm
Location: Phoenix, AZ

Post by AZBrandon » Wed Dec 10, 2008 9:49 pm

bgiddins wrote:...so are these available yet? Not sure if the announcement means they can be purchased today.
Good question. I checked their website and found two things. First, the product page is already up which implies that if it's not already available, it probably will be real soon. Also, the product page confirms the cache size really did grow that much. It shows 32mb cache for the 30-60 GB models and 64mb cache for the 120-250 GB models. OCZ seems to have pretty responsive customer service so if you call or email them asking what the release date is, they'll probably get back to you quickly.

Bar81
Posts: 261
Joined: Thu Sep 04, 2003 4:19 pm
Location: Dubai

Post by Bar81 » Thu Dec 11, 2008 10:39 am

Probably quickest just to check their SSD forums, the guys there try to get the info out asap.

Post Reply