How To Prevent SSD Degradation Without TRIM

The problematic:​

SSDs tend over time to degrade both in capacity and speed.​

Capacity:​

As every write to a cell wears it, the more you write to a cell, the faster it will become unusable, leaving you with less space to write to, decreasing therefore the capacity.

Speed:​

Every block of cells that gets a write on it will respond slower if there's already something written on any cell of the block; so the more you write to your SSD, the more blocks will be partially written to, leaving you with slowest writes, decreasing therefore the speed.

The cure:​

SSD's manufacturers, along with O/S's and Storage Controller's manufacturers, came with the following three solutions:​

WEAR LEVELING:
This is an algorithm integrated to the SSD, which guides the data to be written toward the cells that have had the less writes so far, thus, peventing some cells to die way before others, allowing them to wear pretty evenly and, therefore, keeping then the capacity of the drive.

TRIM:
When you delete a file, windows only deletes the index pointer, not the data itself. This is not a problem with an HDD, which will simply overwrite the data, it doesn't make a difference for it if it's empty or not. To the contrary, an SSD will take up to four times the time it takes to write something on a partially written block, as opposed to a fresh block, would the data be marked as valid or not. TRIM makes sure that when a delete command is sent to the drive, the drive actually indeed eradicates the data so the cells are really empty, so the next write to them will be as fast as possible, keeping then the speed of the drive. TRIM was implemented after ITGC, because of it's inherent limitations.

ITGC:
Before the TRIM command was implemented, SSD's manufacturers (Samsung namely) came with Idle Time Garbage Collection. This, as its same says, in Idle Times, when your SSD is not sollicited, Collects the Garbage, the data marked in the index to be invalid but still present on the cells, and consolidate them on single or contiguous blocks; it basically defragments the drive, so the next writes will be done on empty blocks. There may be technically an argument about to wether or not ITGC recognize or not the valid from the invalid data, but it's not significant as would it defragment only the Garbage or all the data, the result bottom line is the same: less blocks partially written, keeping then the speed of the drive !

So, what's this all about you ask ?​

Well, while everyone can enjoy the benefit of the wear leveling, not everyone have the opportunity to work with TRIM, be it because they're in RAID, or simply because either their O/S or their SSD does not support it, and while ITGC is then a must, and every drive most likely has it's own version of it, still, it has it's own limits, and my guess is that nobody told you about it yet...​

Life's Good !!​

So, say you just put a brand new SSD in your machine, first time you use that, coming from a regular 5400RPM HDD, and say YOU DO NOT HAVE TRIM. No matter what brand the SSD is, you'll be simply amazed by the newly speed of your system; you might even not believe it at first... So you start to enjoy your brand-new-system-feeling-like, you play, you work, you download, you install, you uninstall, you re-install, you delete, you name it: _____________, and life's good, you simply love your new system, you almost wish you would never have used anything else in your life: You are now an HAPPY-SSD-BUDDY !!!​

Then, the CRASH...​

So one day, you wake up and decide to benchmak your puppy, just out of curiosity, not necessarily that you noticed some slowdown recently, or maybe just a little tiny bit, but you're somewhat of a Geek, and you like to know your system is working optimally. So you start a disk bench and then, here's the shock, the horror, the frightfulness, you just can't believe it: your perfs are down to almost those of an ancient, antique, unfashionable, dowdy, outdated, obsolete and ridiculous HDD...​

What The Heck Happened ???​

Well, as life goes on, things tend to amass, to pile up, so all those little writes that occured on you SSD, plus all the files you downloaded and deleted after, plus all those temp files that only windows knows their secrets, plus your browser's cache and cookies, and your pagefile's writes and so, combined to the fact that, as nobody told you about that before, YOU AT LEAST ONCE FILLED YOU DRIVE OVER SAY 75%, 80%, 85%, 90%, your call, it's on me _____%, AND NOW, YOUR SDD IS F....., FROZEN, FRIED! and so are you...​

STILL: What The Heck Happened ???​

Remember we've established you DO NOT have TRIM. So all you got is an ITGC, which, as previously said, has it's own limits. So here we go. When your drive is from 0 ~ say 70% full, there's plenty of empty or partially written blocks that can accomodate Garbage or valid data. However, once your disk gets written to over a certain threshold, never mind the overprovisionning the manufacturer has provided, and considering wear leveling works in conjuntion with ITGC, it then becomes impossible for ITGC to find any empty block, so next writes won't find any neither, and then, every single K of write which will from now on occur will take up to four times it should have taken, because they're all going to be written on already partially written block, thus your perfs down to almost the ones of an old HDD, especially with the 4Ks range...​

So, What's In For me ​

Well, thing is that, while it would certainly have served their honesty, it might have prevent some cash to come in, so manufacturers did not broadcast that info upfront, and my guess is that they won't neither. But we, maniacs, fans, whatever, we know we're not dummy customers, so we have to take care of ourselves. So far, if you filled your drive at least once up to XX%, you're as f...ried! as your drive, as far as I know. For instance, back with my first M17X, which was to be replaced for a problem not regarding the two Sammies SSDs Raided inside, I remember I ran a kind of FFs program on them, and this screwed those drives like crazy. Never was ITGC able to get them back on their feet. Basically, this program had written to every single cell of both of them, and they were f...rozen ​

Still: What's In For me ​

Well, at that point, all I can provide you with is the following advice:​

→ NEVER FILL YOUR NO-TRIM ANY-BRAND SSD MORE THAN SAY 70%
OF IT'S CAPACITY, AND YOU'LL REMAIN AN HAPPY-SSD-BUDDY ← ​

I personally use two Intels 160GB in RAID, and was previously using two Samsungs 256GB in RAID, so basically, I never got TRIM to work for me so far. Take it for what it's worth, I never did, and never will fill them over 50% if I can, and for sure, never shall go over 67%. In both cases, I've been able to maintain pretty much my initial performances, WITHOUT TRIM.


Hope this advice shall serve as much and as many people as possible...

 

You're theory assumes the SSD needs ~ 30% of the drive in order to properly execute internal GC, which IMHO seems quite excessive.

Either way, if a non-TRIM SSD can't handle anything over 60-70% then you're actually better off buying a much better quality drive with a lower capacity for the same price. For example, instead of buying a 256GB Samsung, get a 160GB Intel or 180GB OCZ Agility 2 @ $400 for less than or no more than the cost of the non-TRIM SSD.

A secure erase should correct the issue though, but that requires losing all data on the drive, and reinstalling or reimaging (which has its own set of issues) the drive as it was previously.

 

I said 30% ~ 70%, just a guess; I don't know the exact thing here...

And what about if you are in RAID?

ANY SSD in RAID will not handle TRIM, until drives with ATAPI-8 compliancy hit the market...

I have two Intels 160GB in RAID; both supports TRIM, but NOT together...

My whole point wasn't about or not buying a non-TRIM-able SSD; it was more kind of, if for ONE reason or another, you use an SSD in a non supported TRIM environment, you'd be better off not fill it more than 70% of it's capacity.

As far as all cells being used once, it's obvious, would it be only because of wear leveling. Question is where are the cells containing deleted data: among the blocks that contain valid data, or spread among blocks that contain NO other data, which is the worst case scenario; ITGC put those scories among block already containing other data, fake or real, and this cleans up empty blocks for subsequent writes.

But when you fill your drive almost 100%, there no such empty cells (pages) within the partially filled blocks, so they start to use the empty ones cause they have no other choice, so everything clugs, chokes, slows, dye...

 

Just want to state that with even non-raid SSD setups, these minimums are still a good idea.

Maybe I'm just unlucky in the SSD I have chosen a SandForce based Inferno (look in the oxz forums where Bill Gates states that TRIM is essentially ignored on the SandForce drives), but I definitely saw a performance increase from it when I deleted/removed about 14GB from it and am now running with 70% used capacity (from a 93GB available).

When do I start noticing that an SSD has much less 'zing'? At around 40-50% filled capacity. Anything below that and the SSD seems magical.

eYe-I-aie, what is the primary use of your system? Does it endure a lot of writes per day or is it mostly beneficial for the 'snap' it offers in your usage?

 

eYe-I-aïe... I edited my post quickly and you must have snagged it before I got done.

I don't understand why people want to RAID their SSD's, especially considering lack of TRIM. It's like taking two Hemi 5.7L V-8's and making a V-16 out of it. It's definitely bigger and probably a bit faster, but nothing that you usually can take advantage of.

If 50% is truly the max speed factor then SSD's have a lot further to go than people lead on, not to mention double the price point, since you need a drive twice the size of what you really think you'll utilize at max. This makes me want a Momentus XT more and more now. A defragged hard drive can be at 90% capacity with marginal performance difference from one at 10%.

However I would think even if the drive is filled to nearly 100% it shouldn't matter once you've gone back to a 50 or 60% (or less) filled drive. The blocks that are partially filled should be "erased" by the GC routine eventually. Because after I filled my drive, then deleted the "filler data" so I was back to about 20% full, I let it idle for 24 hours and performance came back like fresh. For a while. Now all of a sudden it's giving me this crappy performance. How can you explain that?

 

Basically, people RAID their SSDs to get better performance out of it/them.

It's even a better idea if the drives do not support TRIM as it then makes no difference if you use them in a RAID setup or not, as they cannot handle the TRIM command anyways...

Your motor car analogy does not apply here in my humble opinion for at least this reason:

A V16 is indeed bigger, and most likely faster, but you're not allowed to use it at its peek, at its full potential, simply because there's speed limits on roads and even a V4 is able to reach and even exceed those limits. So I agree with you that there's no point really for doing that.

To the contrary, there's no such speed limits for an SSD, regulation-wise, so the more SSD you RAID together, the snappiest your system will run. Just to take one example, I gained 5 seconds of boot time (5 seconds less to boot) with raided SSDs than with only one. 5 seconds may seem to one like nothing; for me, however, 5 seconds out of 25 seconds is HUGE indeed, it's actually 20% faster, not really something I cannot take advantage of...

I already explained why an HHD won't suffer from being almost filled, and this is because it takes for it the same time to write data on a part of the disk which already contains data (invalid) than it takes to write the same data on a part of the disk where there's no such already written data. That's why you will most likely not experiment sudden slowness with an almost filled HDD.

Again, I don't have the exact numbers (neither Samsung nor OCZ nor WD nor Intel nor whoever would discard them to me), but, as a preventive measure, I suggested to anyone who use a NON-TRIM SSD to not fill it over 70%, to prevent speed degradation. Your call is as good as mine on this one I guess, I said 70%, I myself won't fill them more than 2/3, say 67%. That's the way I've chosen and never got any noticeable performance lost this way. Yours to see if it fits your habits...

For reference, when I was shopping around for my M17X, I've read a whole bunch of news, benchies and what you will about SSDs, just to make sure I'd make the right choice about the one(s) I was to pick. This was back in October 2009. At that time, an Intel G2 160GB was selling for ± $650 which is not exactly cheap. I knew that if I would have chosen this guy, I would have filled it up to 70% right off the bath, and thought it was not a good idea. SSD's choice at that time was pretty much Samsung, Intel and any re-branding vendor with Indillinx controller. I ended-up going with two Sammys in RAID, because Dell were selling them to me $500 each, which was WAY cheaper than any other offer the market had at the time. Except for the FF writing I did once, I never had problem with those drives, but never filled them more than 30% of their entire capacity (± 135GB out of 476GB total).

I chose at that time to go with this setup (2 X Sam) for mainly two reasons:

1. I had read some Intel users bricked their SSD (the 8MB bug) by updating their FW (to get the TRIM command support);
2. I figured that without TRIM, if I would care not filling them too much, ITGC would do the job, and since I was to get ± 476GB total...

And what I thought was what I got.

Now, allow me to try to explain better than what I did so far about ITGC...

I agree with you, one would expect ITGC to erase garbage but, as a matter of fact, it simply can't, all it can do is to DEFRAGMENT the garbage, like the name says, all it does is GARBAGE COLLECTION, it puts together the invalid data, the garbage, on the same block(s), or on already partially written blocks, and it's doing that in conjunction with wear leveling.

So, once your drive gets filled passed a certain threshold, comes a point where there's no such partially written blocks that could accomodate the garbage; therefore, ITGC have to use fresh blocks to consolidate the garbage, so we reach a point where all the next writes will occur on already partially written blocks, thus, the speed decrease.

At that point, even if you delete half of the data on the disk, still, the blocks where the data was are STILL filled with Garbage, with data marked as to be deleted but still present... So the next writes will still be slower because, even if the data is deleted, its still there, but simply marked as "can be overwriten" in windows's index. Of course, ITGC could then offer you a respite if you let it idle for some time, where you got back your fresh drive after deleting alots of data, because it had time to consolidate the garbage, freeing-up some fresh blocks, but as your drive has been almost filled once, the rest won't last.

We can argue till the end of times as to whether or not any SSD should benefit from the TRIM command, and the obvious answer is YES. However, again, not everyone can benefit of TRIM, because RAID or non-supported-TRIM-SSD or older O/S, but thing is that if you don't have TRIM, you'd be well advized to not fill it say more than 2/3 of it's capacity. Now, if you ask me if one should do everything in his power to have his SSD with TRIM, it's obvious, no question about that, the answer is YES !



Tiller, I use my lappy for both work and leisure. Among other things, I write a book, I play some games, a bit of video, audio and image editing (not as professionnaly as you though), Internet, test many soft (so lots of install-uninstall), and others. My intels are filled at ± 33% of their entire capacity, no TRIM (I even shave my head )

Hope this will help a bit, gotta go now !


eYe

 

I thought some ssd manufactures reserve a portion of the ssd for wear levelling like my corsair 60gb , 56gb free after formatting , so 8gb is use for the fat and wear levelling.

see my thread below.

http://forum.notebookreview.com/har...rket-upgrades/515845-ssd-size-explain-me.html

 

60-56 = 4, not 8

Even then, the stated capacity of the disk falls under the same capacity trickery that's been going on with hard drives since their early days. A formatted drive will always show less capacity than what's on the label. A drive labeled as "250 GB" hard drive has ~250,000,000 bytes, which in actuality is 232 GB.

 

My ssd is sold as an 60gb , but i was told it is actually 64gb , with 8gb reserved for, as i said is use for the fat and wear levelling.

EDIT : below is an review of the 120gb version of my ssd , it contained 16 x 8gb memory chips = 128gb

http://www.storagereview.com/corsair_force_f120_ssd_review

 

I think you're mixing up two things (but it might be me as well )


Advertized capacity differs from available capacity...

For instance, a Sammy 256GB will only give you 238GB once formated, and this has nothing to do with the overprovisioning. To the same, an Intel 160GB will give you 138GB once formated. ANY disk, would it be HDD or SSD will lead to that.

Specifically talking about SSDs, some (every?) manufacturers over-provision their SSDs, as Les replied in your thread, just so if you fill them too much, there's still empty blocks. But those are not infinite, so once you filled your drive over a certain threshold, IN A NON-TRIM ENVIRONMENT, your drive is screwed, never mind the overprovisioning.

makes sense ?

eYe

 

All SSD's are 2^x factor 8, 16, 32, 64, 128, 256, etc GB. They just have overprovisioning and maintenance area of some sort which makes the actual useable size smaller, at least from everything I've read.

Also, regarding your theory, my SSD recovered after 24 hours from internal GC (or whatever it does), even though it was filled 99% to capacity then data deleted. It was only after using it for a few more weeks that I started getting this horrible degredation.

 

Just to make sure I get this correct...:

1. You enjoy your drive;
   
2. You fill it almost to max;
   
3. You experiment speed degradation;
   
4. You delete data;
   
5. You leave it idle for some time;
   
6. ITGC does its job;
   
7. You're back up and running;
   
8. Degradation occurs again.

Am I understanding this correct ?

 

Is there anything you can do to fix cell wear?

I bid on a dead ssd:
BROKEN Mach-8 uSATA 1.8" SLC SSD M8SB1-32UC - eBay Internal Laptop, Hard Drives, Drives Storage, Computers Networking. (end time 09-Sep-10 01:17:16 AEST)

I was hoping to use it to test an internal SSD mod using the esata port before I spend money on a good drive (don't want to solder wires to a $100+ drive, without some practise first).

 

Some users report to be able to erase the wear of their drives. I tried once and got horrible results, so there's no way I am going to go that road again. However, some users here might be able to provide you with programs like HDErase, PerfectDisk, Tony's TRIM and all this funny stuff that you could try on your "bricked" drive. As my drives are clean and healthy, I wouldn't touch any of those programs even with a 10 000 feet pole.

Yours to decide !

 

ITGC does it's job and its running almost as fresh as a new drive for a couple weeks just fine, and verified by a couple benchmarks. Then I didn't use it a whole heck of a lot as I was waiting for my replacement SSD with TRIM. I did a couple benchmarks to just verify it was running same or not. Apparently not. I got those poor values one time, but continued to use the laptop for a few days, ran the bench again and still same result. No other updates installed or anything. Basically same setup as when I did the initial testing.

 

In order to understand what happenned, I just need to know the following, and I can't seem to be able to see that in your reply, but I might be too sleepy...

Did the slowness appear BEFORE or AFTER you filled your drive ?

 

After.

But like I said, it seemed to have corrected itself for a couple weeks even. Then all of a sudden its performance diminished significantly.

 

Ok...

Let's take an analogy then...


Say you go to the brand new mall in your car;

You'd like to park your car the closest to the door because you're in a hurry;

Since the mall just opened, there's plenty of parking space near the door, so you're an happy camper.

Life goes on, and life's cool.


However, one day, you're also in a hurry and get to the mall but this time, because so many other persons now know about the new mall, the parking is filled-up, and they even reserved the spaces closest to the door for disabled persons... You then have to park your car much more far away than the previous time, which takes you longer to do your things.


Now, if all the cars in the parking leave for night-time, does that mean there's gonna be no car in the parking the next day ? Does that automatically mean the spaces nearest to the door will be free the next morning when you'll get there ? As a matter of fact, the soonest you'll get to the mall, (right after ITGC did it's job) the more the chances you'll find an empty spot near the door, (to write on an empty block) and this, even if the parking has been emptied in the night...

As time goes on ( as it's been more and more time ITGC did it's full job - as the more you wait to go to the mall, afternoon instead of morning), it's getting more and more difficult to find free blocks to write to, it's getting more and more difficult ot find a free parking spot near the door...

Or, another way: if the spaces reserved for disabled persons are empty, (a block containing invalid data) does this fact give you the right to park there just because the space is free (to write on the block without any downside, like getting a ticket, or taking a speed hit) ?

I fully acknowledge I'm having a hard time to express myself clearly, while I'm doing my very best; maybe with another coffee or so, I'll be able to be clearer..

Meanwhile, sorry for being so nebulous



eYe

 

Sorry, I didn't follow that analogy. Either way, the drive worked like new for a couple weeks after ITGC. Then I BARELY used it and performance DROPPED. I don't think any analogy will explain that.

No disrespect to you, but I think you're guessing and have no solid data to back up your theory. What we need is an individual that is highly educated in the area of SSD's, like an engineer, to explain some of this stuff.

I asked Samsung if they could clarify how GC works and what the user should do as far as idling their PC, when will GC activate, etc? They had no clue what I was talking about.

 

Well, you're right, I'm only guessing here...

And, you're right, I am no engineer and I have no solid data to back me up...

And you're right, we would need such a person...

And you're right, Samsung are useless...

And you're right, my analogy is poorly build...


But...

I just thought I could try to maybe shed some lights, arouse some ideas, and I unfortunately realize that I totally missed it, so I would like to apologize for mixing people up.

Sorry for this time, I'll make a point of not doing this kind of stuff anymore without being fully awared of what I'm talking about, with solid data to back me up, with the help of an engineer if ever possible ...

Meanwhile, my apologies again for this time...

 

Sorry, I meant no disrespect to you. I guess I'm just a bit bitter. I'm not saying you need to be an engineer, and sharing opinions or theories is what these forums are about. But so far ITGC seems to be a mystery that the SSD developers don't care to share at the moment either.

My point is regardless of reason of what's happening, the the OEM's need to share the restrictions with the customer. If utilizing your SSD beyond 50 or 60% results in a significant degredation in performance, they need to share that with us. If filling your drive to near capacity will destroy its performance they need to share that too. They need to share what maintenance procedures customers should follow (like idle time, will it work during PC sleep, etc) for ITGC.

But all I see so far is these holy grail of storage devices that cost an arm and a leg that everyone seems to justify their high costs. Since it was a simple swap with an HDD I (and others) expected it to work like an HDD. There's no documentation or instructions to say otherwise.

I can see from the two smaller SSD's that I currently own (both Intel's, a 40GB and 80GB), that support TRIM, have a noticeable improvement to overall system reponsiveness. I haven't taken any special precautions to make sure their performance remains near new. It just does, which is the way it should be.

 

From watching Intel videos and reading everything I could find about their SSDs I learned that the reserved, wear-leveling portion of the drive must be unpartitioned, never-used space. The IDF (2009) Video makes that very clear. Secure erasing is the only option if you did not initially reserve extra space when you did the original partitioning.

Intel claims they automatically reserve about 7% of the 80GB G2 drives for this purpose and they recommend that when you initially partition the drive to raise that to 10% or more. The longevity of your drive is increased dramatically up to about 27% reserved space and then you reach a point where the loss of space outweighs the wear-leveling benefits.

I knew this beforehand and partitioned my 80GB G2 leaving a total of 10GB (12.5%) free. My OS partition (the only one there except for the tiny System Reserved one) is 70GB.

When I started my initial clean-install of Windows 7 it reported a 74.5GB drive. That reflects the ~7% that Intel reserves. I created a 70GB partition and formatted it for 64K clusters on the recommendations of several people over on the forums at storagereview.com.

All is well, so far. I've only been using it for about 4 days now. I will never change the partition boundary or create more partitions because once a cell has been written to it cannot be used for the wear-leveling process unless you secure-erase the drive. And reports are that after secure-erasing, you don't have the performance of a new drive at all, the degradation being quite evident in both benchmarks and real-world usage.

 

4 months on my MacBook Pro...SSD. No TRIM in OSX, but I've not noticed any performance degradation.

 

That is very interesting information, however, not everyone using an Intel SSD is aware of such IDF 2009 Video and thus, watch it... Point here is that this information should be disclosed along with the drive, Kind of READ ME FIRST type of thing, not ONLY in the Intel Developper Forum... Besides, my guess (yes, it's ONLY a GUESS) is that if you do not reserve such unallocated (unpartitioned) space for overprovisionning, as long as you don't fill your drive(s) say more than 67% of its (their) capacity, you should keep the performance, even without TRIM. That is as a matter of fact what I do experience myself.

Then, why wouldn't the manufacturers UPFRONT overprovision adequately their drives, so we, users, don't have to deal with such manipulations that you, me and others here may deal with pretty easily but for which THE VAST MAJORITY of consumers won't understand ?

That's the exact reason why I suggest to NOT fill your drive more than say 70%, so you still get plenty of cells for either TRIM, ITGC and Wear Leveling to work at their best. On top, if Secure Erasing does NOT bring your drive to initial performance, AND I BELIEVE THAT, well, that's another reason to NOT fill it more than 2/3 ~ 3/4 of it's initial capacity.

But again, the point here is that manufacturers do not disclose that UPFRONT, so some users get confused and if they do fill their drives, they're Frozen...

The only information you failed to provide is this: How full is your drive ?

And nvermind the answer, my advice to you is to NEVER fill it more than say 67% ~ 75% of it's capacity, and you'll continue to enjoy it as you actually do !


Bottom line, SSDs are still a new technology, and manufacturers need to realize they have to provide their customers with complete and accurate information for us to keep the benefits of using their QUITE EXPENSIVE SSDs, and foremost, they should build them so WE do not have to go through some obscure manipulations (leaving an unformatted partition) for them to work at their best.

eYe

 

While some of the above might be true, you are very confused about GiB, GB, and how much manufacturers allow for overprovisoning (hint: it is not included in the total capacity that the drive is marketed as).

Intel does not recommend to increase it to 10% or more by default - I did see the video and it says that it is simply a method to use an MLC drive in some SLC specific scenarios.

A consumer might also follow that strategy but unless they want the SSD to be used at normal consumer levels for the next 50 years, what is the point?

I have seen another Intel SSD centric video about a year ago where Intel specifically stated that as long as you don't use the full capacity of the drive (no matter how it was partitioned), all the free nand area is used as spare - in addition to the already included 7% and 27% (MLC, SLC, respectively) spare as delivered.

 

Thanks, this is the overprovisioning, forgot to mention that...

Those cells will then be used for both Wear Leveling and ITGC, providing there's such a micro-program on Intel drives (and others apart Samsung), which I believe is true but cannot prove though...

Thanks for the clarifications Pal !

 

I understand the binary math of GiB/GB and the marketing strategy of drive manufacturers. So, you are saying that the originally listed unallocated space on the drive (74.5GB) does not represent 5.5GB of "reserved" space for wear-levelling? That is what I assumed since they say that about 7% of the drive is reserved and, I assumed, unavailable for partitioning.

But, the math supports you. 80 billion bytes / 1024^3 = 74.5GB. So where is this "reserved" space? Is it invisiible to Windows and only seen by the controller? And did I simply waste space by creating only a 70GB partition for my OS? My intention was to increase the wear-levelling area to 12.5% (10GB). But here again, perhaps my math was faulty.

I am absolutely certain that in either the video or somewhere on Intel's website it was stated that only unpartitioned, "never-written-to" cells can be used for wear-levelling.

 

That is exactly my understanding, there are cells hidden (on top of the capacity (real or advertized) available to partition), that they call overprovisioning, but it seems there's not enought of them as if one fills his drive almost to full, wear leveling seems to be unable to do it's job...

My guess is that this space is still available for wear leveling anyways...

just my 2 cents though...

 

Thanks for that.... I'd really hate to be wasting space on the drive. I partitioned for (slightly) increased longevity.

It is absolute negligence on Intel's part not to make this clear to consumers. I bought a new, OEM drive from Amazon and there were no instructions or advice of any kind in this regard... or, in ANY regard. I suppose they expect you to just slap it into the bay and forget about it. That's fine for the "average" consumer, but some of us would really like to get our money's worth from the product and make it last. I don't plan on buying a Gen 3 drive. I want the one I have to last several years. It wasn't cheap!

 

This is not specific to Intel; just ask htwingnut what he thinks of Samsung...

Basically, he's bringing to the table the exact same concern you state: Why the heck don't the manufacturers provide us with clean instructions about HOW-TO service/maintain/preserve the health of those pretty-expensive-puppies of ours...

Again, SSDs are still in the early stages of their lifespan, so, maybe they did not even see that coming; but as of today, they have no excuse anymore, they should know, and they should let us know...

 

LOL yeah. Samsung won't be finding its way into my PC's any time soon.

Besides giving us basic maintenance info they could also provide guidelines for benchmarking for real world performance. All this data with sequential, 512K, 4K, 4k QD32, 4K-64thrd mean little without how it affects real word performance.

It would be nice to see them categorized by boot times, file copy, application loading, gaming, or whatever. But that's probably asking too much.

 

what about those 2 articles

The 'write cliff' effect
Storage smackdown: Hard drives vs. SSDs

Higher-endurance NAND in MLC SSDs
Israeli start-up claims huge boost to SSD reliability

 

Just as a minor nitpick, the fact that you bought an OEM drive is more probably the reason for lack of instructions and manuals. After all, the point of an OEM drive is a bare drive that's intended to be sent to manufacturers/resellers/distributors to be put into and sold as part of other machines. As a result, they lack the manuals and packaging of a retail drive, which is also why they're cheaper.

Now, this is not to say that a retail drive will necessarily have said manuals and instructions. They probably should, but whether they do or not is, naturally, up to the manufacturer at that point. Still, the point I'm trying to make is that when you buy an OEM or bare drive, you should not be expecting much if any documentation at all.

 

I own an Intel 40gb and 80gb and they have zero documentation to this effect. OEM or not it should be freely available through their website. Buying OEM or retail should have no bearing.

 

My Corsair Force F60 gb, came in retail packing and it has zero instructions , just the SSD features printed on the packing.

 

That wouldn't make sense as boot times, file copy, application loading etc varies depending on the computer, whats on the computer and the person using the computer.

Measuring the sequential transfer rate of a single 100mb test file for instance makes it easier to compare drives because the data is the same and it is being read/written under constant conditions. Its just important to realise that in practice, sequential read/write speed is not an important attribute to look for in an SSD because mechanical hard drives are good at it too, thus you won't notice much of a difference under these conditions. This is because the read/write head in a mechanical hard drive doesn't need to move around alot to retrieve or write data. A sequential read would be like the stylus on a record player just following the groove all the way to the end whereas a random read would involve lifting the stylus off the vinyl and moving it to another part of the vinyl. The sound has to stop reading and theres a delay before you put the stylus back in a groove to resume playback.

One of the biggest improvements with SSDs is the removal of the read/write head and the spinning platter which largely eliminates seek times. This is why people get all hot and bothered over small random read/writes because these are traditional weaknesses of mechanical drives. If your ssd has good iops and random read/writes then it'll be fast in a way thats noticeable over a hdd.

But when comparing random read/write benchmarks alot depends on the compressibility of the data so you sometimes see people saying that you shouldn't really bench a Sandforce drive with AS SSD because the test data is not very compressible (in contrast with ATTO which is highly compressible and naturally recommended by all the Sandforce partners and distributors as these drives deal well with compressible data and ATTO benchmarks give them more favourable results).

In the end I'm just happy with the ssd I've got because its silent, its noticeably faster than my old hdd in normal use and I've got a nice long warranty. Its a step up in every way so what more could you possibly ask for?

 

Any comments on this solution?

Guide A simple guide for speeding up EOL OCZ SSD's..

 

That looks very interesting....

BUT...

I am not going to be one of the first ones to test this solution, mainly because I once scrapped my 2 Sammys in RAID with his Tony's Trim program...

Perfect Disk might be the right tool to defragment/consolidate/erase SSDs, however, I will wait till there's a WHOLE bunch of a LOT of people with different SSDs that have tried it and reported that everything is so cool that they wish this tool was available before SSDs themselves...

Sidenotes:

This guide is aimed at people who uses OLD OCZ's SSDs, meaning, no TRIM, or RAID and so and so...

As he states it, it has to be used (performed) 1 or 2 times every week... (hello wear levelling !!!)

As a rule of thumb: if it's not broken, why is the fix for ? (In this case, if the drive is performing well, why wear it to gain little or nothing?)


I fully understand that Raxco (who makes Perfect Disk) are trying to come out with the first SSD (good-working) defragmenter, but until I see SSD's manufacturers stating that this tool is " approved" to use with their products, I don't think I'll play the guinea pig and risk to scrap my Intels which, again, are in RAID (so no TRIM) and only took a small, tiny performance hit since I bought them ± 3 months)

However, again, this looks pretty interesting anyways; wait and see eye guess...

Thanks for posting your question here !

eYe

 

eyeiaie...

Have you ever successfully done a image backup and restore onto the same SSD? I hear mixed results. Seems like if its off the same SSD it should be fine as long as it was aligned properly the first time?

I ask this because I don't want to waste Windows 7 activations for a fresh install, but would rather just do an image of a good clean install (win 7 activated) and when needed, do a full image recovery.

 

Nope !

Makes sense !

Do you mean there's a limit to the number of times you an activate the windows version you bought ???

I would by far prefer to clone back myself than reinstall; just a matter of lazyness in my case...

And I do read the same mix results you see yourself; sorry for not being more helpful here pal...


I used (long ago) to have my very-little-own-simili-RAID-solution in an lappy where I had two 60GB disks each divided in four partitions:

D1P1: Windows→→→12GB
D1P2: Programs→→→ 8GB
D1P3: Temp area→ → 2GB
D1P4: Data BKP→→→38GB

D2P1: DATA MAIN→→38GB
D2P2: SWAP area→ → 2GB
D2P3: Windows BKP→12GB
D2P4: Programs BKP→ 8GB

It was almost as robust as a RAID-10 solution...
OK, maybe a RAID-5 solution...

Anyways, thing is that if ANY of my disks would happen to fail, I was able to get back on my feet instant, as I did have everything on every disk; just boot onto my norton ghost's diskette (ya, you read correct), and that was cool... ...

Old antique technology...

 

Yeah, MS just reduced number of keys available per product from 10 to 2 for standard technet subscription and from 10 to 5 for professional. So as you can see, significant change.

 

Purchase Diskeeper Home with HyperFast Online | Diskeeper

Anyone have input into this product? My SSD is an OEM Toshiba with no TRIM support.

Thank you

 

I have been using the trial version for a while , only the hyperfast feature works with ssd`s , and it seems to keep my ssd fragment free.

 

I thought there was no fragmentation of an SSD.

 

Actually, there IS indeed fragmentation on SSD as on any disk; however, since there's no seek time for the head to move from one sector to another, SSDs are almost LATENCY free, and thus, fragmentation is just simply NOT as problematic as with an HDD, since it's not taking longer for the SSD to read data on blocks not beeing contiguous, as opposed to an HDD, which has to move his head to read non-contiguous sectors, which increase the latency.

Bottom line, because fagmentation on SSDs do not cause latency problems like on HDDs, some may say there's no fragmentation, which is false; simply, it does not hurt as it does on HDDs.

eYe

 

A bit of marketing hype or not

http://www.diskeeper.com/hyperfast/index.aspx

 

Sorry to resurrect an older thread but I thought it might be helpful to put this info here...

I always do a clean install on SSDs to be SURE I have proper partition alignment. But I don't waste the activation key and license I paid for, I just back it up and restore with ABR (Activation Backup and Restore). I bought a laptop recently and used this program to backup my license certificate and product key and then after clean-installing on the SSD I just ran the restore. Works perfectly. Just be sure you install the same EDITION of Windows 7 that the key and license are for (HP, Ultimate, Professional, etc.).

 

Thanks +1. I'll have to try that.

 

I use my ssds since years without trim. no noticable performance degradation happening. there sure is one if i would measure. but why should i care? it works as expected.

 

Subscribed