I can have a HDD turned off for a year and still have all the data intact when I turn it on again. I can use a HDD in a video editing system and not have it die within a few months time due to the massive writes. Also, Flash memory has hit a brick wall already with how far it can scale. At this point 34 nm MLC Flash has 3k writes/cell and a data retention of ~1 year. 24 nm will reduce this to 1-1.5k and ~8 months. MLC memory even at 65 nm already suffers from massive latency fluctuations and writes to one cell affect those surrounding it. NOR Flash is already being replaced by Phase Change Memory (PCM) and NAND Flash will be the next target. Maybe PCM SSDs will end up replacing HDDs. Who knows
SSDs gradually (like in many, many years) start to lose cells due to being written to too many times. At which point your drive will begin to creep smaller and smaller. This is failing due to wear levelling. Anything else is just a product failing, the same way a normal HDD, car or microwave can fail.
Even multi layer cells can sustain a hefty amount of write cycles. Just because the possibility is there doesn't mean you should worry about it. A review site did the math. Most multi layer cells have 20k write limit but there are many cells and the SSD uses these cells in rotation to keep the durability up. If you have something like a 80GB SSD, you can write in excess of 400GB onto it and still have to last 10 years. Chances of controller failure is probably higher. I can't imagine anyone writing 400GB on a SSD on a 80GB SSD ever. Thats why manufacturers can pack a huge warranty on those things. SSDs can lose write performance from being full. This is because SSDs have to do a delete cycle and a write cycle to replace information. Once a SSD is degraded, it wont degrade further, its not like they can degrade off the face of the earth. The only true way to restore 100% of its speed is rewriting the drive with binary zeros. TRIM does a good job too but not as good. We are only talking about write cycles here. The important stuff like random read doesn't degrade.
OMG. This has turned into exactly the kind of indepth thread i hoped for even thou i still dont understand half of it. An so many answers in so short time. Thanks a lot everyone. Please tell me if i got this right !! So on XP it all comes down to TRIM. Running wiper.exe reguarly would be the best ( only ) option for drive maintainence ? So the estimated MTFB for SSD is pure bul1sh1t ? .... What is it exactly that makes these drives die so "quickly" ? Seems ive read a lot about dead drives. I honestly thought that it only had to do with reformating it reguarly. Even some of the answers here lead me to believe this. I thought that it had to do with the way SSD´s read and writes blocks. So my common sense made me think that a format would reorganise all blocks leaving the drive as if it was all "new". ? So it will eventually make the drive perform slower with time anyway ? And yes, i mostly needed it for a totally silent pc. No fans at all. So no SSD for my XP server. Might just update this to win7. Let me think 5. Another question came up out of one of the answers or the debate in this thread. I also wanted SSD´s because i thougt that data was safer to store on SSD´s ( Cause its ram ) than on HDD´s. So its not afterall ? 6. The 2.5" of the drives is an issue for me. Now it seems that im better of running 4 2.5" 500GB HDD´s in raid1 than buying SSD´s just yet as the technology is still "this young". Any suggestions on some good 2.5" @ 500GB disks for a desktop ? And all you guys thanks for all the answers!! Please keep the debate rolling in here. im learning a lot !!
SSDs are no more reliable than HDDs. If anything SSDs are less reliable because any kind of failure means that your data instantly vanishes. Only HDDs have the option of data recovery by replacing the controller PCB, or doing platter-level recovery. In terms of MTBF it should be noted that SSDs haven't been in existence long enough to prove their '100,000 hours', not to mention that the type of usage can have dramatic effects on this MTBF. HDDs are far more consistent in this regard. I have always been partial to Seagate and Samsung when it comes to HDDs.
Oh crap. I thought of SSD´s as the lost ark till i read this. So i bought me 4 Seagate Momentus 7200.4 ST9500420AS. Ive had a few seagate drivs failing on me some years ago. Lets se how these will last. Thanks for the respond Elledan So do i understand these correctly ? So on XP it all comes down to TRIM. Running wiper.exe reguarly would be the best ( only ) option for drive maintainence ? So the estimated MTFB for SSD is pure bul1sh1t ? .... What is it exactly that makes these drives die so "quickly" ? Seems ive read a lot about dead drives. I honestly thought that it only had to do with reformating it reguarly. Even some of the answers here lead me to believe this. I thought that it had to do with the way SSD´s read and writes blocks. So my common sense made me think that a format would reorganise all blocks leaving the drive as if it was all "new". ? So it will eventually make the drive perform slower with time anyway ? And yes, i mostly needed it for a totally silent pc. No fans at all. Thanks in advance
As bit-tech's resident SSD expert, allow me to clear this up properly 1) Even when you're not installing/writing files to your drive, you'll still be writing to it regularly as part of just running XP. The page file, temporary files, booting up, all write data to the drive. In XP you can use the wiper.exe to manually initiate TRIM though, so as long as you run it now and then (say once a day) you shouldn't hit any performance issues. Of course, your SSD will need to support this, but Any Indilinx or Intel based drive will fit the bill, as they all have manual TRIM tools. Wiper is only a short term fix though, so will need to run it regularly. 2) While NAND does have a set number of read write cycles, the number is in the tens of thousands. I spoke to an SSD manufacturer regarding this and they said you'd need to be thrashing your SSD 24hrs a day for over 10 years to hit the limit. Nothing to worry about. Of course, all consumer electronics are susceptible to hardware failure, but you should always have backups of irreplaceable data. The fact that you can't recover data from an SSD isn't a convincing criticism - theier storage capacity is so low it's easy to back up the contents. 3) Writing files to a drive once won't make it slower. It's the process of re-writing and moving cells of data on the drive that causes performance degradation. Dropping your MP3s on your SSD won't slow it down, although it will be a bit of a waste, why use an SSD for bulk storage? 4) Using an SSD in a server is fine, especially if all it's doing is dedicated reads. Again, you can manually use TRIM via wiper.exe or Intel's toolkit to keep performance up if you start chucking a load of re-writes at it. I would seriously consider making the move to Win7 though. It's been made with SSDs in mind, and automatically detects an SSD and disables defrag etc automatically. Not to mention supporting TRIM as standard, which is a BIG plus. Seems a shame to drop £160+ on an SSD and not go for a £75 Operating system upgrade.
34 nm MLC Flash has ~3k write cycles, as stated by a company official. The figures they gave was more like 20 GB/day resulting in a 5 year lifespan. The size and location of the data being written (many small files versus a few large ones) will of course also heavily impact this. Believe me, it's perfectly possible to completely trash a modern SSD within a week with the right writing strategy. Just get me an SSD and I'll unleash an app on it
The reliability of a PC device is usually seen to be the chance of a failure. In this case, even low end consumer SSDs have a lower chance of catastrophic failure, even at this immature stage (according to Dell, this is), and manufacturers rate them as such. Google has found that the annual failure rate of hard drives is around 8%. Explaining how you can replace the platters or controller board on a hard drive is almost irrelevant because if you're prepared to spend the thousands it costs to have the platters changed by a qualified technician in a clean room, you should have just bought an enterprise level hard drive, which would be a lot more reliable than any normal hard drive, or even an enterprise level SSD, more reliable still. Intel's MTBF is actually 1.5 million hours, this is obviously worked out mathematically. Does this mean that hard drive MTBFs are any different? Of course not! The MTBF of the Samsung Spinpoint F3 is 1.2 million hours, a very convenient 50,000 days, which implies that you should be able to use the spinpoint f3 continuously for over a century. That's not really true, either. Finally, you'd be mad to use an SSD as a storage disk. SSDs are currently only economical in small amounts, such as 80GB, which means that you can put your boot stuff and games on the SSDs, but your important data on hard drives. In general, no one hard drive manufacturer is better than another. They all make their mistakes, and over the years they've made different products. I've got hard drives from Samsung, Hitachi, Seagate, Maxtor and Western Digital that are between six months and nine years old, and they all work fine. Buying a new product because you had a good experience with an old one is not such a good idea in the hard drive world, because they're all basically the same. You're only getting significantly better quality when you get enterprise level hard drives.
It depends on the drive. Even a number as low as 3000 write cycles (IIRC the Jedec spec specifies 10k write cycles as a minimum) will still last an absolute age on a 1TB SSD with normal usage, whereas if you have a 20GB SSD it won't last nearly as long. In terms of destroying a drive in under a week, it's perfectly possible to do. I know how you think you'll do it, ie lots of 2kb writes, thus causing all the blocks to be repeatedly modified. The problem is that it's not representative of normal usage in any way. If you set out to destroy anything through use that is not representative of normal use, of course you'll manage it easily. SSDs are good for boot drives and enterprise storage. If you want a big swap disk, I wouldn't put an SSD up for it, but otherwise SSDs are good for normal use and will provide a big speed boost.
Enterprise storage? Only thing I have seen SSDs used for in enterprise settings is for database query caching and similar high IOPS tasks. And on a sidenote, it's perfectly possible to destroy an SSD within a week or so by writing lots of big files to it as well. Remember that 20 GB/day means 5 years lifespan. What if you were to write not 20 GB but a few TB a day? There are only a number of limited use cases where SSDs actually make sense.
I'm ignoring the Intel number because that factors in small file transfers as well which would not be part of your algorithm. transfer speed: 70MB/sec rated size: 80000MB time to overwrite entire hard drive: 1142.857 seconds decimal value of percentage of day to overwrite hard drive: 0.0132175 number of cycles: worst case 3000, average case 10000 worst case time to kill disk: 39.6525 days average time to kill disk (if you're listening to JEDEC): 132.175 days So no, with large files you wouldn't be able to kill the drive that quickly. Again, is this normal use? There are plenty of uses for SSDs in enterprise environments. Manufacturers are generally billing them as replacements for the traditional 15k rpm hard drives, since they use far less power. Furthermore, they last a lot longer because they're generally SLC NAND - the best of which lasts around 200,000 cycles.
Wow, this does sound trustworthy. Thanks a lot for clearing all this up. I would use 1ssd for OS and another for music/ musicvideoes only for a totally soundless pc. On another pc im running crossifre with some rather old x1950 pro GPU´s. These arent 100% supported by win7. So it wouldnt be os only i would have to pay for. Im borrowing a win7 OS from a friend in a few hours to se if the GPU´s will work with what i need of apps. Cause i can see that SSD´s is exactly what i want. Once again thanks for the indepth explanation.
If SSDs were a competitive price vs HDDs I suspect everyone would be using one, but it does seem to be the case that in bang for buck terms there isn't a mainstream reason to use one currently and mostly those getting one are doing so for a less regular reason (early adopters, silent PCs etc as well as people who don't really get it but feel the need to grab one anyway). Are there any benchmarks of normal use between a regular HDD and an SSD in operation for slightly higher (ie not cannot turn it on without a label but not splitting the atom on it) users?
Here's the boot time between various SSDs and a fast hard drive. http://www.bit-tech.net/hardware/storage/2010/04/12/corsair-nova-128gb-ssd-review/10 As you can see, the SSD boots up in much less time than the hard drive.
Thanks, that's an interesting link. Feeling a bit foolish as its bound to be on there somewhere, but what's a clean vs dirty bootup (in the non Benny Hill sense)? It does seem a lot faster, though I am not sure that 71 seconds is a terribly long time to wait, I guess that goes down over time having installed everything. The reviews are interesting too..missed those. I still think for a mainstream gamer the difference might be spent more profitably on a graphics card uplift, if buying a new PC vs an SSD, But that may be coloured by having researched a new build lately.
The point is that with an SSD you're loading everything up faster, not just Windows, and most people will support a significant boost in just about everything. The difference between 'clean' and 'dirty' is basically that as SSDs age, they would traditionally get slower. This is because SSD blocks can't be modified, they have to be read and then rewritten with the new data inserted, which means that they can be a lot slower. With Windows 7, TRIM has been added, which is a way of fixing this. It makes sure that as few of the cells as possible need the read write process by getting Windows to tell the drive when a file is properly deleted and suchlike. The 'dirty' workload is used to make sure that the performance doesn't suddenly degrade once you've started using it. Relevant pages: http://www.bit-tech.net/hardware/2010/02/17/kingston-ssd-now-v-series-128gb-review/2 http://www.bit-tech.net/hardware/storage/2010/02/04/windows-7-ssd-performance-and-trim/1
