SSD Discussion - Split from T410 First Impressions Thread

[Navck]

Just a personal recommendation, but don't get a SSD.
1. Immature technology
2. 1,000-10,000 (Realistic) write cycles per cell, wear leveling only mitigates the quick death
3. Power consumption is as equal to HDDs in similar situations
4. Heavily advertised to sucker people into buying one, expensive per GB

I got my T410 to boot in 22 seconds with the 320GB-7200RPM option

Defragging all files and moving them to the end of the disk combined with system optimization will let you beat SSDs all day overall.

[khtse]

Just a personal recommendation, but don't get a SSD.
1. Immature technology
2. 1,000-10,000 (Realistic) write cycles per cell, wear leveling only mitigates the quick death
3. Power consumption is as equal to HDDs in similar situations
4. Heavily advertised to sucker people into buying one, expensive per GB

I got my T410 to boot in 22 seconds with the 320GB-7200RPM option

Defragging all files and moving them to the end of the disk combined with system optimization will let you beat SSDs all day overall.

I don't want to get into the SSD versus HDD flame war, but you last comment is very wrong. No defragging or optimization will make a 7200RPM as fast as a proper SSD (except perhaps those early MLC SSD with Jmicron controller).

I have been using SSD since the first Intel MLC was released. There are a lot of hardware upgrades that cost a lot, and 90% of the users won't notice the difference, like CPU upgrade or RAM upgrade. But the speed improvement going from a 7200RPM HDD to SSD is so huge that everyone can notice the difference. EVERYTHING runs faster and snappier.

It is a relatively immature technology, but it's mature enough for a lot of people to put them into their main work computers. You are grossly understating the write cycle per cell for SSD, and that the range of write cycles may well be right, but is not realistic at all. SSD does not work by repeatedly writing into a single cell, wait till it is dead, and then move onto another one. Leveling algorithm works by spread read/writes over all the cells on a SSD and thus the write cycles of a typically SSD, realistically speaking, is hundred to thousand time more than write cycle per cell, and these leveling algorithm does not mitigate only just quick death.

It's true that most SSDs in the market do not exist long enough to see how long they will last, but don't forget that HDD is never the most reliable part on a computer either, espcially on mobile platform. I have been using 3 SSDs on various computers, the oldest one being the Intel X25M G1 (1 year + 4 months), and none of them has failed yet. In fact, I have a 1TB desktop HDD that I used for storage failed during this period and it is less than a year old. Another advantage is that SSDs are silent. If you always work in a quiet environment like me, the lack of grinding HDD noise can already be enough to motivate you to get an SSD.

I agree with you that the low power consumption promise of SSD does not translate into significantly better real world battery life, but SSD definitely consume less power than HDD. It's just that laptop HDD is not very power hungry to start with. The only real con of SSD is price. They are very expensive compared to HDD, $/GB, and I really don't recommend getting one if you need 200GB+ storage on a laptop. But if 128GB or 160GB is enough for you on your on-the-move laptop, and you have $350-450 to spare, you should get a SSD.

[ThinkRob]

Just a personal recommendation, but don't get a SSD.
1. Immature technology
2. 1,000-10,000 (Realistic) write cycles per cell, wear leveling only mitigates the quick death
3. Power consumption is as equal to HDDs in similar situations
4. Heavily advertised to sucker people into buying one, expensive per GB

Wow. So much FUD, so little fact.

1) Immature? Well... yes, in the sense that 802.11n, LED-backlit screens, and e-Ink are immature. As with those examples, the actual tech's been around for quite some time, but the current application is only a few years old.

2) Technically true, but wildly misleading. Yes, each cell has a limited number of writes -- but that's why we have ECC and wear-leveling. Sure, if you wrote to the very same cell every single time for 10K writes you'd kill it -- but that's exactly what a sane controller avoids. Intel's X25-M drives are rated for something on the order of 20GB per day, 365 days a year for 5 years before the first cell fails. Given that modern HDDs are only warrantied for 3-5 years, I don't see SSDs as a step down longevity-wise.

3) Only for early-model drives. Good MLC drives have lower draws under load, plus massively lower idle consumptions. Furthermore, since they can write/read much data much more quickly than a spinning-platter drive, they spend more time in the idle state (and thus spend more time at a lower rate of power consumption.) Add that to the fact that they don't have any spin-up (and thus no spikes in power draw due to said spin-up) and you can see why a good drive yields an improvement in battery life.

4a) Yes, they are advertised. That's not to say that everyone who buys one is a sucker, nor that everyone who buys one did so due to said advertising. I've got three SSDs at the moment, and not a single one was purchased in response to advertising. I purchased them because their random IO performance and decreased power draw meant that I can get more work done (since a lot of my work is IO-bound) and stay unplugged for longer. Given that I get paid by the hour, this means that they have paid for themselves several times over already.

4b) This is your only 100% correct point: they are indeed more expensive per GB. This will change with time (all tech. gets cheaper at some point), but for now they certainly are much, much more expensive than magnetic media as far as the price-per-GB goes. Still, even at current price points they make good financial sense for some applications (such as fileservers, DB servers, dev. boxes, etc.)

I got my T410 to boot in 22 seconds with the 320GB-7200RPM option

If my X200s booted in 22 seconds, I'd wonder which one of my init scripts was broken.

I just timed my X200's boot speed. From pressing the power button to the GDM login prompt, it takes exactly 21 seconds. That includes not only the time it takes the BIOS to POST, but also the three second GRUB timeout, as well as the time it takes for me to type in an 80 character LUKS passphrase.

Defragging all files and moving them to the end of the disk combined with system optimization will let you beat SSDs all day overall.

That's complete and utter BS. Sorry to be so blunt, but your claim is just patently false. Yes, doing that will allow you to beat a poor SSD (such as one of the JMicron-based disasters) under a certain set of circumstances (random writes or reads under certain loads), but it simply won't matter if you're squaring off against an X25M or (even worse) an X25E. There's just no contest. The X25 will completely and utterly destroy the HDD when it comes to random IO, and will best even the fastest HDDs in sequential transfers.

[Navck]

I would like to ask you to take a 1TB OCZ SSD with multiple controllers (Imagine the super fun overhead) againist three striped VeloRaptors but that wouldn't be fair would it? (Desktop application, but the price range is the same including RAID controller for the Raptors, not just a mobo solution)

I also would like to bring up that no matter how the manufacturer's marketing teams (You know, the ones who don't let you see the internal specsheets?) wouldn't want you to know that realistically some of their drives *do* get fragmented in a specific way that *does* slow SSDs down (They do their own equivilent of a defragmentation process*, you'd have to ask some of their internal engineers on how that works, some SSDs don't and you'll be going to hell within two years of heavy usage.) Did I also mention SSDs still have *technical* seek time because they don't run natively in Windows? Whats this about all this wear leveling, controllers? Still have to deal with those. The time it takes for a harddrive to move its head a milimeter from the edge of the disk in a little? Sub 1ms. You can still beat SSDs when you have the proper disk defragger and have it arrange files properly.

Edit: This is all ignoring all the "dies after a few years practically", you do know the SSDs that are avaliable to consumers are binned with the lowest grade of flash memory from the manufacturers and some of them only rate *the first _____* (Depends on manufacturer, some only assure "x" cells, others do "across these chips, that one cell per...") will be able to survive 10k rewrite cycles (You know, not like over time every little NAND cell will be happy about that?) Any of the stuff you read is the stuff the SSD mfgs feed you as marketing. If you can get their internal specsheets, you get a different story ("Binning" LED manfuacturers understand this conecpt too.) When I said overall, I mean *overall* as a storage solution as well.

LEDs are not mature, using them as a backlight is a relatively immature technology. (I'm sure you'd like to read about how you can produce 200-300 lumens/watt and the thrilling uses of high performance flashlights.)
Also try comparnig a WD Scorpio Black's power usage to a SSD then try to imagine some of the conditions that you can go under. Not necessarily more power efficient for the SSD. They also aren't in deep states of idle, you should see all the fun sorts of things that happen when you ask one to delete a file (Goes into query, SSD decides when it wants to wipe it off that sector.)

[blackomegax]

I would like to ask you to take a 1TB OCZ SSD with multiple controllers (Imagine the super fun overhead) againist three striped VeloRaptors but that wouldn't be fair would it? (Desktop application, but the price range is the same including RAID controller for the Raptors, not just a mobo solution)

I also would like to bring up that no matter how the manufacturer's marketing teams (You know, the ones who don't let you see the internal specsheets?) wouldn't want you to know that realistically some of their drives *do* get fragmented in a specific way that *does* slow SSDs down (They do their own equivilent of a defragmentation process*, you'd have to ask some of their internal engineers on how that works, some SSDs don't and you'll be going to hell within two years of heavy usage.) Did I also mention SSDs still have *technical* seek time because they don't run natively in Windows? Whats this about all this wear leveling, controllers? Still have to deal with those. The time it takes for a harddrive to move its head a milimeter from the edge of the disk in a little? Sub 1ms. You can still beat SSDs when you have the proper disk defragger and have it arrange files properly.

Edit: This is all ignoring all the "dies after a few years practically", you do know the SSDs that are avaliable to consumers are binned with the lowest grade of flash memory from the manufacturers and some of them only rate *the first _____* (Depends on manufacturer, some only assure "x" cells, others do "across these chips, that one cell per...") will be able to survive 10k rewrite cycles (You know, not like over time every little NAND cell will be happy about that?) Any of the stuff you read is the stuff the SSD mfgs feed you as marketing. If you can get their internal specsheets, you get a different story ("Binning" LED manfuacturers understand this conecpt too.) When I said overall, I mean *overall* as a storage solution as well.

LEDs are not mature, using them as a backlight is a relatively immature technology. (I'm sure you'd like to read about how you can produce 200-300 lumens/watt and the thrilling uses of high performance flashlights.)
Also try comparnig a WD Scorpio Black's power usage to a SSD then try to imagine some of the conditions that you can go under. Not necessarily more power efficient for the SSD. They also aren't in deep states of idle, you should see all the fun sorts of things that happen when you ask one to delete a file (Goes into query, SSD decides when it wants to wipe it off that sector.)

you must work for a hard drive manufacturer.

SSD's dont run natively in windows?
a) both HDD and SSD channel their data over SATA controllers via the same methods.
b) windows 7 has very specific support for SSD's

Also, you can not physically defrag an SSD. they simply do not work like that. you can LOGICALLY defrag it, but the actual bits will still be effectively random due to wear leveling. access to these "random" sectors is all maintained by the disk's controller and essentially has no visible seek time.

Other than the interface and the function, there is zero similarity between HDD's and SSD's. even data recovery methods are thrown out the window.

As for binning, intel makes their own chips, and only has a few models of SSD. The SLC chips are what they market for high end these days. MLC has come a vastly LONG way as a consumer tech.

\in all fairness I still use HDD's and keep any important data mirrored, and would do the same with SSD's. Unless you're dealing with military hardware, it's ALL a crap-shoot.

[Navck]

If I worked for a harddrive manufacturer, then I wouldn't of been trying to find a way to find a way to stack EPP+Educational discounts for this T410 a week back would I?

Also, go find a datacenter or servers that uses SSDs for high performance applications. They don't use them because of lifespan being *gone* within two years with capacity.

SSDs *do not* run natively in Windows 7, if they did then you wouldn't have wear leveling algos built into the controller would you? They are still treated in a similar way as harddrives.

SSDs *have* seek times. Period. This is all controller level stuff, it might physically have insignificantly small seek times but harddrives deal with this too. You still have to deal with the controllers. You think wear leveling is a free action? Thats something the controller has to do, as with the DELETE QUERY. There are *many* operations that the controller on SSDs have to handle, just as harddrives need their own controllers (Speculating: This is why WD did that dual controller/processor thing on their RE3s)

Find your personal SSD buddy who can tell you about binning, I can summarize it like this:
Super rejects: Flash sticks (You know those super cheapie 4-16GB things on sale for 9-20 dollars? Bingo.)
Rejects: Consumer drives (Esp. Retail outlets)
Good stuff: Industrial applications (IE: The drives that are bought ONLY in bulk by companies and not avaliable to you.)

SSDs *do* fragment on their own level, they might not "appear" fragmented to you but the controller is dealing with that crap individually with each chip within the "magical solid box" right there, some controllers have shoddy firmware that doesn't optimize for that and leads to slow performance!

Harddrives and SSDs also share one thing, they're vulnerable to heat, SSDs are moreso vulnerable to heat than harddrives. They heat up faster during sustained write ops too.

The reason I say this is because of the marketing hype from the SSD makers trying to FUND their own operations by releasing early tech to the consumer market to sucker money for R&D so they can release a competitive product that (... Decades?) can sustain rewrite cycles to each cell to the level harddrives can, solve data retention and all sorts of issues.

[yak]

Guys, take it easy or take your flame wars elsewhere. This is a topic about T410 impressions after all.

[blackomegax]

Yeah, i just don't get the hostility towards SSD.
it's a perfectly fine tech for what it's normally used for. (in the case of a laptop, occasional light use, it'll last for decades with the swap file off)

and, from personal experience, putting the intel MLC drives under heavy load in an actual data center, it's still ticking along just fine under very heavy strain, and for the *specific* application used there, increased performance from the 15k HDD array almost 500 times. *subnote* for everything else, they use seagate high grade stuff, so, the markets will co-exist for a very long time before either one can "win"*

[ThinkRob]

I would like to ask you to take a 1TB OCZ SSD with multiple controllers (Imagine the super fun overhead) againist three striped VeloRaptors but that wouldn't be fair would it? (Desktop application, but the price range is the same including RAID controller for the Raptors, not just a mobo solution)

No, it wouldn't actually. A more fair comparison would be something like three X25-Ms (or Vertexs if you prefer) in RAID-5 against three VelociRaptors in RAID-5.

Just for fun, take a look at a 1 v. 1 comparison: http://www.anandtech.com/cpuchipsets/sh ... i=3607&p=4

Somehow, I don't think relative performance is going to change that much when you put 'em in a RAID.

I also would like to bring up that no matter how the manufacturer's marketing teams (You know, the ones who don't let you see the internal specsheets?) wouldn't want you to know that realistically some of their drives *do* get fragmented in a specific way that *does* slow SSDs down

Indeed. That's why good reviewers (such as Anandtech) also post benchmarks after such degradation. In fact, they've actually gone so far as to compare a degraded X25-M to a couple other drives including the VelociRaptor. The performance degradation is around 3%.

Did I also mention SSDs still have *technical* seek time because they don't run natively in Windows?

What? Please elaborate.

Whats this about all this wear leveling, controllers? Still have to deal with those.The time it takes for a harddrive to move its head a milimeter from the edge of the disk in a little? Sub 1ms.

Then where, pray tell, is the extra 16 milliseconds coming from?

You can still beat SSDs when you have the proper disk defragger and have it arrange files properly.

No, you can't. Every single benchmark of a modern, non-JMicron SSD shows that this simply isn't true. If you have benchmarks that claim otherwise, please post them.

Edit: This is all ignoring all the "dies after a few years practically", you do know the SSDs that are avaliable to consumers are binned with the lowest grade of flash memory from the manufacturers and some of them only rate *the first _____* (Depends on manufacturer, some only assure "x" cells, others do "across these chips, that one cell per...") will be able to survive 10k rewrite cycles (You know, not like over time every little NAND cell will be happy about that?) Any of the stuff you read is the stuff the SSD mfgs feed you as marketing. If you can get their internal specsheets, you get a different story ("Binning" LED manfuacturers understand this conecpt too.) When I said overall, I mean *overall* as a storage solution as well.

Parsing that was kind of hard, but I gather that you're making a couple of claims here:

1) The manufacturers use low-quality NAND in consumer SSDs.

2) They warranty the drives for a certain number of years, but have no intention of making drives which actually perform at the rated levels.

3) There exist internal specification sheets that make this fact obvious.

Given the above, I'd like to ask three questions:

1) How do you have access to these internal specification sheets?

2) Why does Intel extend the same MTBF promise and the same length warranty to their SSD customers as many hard drive manufacturers do (i.e. three years)? Surely if hard drives are inherently more durable, Western Digital would warranty, say, their Scorpio Blue drives for a longer period than the X25-M.

3) Are you aware of the differences between SLC and MLC flash, and the lifespan implications of the same?

Also try comparnig a WD Scorpio Black's power usage to a SSD then try to imagine some of the conditions that you can go under.

Ok. Let's do that.

WD Scorpio Black (WD3200BJKT)
Read/Write 2.50 Watts
Idle 0.85 Watts
Standby 0.25 Watts
Sleep 0.15 Watts
source

Intel X25-E
Active (32GB) 2.40 Watts
Active (64GB) 2.60 Watts
Idle 0.06W
source

Intel X25-M (50nm)
Active 0.150 Watts
Idle 0.06W
source

Intel X25-M (32nm)
Active 0.150 Watts
Idle 0.075W
source

So yes: the Scorpio Black consumes 100mW less power when writing than the most power-hungry SLC drive that Intel makes. Given the performance of the various drives... well... I'll let the readers draw their own conclusions about which drive might be spending more time in an idle state.

They also aren't in deep states of idle, you should see all the fun sorts of things that happen when you ask one to delete a file (Goes into query, SSD decides when it wants to wipe it off that sector.)

What sort of fun things? You are aware that hard drives don't automatically zero free space upon file removal, right?

[Navck]

"Then where, pray tell, is the extra 16 milliseconds coming from?"
Sure, full seek from spindle to edge + any vibration of the head and arm. I don't think you're going to ask the harddrive to do much when you've defragged + moved the data around the surface of the drive where its positioned so that the most frequently accessed data is on the very outer edge (Highest transfer rate) in a very close group (Hey check that out, head only has to move fraction of the distance now.)
Speaking about "seek" times:
http://www.wdc.com/en/products/Products.asp?DriveID=623
Seek Time 250 µs (Typical)
(Just like it isn't a free action for HDDs, either)

"What sort of fun things? You are aware that hard drives don't automatically zero free space upon file removal, right?"
Oh sure, they just ask NTFS to mark those spots as free, SSDs still have their own process that clears the cell for a secure delete (Just as HDDs support secure delete features. Hey did you know the fun things that happen on a SSD when you ask it to write over a sector over and over? They'll actually ignore you when they figure out you're trying to write a pattern to them. Little more life saved there from things like Spybot's secure shredder), instead SSDs will place that into a delete query when it determines when it wants to clear that cell. The time from that could be measured between minutes to weeks, ask a data forsenics guy near you.

By the way, on warranties, I have a few Maxtors that have loooooooooooong expired their warranty by 8 times (One year warranty) and they still were running quite strong. Mind you, early 2000s, no ramps, none of that fun stuff you'd find on harddrives from the past few years.
Mind you most people have very bad luck with Maxtors (Must of gotten lucky or something with no bearing failures. Back when Maxtor still used solid bearings on drives.)

On three part question:
1. Data recovery/Forensics people like to check the next hot thing in storage, try making a friend with someone who works with one of the SSD mfgs.
2. Compare Intel againist OCZ/Toshiba. Harddrives are more durable in rewrite ops + Heat conditions.
3. Thats ideal and theorectical. They only rate those as I said before "These many cells will withstand these many cycles" or whatever the mfg. decides to "specifically" rate them as. Ask a SSD mfg. rep.

On power consumption, check out THCs thing on 7200RPM HDDs and see how they actually use power, then find a SSD review with that. Also, lets include some of the "ever popular" non Intel choices too (See it would be like me telling you how awesome WD Raptors are againist Seagate 7200.11s.) I also don't think my harddrive is constantly idle (Nor the SSD) because the access light blips on every few seconds to write something off the RAM to Firefox's cache occasionally, just like foobar2000 will read a small chunk every ofetn as well. Did I mention Intel and Samsung have quite the nice SSDs compared to some of the other offerings? You should check out some of the other "popular" choices, like OCZ. Also, SSDs are still *not* natively supported in Win7, why don't you ask Microsoft for NTFS not kill a SSD without full wear leveling implementation? CF cards are formatted in fat32 as with flash drives for a good reason. (There is another FS I remember that is relatively gentle FS that by chance works great with SSDs, *nix/BSD supports it, I'll remember it eventually.)


Also about VeloRaptors and RAID(-0, 4 drives):
http://canislupy.com/images/HDTune_4VR300R0_16.png
... I think thats a lot of change, what do you think?

[ThinkRob]

"Then where, pray tell, is the extra 16 milliseconds coming from?"
Sure, full seek from spindle to edge + any vibration of the head and arm. I don't think you're going to ask the harddrive to do much when you've defragged + moved the data around the surface of the drive where its positioned so that the most frequently accessed data is on the very outer edge (Highest transfer rate) in a very close group (Hey check that out, head only has to move fraction of the distance now.)

Yes, under ideal circumstances you might get lower than the average access time. With an SSD, on the other hand, there is no significant difference between random and ideal access times.

Speaking about "seek" times:
http://www.wdc.com/en/products/Products.asp?DriveID=623
Seek Time 250 µs (Typical)
(Just like it isn't a free action for HDDs, either)

And? Yes, that's one of the main benefits of SSDs -- exceptionally low average seek times. Show me a hard drive with a <1ms *average* seek time and I'll eat my hat.*

Oh sure, they just ask NTFS to mark those spots as free, SSDs still have their own process that clears the cell for a secure delete (Just as HDDs support secure delete features. Hey did you know the fun things that happen on a SSD when you ask it to write over a sector over and over? They'll actually ignore you when they figure out you're trying to write a pattern to them. Little more life saved there from things like Spybot's secure shredder), instead SSDs will place that into a delete query when it determines when it wants to clear that cell. The time from that could be measured between minutes to weeks, ask a data forsenics guy near you.

You'll see a similar behavior with hard drives. In both cases, the "soft delete" behavior is a function of the filesystem. Overwriting each with random data will in fact change the underlying storage medium, as will issuing the ATA secure erase command. Frankly, I'm not really sure what your point is here.

By the way, on warranties, I have a few Maxtors that have loooooooooooong expired their warranty by 8 times (One year warranty) and they still were running quite strong. Mind you, early 2000s, no ramps, none of that fun stuff you'd find on harddrives from the past few years.
Mind you most people have very bad luck with Maxtors (Must of gotten lucky or something with no bearing failures. Back when Maxtor still used solid bearings on drives.)

And up until recently I had a 5.25" drive that worked great, but you won't see me arguing for the superiority of that tech. Anecdotal evidence is pretty useless.

Data recovery/Forensics people like to check the next hot thing in storage, try making a friend with someone who works with one of the SSD mfgs.

Amazing, really, that these secret spec. sheets have never been mentioned outside of this forum. Sorry, but I'm still a bit skeptical.

Compare Intel againist OCZ/Toshiba. Harddrives are more durable in rewrite ops + Heat conditions.

Again, we're back to the spec sheets. The publicly-available ones say otherwise. If you can refute this by posting internal spec. sheets or large-scale trials, etc. then please do. Otherwise, you're basically just stating this with no supporting evidence.

Thats ideal and theorectical. They only rate those as I said before "These many cells will withstand these many cycles" or whatever the mfg. decides to "specifically" rate them as. Ask a SSD mfg. rep.

See above. All the publicly available data states that SLC drives have a very, very long lifespan. The whole "I have some secret data that proves my point, but it's not available to anyone else" bit strikes me as a bit... conspiracy theorist-like.

Also, lets include some of the "ever popular" non Intel choices too (See it would be like me telling you how awesome WD Raptors are againist Seagate 7200.11s.)

I'm not going to defend the JMicron-based drives because... well... they're junk.

That said, Samsung and OCZ both make drives that provide low power consumption (lower than the drive you mentioned, but not lower than the X25s).

Also, SSDs are still *not* natively supported in Win7, why don't you ask Microsoft for NTFS not kill a SSD without full wear leveling implementation? CF cards are formatted in fat32 as with flash drives for a good reason. (There is another FS I remember that is relatively gentle FS that by chance works great with SSDs, *nix/BSD supports it, I'll remember it eventually.)

Yes -- drives designed without wear leveling will have issues when used for typical desktop usage. Fortunately, any modern SSD has wear leveling for precisely this reason.

And could you be thinking of JFFS2 or NILFS?

Also about VeloRaptors and RAID(-0, 4 drives):
http://canislupy.com/images/HDTune_4VR300R0_16.png
... I think thats a lot of change, what do you think?

I didn't mean change relative to single-drive performance. I meant that I don't think that a RAID-0 array of VelociRaptors will best a RAID-0 array of good SSDs.

Now HD Tune is pretty poor for this sort of thing (since its measurement of random I/O workloads is pretty poor), but you did provide an interesting set of numbers.
I noticed you forgot to post the other, equally-relevant image from that site: two X25-Ms in RAID-0: http://canislupy.com/images/HDTune_Arec ... SSD-R0.png Oh, and average access times are a less than 1/60th of what they are for the array you mentioned.

You're right. It *is* a big change.

*Note: I do not actually have a hat. Should such a drive appear, I will purchase a hat prior to consumption.

[Navck]

You're ignoring the point that a harddrive can get sub 1ms seek times when data is close enough and within the timeframe of the head to go from A to B. It is *not* impossible for them to access things on close tracks, it is a requirement for harddrives. Properly defragged harddrive with frequently accessed data placed on the outer tracks can beat SSDs easily.

Harddrives are completely capable of rewriting to a sector billions of times in a day, SSDs die from that. The delete query and capability to ignore commands is used to prevent this from happening.

I'll try to get some internal data but this isn't something I can make magically happen in an instant. (Key word: Internal specsheets, the ones NOT for public consumption and NOT spun by the marketing department with help from engineers to give you theorectical ideals. These are the ones they have for reference when designing products.)

Wear leveling is to only mitigate death so it takes 5 years before you have 10GB usable left, badly implemented wear leveling (Much more common than you think) means you'll be hit by capacity loss within two years (By the way, good luck with data retention. Imagine what happens when your data disappears in a year or two from some drives) and so very little space left by the third.

Look, SSDs won't be mature enough in another 3-10 years, come back to them when each cell can withstand trillions of rewrite cycles and overall they have the research and funding behind something ancient as harddrives behind them. You're paying for their R&D right now on half baked products. Give it a while and don't pay 300 dollars for 160GB.

Here, let me share a secret with you that I learned from a harddrive engineer:
Harddrives cost less than ten dollars USD to build and even if they charge you two hundred a pop. Most of that goes into R&D when you're trying to make things fly nanometers off the platter at speeds scaled so that the head would be a 747 going Mach 10 and produce them on a massive scale. Just like harddrives can adjust fly heights of their heads with a small element that heats up to change shape (Airfoil).
Those are the things they don't release to the general public, go and call your favorite SSD maker posing as a data recovery person and try to get friendly with some of the staff, maybe they'll let you know some of their cells are only capable of withstanding 1k while another mfg. proudly states theirs will do 10k.

By the way, I believe the FS was JFS.

To anyone else who wants to hear about the T410:

I have the Ultimate 6300 wireless option for my T410. In comparison to the T43, there is this one spot downstairs where the signal is less than 10% and sometimes the AP is unseen. To the T410? Thats a 100% signal right there. Similarly, there is a room on campus where I am unable to get any signal with my HTC Fuze until I walk into a hallway opposite of the room and navigate down the building. The T410 picked up the student AP with 70-80% signal and connected perfectly. The T43 has the 2915ABG option from Intel, I definitely can advise everyone not to skimp out on the wireless if you want the capability to inhale wireless signals from range. (Note: Have not used Wireless N capabilites yet but it seems there a few APs in my neighborhood.) I know I am picking up signals from hilly terrain a few... Streets up and down right now. If they were unsecured I'm more than sure that I could get a stable connection right now.

[Volker]

Smaller formfactors favors SSD, its as simple. For 1.8" drives (T410s hopefully some day), you'd be crazy not to go with SSD. Now that even Windows has finally support for TRIM the performance-degradation problems are a thing of the past.

Once you get to a multi-disk RAID, on the other end of the scale, HDDs are the way to go. Cheaper, better known failure modes, SSD TRIM does not work with current RAID chip, ...

[ThinkRob]

You're ignoring the point that a harddrive can get sub 1ms seek times when data is close enough and within the timeframe of the head to go from A to B. It is *not* impossible for them to access things on close tracks, it is a requirement for harddrives. Properly defragged harddrive with frequently accessed data placed on the outer tracks can beat SSDs easily.

Yes, if the data you're interested in accessing is exactly under the head at the moment that you're trying to access it, a hard drive can beat an SSD... maybe. But realistically, every single spec. sheet and every single benchmark shows that seek times aren't even close. I really don't know how else to put it: reality disagrees with you.

Harddrives are completely capable of rewriting to a sector billions of times in a day, SSDs die from that. The delete query and capability to ignore commands is used to prevent this from happening.

And? So they don't zero cells immediately, and as a result they last longer. I'm sorry, but I'm just not seeing what's wrong with this. And SLC SSDs are capable of having > 1TB of data/day written to them every day, for years on end. At that point, lifespan concerns are really not that much of an issue.

By the way, I believe the FS was JFS.

I'm pretty sure it's not. JFS was not designed with any consideration for SSDs and has no support for wear-leveling of any sort. (I run it on as my main FS on all my machines, but for other reasons.)

Now that even Windows has finally support for TRIM the performance-degradation problems are a thing of the past.

Plus, if benchmarks by Anandtech and others are to be believed, Intel's drives only suffered a 3% degradation in the worst-case scenario, which isn't really something you'd notice. But yes, widespread TRIM support makes SSDs an even better choice.

Back on topic: can anyone with access to a T410 and an X200s compare their wireless reception? I'm curious if the move to the carbon-fiber lid has improved wireless reception like it did for the X200s.

[Navck]

A harddrive defragmented with data arranged onto the outer track with a proper utility that will arrange the data within specific clusters *will* allow for harddrives to have SSD-like seek times for those files. Just me but I have delt with people who are fanatical about their 500+ dollar purchase and still have an inferior loading time for various games compared to my desktop (74GB Raptor, 320GB RE3, 2x 640GB Caviar Blacks)

I did not state that JFS was designed with SSDs in mind, there was something about JFS being "gentle" on SSDs unlike NTFS.

Show me a SSD that survives without wear leveling, then you can brag about lifetimes. Datacenters do not use SSDs for a reason, even with all the ways they try to get away with it. There is a finite lifespan and they will hit them fast in those conditions.

On the wireless, the difference between lid construction is going to make less importance than antenna design. As far as I know, the T410's lid has properties similar to the lid of the T43 on the surface (The metallic flakes, possibly related to the magnesium construction on the T43's lid?) The biggest advantage I believe comes from the multiple antennas on the T410 + 6300's capability to use all three.

[blackomegax]

a proper utility that will arrange the data within specific clusters .

which specific piece of software are you talking about?

[khtse]

Show me a SSD that survives without wear leveling, then you can brag about lifetimes. Datacenters do not use SSDs for a reason, even with all the ways they try to get away with it. There is a finite lifespan and they will hit them fast in those conditions.

Does this even make sense? Why would you want a SSD with no wear leveling? SSD has their disadvantage, limited write cycle, and wear leveling algorithm is designed to mitigate this and make SSD work. "Show me a HDD that doesn't spin, then you can brag about the lack of mechanical failure." Does this make sense? It makes as much sense as your statement.

Datacenters do not use SSD for another obvious reason too, they are too expensive for this purpose at this moment. Datacenter by definition require huge storage, and this is not the market SSDs are aiming for right now, obviously.

But you know what? Datacenters don't buy a bunch of hard drives and use them for 20 years. (1) Hard drives will and do fail, (2) many hard drives are replaced before they fail for precaution reasons, and (3) technology improves, and old hard drivers are replaced with bigger, faster, and more energy efficient ones from time to time.



Have fun defragging your bunch of hard drives few times a day (with a proper utility of course) to have some games load faster than your friends' SSDs. Let me know when you figure out how to fit the raptor into your thinkpad, and how to silent the raptor roar.

[Navck]

which specific piece of software are you talking about?

Various defraggers with the capability to rearrange files on a harddrive in a specific pattern by last access time/folders. A single go after any major install of things will get you set in minutes really. Otherwise I have my systems set to defrag on idle time when they are let alone for a few minutes.

By the way, those people aren't friends, they're just acquiantances that feel the need to justify their 500 dollar purchase by yelling about how I have an "inferior mechanical device" to their "glorious piece of money". I still outperform their systems overall (Strange how I have a faster system without dumping 4 grand more and having higher specs...) The only area I guess I don't outperform them is in benchmarks but I think I make up for that by practically outperforming them in real life situations.

Datacenters, servers and other serious markets will not use SSDs until they are able to sustain unlimited rerwrites or something close to "forever".

By the way, data retention is still your enemy on SSDs.

[blackomegax]

Various defraggers with the capability to rearrange files on a harddrive in a specific pattern by last access time/folders. A single go after any major install of things will get you set in minutes really. Otherwise I have my systems set to defrag on idle time when they are let alone for a few minutes.

By the way, those people aren't friends, they're just acquiantances that feel the need to justify their 500 dollar purchase by yelling about how I have an "inferior mechanical device" to their "glorious piece of money". I still outperform their systems overall (Strange how I have a faster system without dumping 4 grand more and having higher specs...) The only area I guess I don't outperform them is in benchmarks but I think I make up for that by practically outperforming them in real life situations.

Datacenters, servers and other serious markets will not use SSDs until they are able to sustain unlimited rerwrites or something close to "forever".

By the way, data retention is still your enemy on SSDs.

"various defraggers".
right....

PS, i also move to get all this crap out of this thread.

[Navck]

Fine if you don't believe me, I'll list a few off for the few who haven't had a chance to use an actual defragger on their poor HDDs that are fragmented to hell, giving them an impresion that their SSDs are totally justified purchases:
O&O Defrag, Diskeeper, PerfectDisk, UltimateDefrag, SmartDefrag

All of those will do a defrag where frequently accessed/modified files are moved to the start of the disk, some of those will do a better job than others and I know one of them lets you move select files (And only those) to the start of the disk.

[ThinkRob]

Datacenters, servers and other serious markets will not use SSDs until they are able to sustain unlimited rerwrites or something close to "forever".

Not so.

A harddrive defragmented with data arranged onto the outer track with a proper utility that will arrange the data within specific clusters *will* allow for harddrives to have SSD-like seek times for those files.

In the absolute best-case scenario, and for those specific files, yes. Compare and contrast to an SSD which will give you <= 1ms seek times for all files.

Show me a SSD that survives without wear leveling, then you can brag about lifetimes

That's silly. That's like if I were to say "show me a HDD that can be used underwater and then you can brag about lifetimes" -- it's a completely absurd premise, and while the drives would indeed fail in those conditions it proves absolutely nothing about the suitability of the current solutions.

Mods: can we please split this into a separate thread?

[Harryc]

Thread split off from "T410 First Impressions".
http://forum.thinkpads.com/viewtopic.php?f=45&t=84606

[Zender]

Just a little side note.

Seek time on typical mechanical hard drive consists of two things, neither of which is negligible:
1) the need to move head "left-right" to the proper track
2) the need to wait for the track to rotate to the required position

Even if you manage to have your data in nearby tracks and get <1ms for the first point (which I doubt), the second point will cost you an average 4ms on 7200rpm drive (and average 5.5ms on 5400rpm drive). Even if account for SATA NCQ, which will allow the drive to optimize reading so that sectors which were requested later are read earlier thanks to being rotated properly, I doubt you'll get this anywhere near 1ms on average.

As for moving frequently used files to outside tracks - yes, you can do that. But that won't help you with writes, which are done to places decided by OS/filesystem, and it won't help with this newly written data. And if you use only the same data all the time, you actually don't have to bother with SSD degradation/write cycles.

As has been said, the real world does not agree.

[radiodavid]

In this thread, which started in a Laptop section, I'm surprised that a significant advantage of the SSD has not been mentioned: no mechanical/moving parts. For anyone who has lost a platter-and-head drive due to shock, vibration, and the other routine artifacts of laptop use, paying hundreds of dollars more for security is not as significant a cost as it might seem. There is a tiny weight advantage, too, but spinning platter issue is a major force behind using SSD's. The access time, fast boots, improved writes, lesser heat are all bonuses.

I don't put server farms in my briefcase. I carry a laptop. I travel over 100 k miles a year, and the SSD makes enormouse sense.

[tylerwylie]

SSD's are used in data centers, look at the FusionIO drives and PCI-E SSD's that exist, they are made for data centers. I have one in one of my servers at work. The OC-Z Vertex Turbo's are very well suited for laptops, big storage should be on a server, I have a BSD box with a few big HD's as my storage server, so that eliminates the need for giant storage devices in my computers and laptops. SSD's will only get better from here on out. I'm about to get another FusionIO SLC drive for work too, those things are amaaaaaaaaaaaaaaaaazing.