SSD Price Drops Signaling End of Spinning Media? 646
gjt writes "When Intel and OCZ recently announced new 'affordable' Solid State Disk drives — offering a meager 32-40GB — we initially yawned. But, then we took a closer look at the press releases and the in-progress research and development in SSD technology and opened our eyes. While the new drives aren't affordable on a cost per gigabyte basis for everyone, it does set a precedent — and most importantly a barometer price of $100. And it really does start the death clock for hard drive technology."
Re:Nikon F6 and FM10 (Score:2, Informative)
Re:Child pornographers. (Score:5, Informative)
Inside of all harddrives for the last 10 or so years are multiple, very powerful neodymium iron boron magnets that move the actuator arm over the surface of the discs. If magnets outside of your drive would erase data, then surely these intensely powerful magnets inside would do the same, no?
The most conclusive testing I've seen done on this was several years ago. A guy had stacks of dead hard drives, and he decided to harvest the magnets from them. He had a stack of 50+ very powerful NIB magnets. He then took a working HDD, full to capacity, and covered the entire hard drive in them- front and back, with layer upon layer of magnets. Then he set the drive in a desk drawer for a few weeks, after which he plugged the drive up, and all of his data was still completely intact. Not 1 file was corrupted in any way.
Now, if you put a
Re:In 5 years (Score:5, Informative)
Yeah, if you really want to compare apples to apples, measure MTBF.
Oh, and let's not forget the SSD's far superior ability to decay gracefully.
Re:Interesting assumptions (Score:5, Informative)
Typical users keep music on their computer.
Re:What about random writes? (Score:3, Informative)
If by "cheap" you mean the craptastic OCZ Core series, and the other SSDs of the same gen that used the god-awful JMicron controller, the "cheap" end of the SSD market is full of high-performance drives that don't choke on random I/O. As soon as Intel and Indilinx came out with controllers that were worth the sand they were printed on everyone started doing it, and we've now got a market where the performance delta between "cheap" and "prosumer" SSDs is much, much smaller.
The vast bulk of the cost of the SSD is the flash, which is why if you can find someone still selling Core series SSDs they're only marginally cheaper than an SSD that's actually usable.
Re:Reports of HDDs' demise greatly exaggerated (Score:4, Informative)
How much for a hard drive that's as fast as that $125 SSD?
The 1TB Seagate hard drive that I recently tested gets random 4k read rates in the ~1MB/second range. My 80GB Intel X25-M gets ~38MB/second.
That's about 40 times more performance for THE SAME PRICE!
Storage capacity is irrelevant in many situations.
A 40GB SSD is more than sufficient for your average manager/executive. They'd almost certainly prefer opening Outlook and Power Point in a tenth of the time it used to take to having an extra thousand gigabytes of unused space on their laptop.
The 80 GB drive I have in my system was the best upgrade I ever bought. Kernel compiles are crazy fast, and all of the media I need can be streamed off the network (sharing a single one of those 1.5TB drives with a dozen or so other people).
Not so fast. (Score:5, Informative)
Don't be fooled, people: http://en.wikipedia.org/wiki/Solid-state_drive#Disadvantages [wikipedia.org]
Re:Reports of HDDs' demise greatly exaggerated (Score:5, Informative)
Wrong.
A full Win 7 Ultimate install with Office 2010 + Visual Studio 2010 + Project & Visio 2010 sits at around 25GB.
you still have 15GB left. Take off VS2010 and you are sitting around 20-25GB free.
Re:In 5 years (Score:4, Informative)
Mod parent down for being simply wrong. Power consumption is directly proportional to clock frequency, not the square of the clock frequency.
An input to a CMOS gate can be approximated as a capacitor, so each time the capacitor is discharged, an energy is consumed equal to the energy stored in the capacitor. The energy in the capacitor is 1/2*C*V^2 where C is the effective capacitance, and V is the supply voltage. The total power consumption is n*k*f*C*V^2 where f is the clock frequency, n is the number of gates and k is the activity level which describes the number of times per clock cycle each gate will change at its input (on average). The 1/2 gets absorbed into the k.
If you double n (two cores), but halve f, the power consumption doesn't change.
Re:In 5 years (Score:5, Informative)
The point of the shift to 4k sectors (e.g. the WD "Advanced Format" drives) is that the amount of space needed for error correction at ever increasing densities was entering into the bounds of diminishing returns. Larger blocks mean less error correction is needed and thus more storage space for a given platter density. Anand has a pretty good writeup on it here: http://www.anandtech.com/printarticle.aspx?i=3691 [anandtech.com]
Re:Ram Disks (Score:3, Informative)
Um, no.
RAM disks -- being RAM -- aren't permanent storage, and were never held up as a replacement for any kind of permanent storage. They were always a work around for permanent storage being too slow for applications that had a demand for responsiveness and could accept the risk of data loss to acheive it, and even then were largely useful for applications that had been designed for older (relatively) constrained RAM situations, and therefore did stuff "on disk" not because it needed stored at that point, but to avoid keeping too much in RAM at once.
Re:...Or an arms race (Score:4, Informative)
(most SSD are 2.5", not 3.5")
PCIe "hard drives" already exist.
Here's a 1TB model: http://www.newegg.com/Product/Product.aspx?Item=N82E16820227500 [newegg.com]
There are others in 250GB, 256GB, and 512GB capacities.
I doubt that the cost goes down much though. The PCIe interface chip isn't free, and neither is the card bracket. The PC board itself is also much larger, and has to be thicker than those used on most hard drives. The cost differences are probably a wash.
Re:...Or an arms race (Score:4, Informative)
There is a fixed minimum cost for building a hard drive. Spindle, motor, etc. It's about $70.
Not quite. A shop near my house (Rome, Italy) has 320GB drives on sale at 35.4 EUR (roughly $48) including 20% sales tax - and this is just the first I bothered to check, it's a street price, it includes their own profit, and it's a 3.5" unit. When 3.5" go obsolete once and for all, the 2.5" drives will stop costing a premium and actually become cheaper, most likely.
So - while there definitely is a price level where mechanical units stop making sense, it's nowhere near $70 and probably it will keep shrinking over time.
Anyway - the entire point is moot. A sum of other factors (weight, power consumption, heat generation and tolerance, shock tolerance) will most likely push hard disks away in the lower capacity ranges.
Tape vs. Spindle (Score:3, Informative)
The price per GB is one concern, reliability and data transfer rate are two others. There are more - thermal considerations/power consumption, portability, media life (bit rot).
Most people have storage tiers - you can have fast/slow JBODs ready and waiting to accept and retrieve data, incorporating slower, offline tape which is SAN-connected, and managed by a robot, which can be transported via station wagon for great justice.
Price-wise, LTO-4 cartridges hold 800GB at a cost of around US$35, which also requires a minimum of one US$300 device to read/write the tape, and likely a dedicated connection on a dedicated interface (some flavor of SCSI), which may tack on another $100-$400.
I can go to Foo-Mart and buy 1TB of SATA for $100 or less - perhaps with it's own (slow by comparison) interface, enclosure and power supply.
Are they interchangeable as a solution? Nope.
Re:Careful on Your Terminology There (Score:3, Informative)
Streaming video is great for a computer screen or an SD TV. But streaming enough pixels to fill a 1080p screen requires a lot of bandwidth. Blu-ray spec is 54Mbps at the low end.
While some people may be lucky enough to have that in download speed, most don't and upload speeds like that are a long way off.
Re:Price isn't everything (Score:4, Informative)
Google, being a very big consumer of HD's, has published such data.
3% die in the first 3 months, another 4% die by the end of the 1st year.
8% more die in year 2.
Maybe you have missed the fact that these monster capacity drives actually suck in the reliability department? yeah.. 3 out of 100 convert themselves into worthless crap within 90 days.
Re:...Or an arms race (Score:3, Informative)
I believe mainframes already do the whole memory page == disc block thing, but I'm not an expert.
I asked the same question on /. a while back regarding more common operating systems and got this response [slashdot.org] from m.dillon, which seems to indicate it's not really feasible unless the whole software stack is (very) radically altered, or the performance delta between memory and storage becomes alot smaller.
Re:In 5 years (Score:5, Informative)
Pity the lesson of Y2K went unheeded - where every COBOL programmer was paid whatever they asked to fix their code, but after should have all been taken out to a field and shot in the head.
You don't remember the days of limited storage, do you? Those 2 extra bytes times 100000 records * 20 date fields was 1/10 of your drive back then.
Now get off my lawn!
Re:...Or an arms race (Score:3, Informative)
Because otherwise it might run too fast? The SATA interface is the big bottleneck holding SSD speeds down. If you put the whole thing directly on the PCIe bus it would be a lot happier. Fitting a fast SSD into a PCIe slot and then tying it to SATA is just cruel.
Re:In 5 years (Score:5, Informative)
Oh, the ignorance of youth. Computers were over fifty years old in Y2K, and your cell phone is more powerful a computer than any built in the fifties. Hell, a Hallmark card is more powerful and sophisticated. They used two digits for years because they had to. There simply wasn't enough data storage (which oddly makes this otherwise offtopic comment on topic). You take your terrabyte disk drives and your gigabyte SSDs for granted, but early system were measured in kilobytes.
An example is the IBM 1401 [wikipedia.org] that was announced by IBM on October 5, 1959 (I was seven years old at the time).
Legacy data and cheapassed managers kept the two digit dates around, and programmers and systems analysts warned management of the coming doom, but were ignored until it was almost too late.
A COBOL programmer in the 1950s would be dumbstruck by what we have today. Actually, I'm dumstruck as well; cell phones, flat screen computers, and self-opening doors in Star Trek were impossible; science fiction. You young folks can't imagine how primitive things were when I was a kid, and nobody dreamed we'd see SSDs.
Re:In 5 years (Score:3, Informative)
Well, I'll question the more reliable part. Despite having owned way more harddiscs in the last decade that I've owned even tape media (tape was a backup solution only for some years), I had more unreadable tapes than unreadable hdds.
Re:Price isn't everything (Score:4, Informative)
Re:I think so. (Score:4, Informative)
More than two years ago the balance shifted. It is now cheaper to build massive storage servers with SATA RAID in-house and off-site and backup to both than to put a Tape Library in your office and rotate tapes off-site.
This is true even when you assume $0 for transporting tapes and free off-site storage.
Re:In 5 years (Score:3, Informative)
You can't use resistors as your model of computer components- well you can, but you get unreasonable conclusions like the one above.
Take as an example CMOS tech, theres a pretty good run down of why any type of semiconductor doesn't act at all like a resistor when it comes to power dissipation.
http://en.wikipedia.org/wiki/CMOS#Power:_switching_and_leakage [wikipedia.org]
FTR- I'm not saying P=I^2 R is wrong... Its certainly correct, the challenge is coming up with 'R' when you're talking about switching semiconductors (or anything else really).
Re:Not so fast. (Score:5, Informative)
I'll edit the article later. For now...
- Wear leveling used on flash-based SSDs...
Oops! Wear leveling is done on HDDs too. And it isn't a disadvantage: it is a solution.
- More expensive, lower capacity
Don't need to address that, since that is the topic of the article...
- Asymmetric read vs. write performance
Oops! Platter drives have this problem too!
- Requires TRIM
Solved.
- Limited lifetimes
Funny, that's considered the major downside to platter drives. Anyway, this is the same as the first point.
- Performance of SSDs degrades with use.
Solved. See TRIM. Note that this is also a problem on platter drives.
- SATA-based SSDs generally exhibit much slower write speeds.
This one doesn't even make sense.
- DRAM-based SSDs
Aren't what we are talking about, so that's irrelevant.
Re:HDDs are Done When Google Says They Are Done (Score:3, Informative)
This article claims Google is using Intel SSDs [informationweek.com]. There's no source though, and Google declines to comment. Oh well.
Re:In 5 years (Score:3, Informative)
More significantly, long-term storage would run you $0.10 per byte per month. Those extra two bytes times 100,000 records times 20 date fields would run the bank a half-million dollars a year in increased data archiving costs.
Re:In 5 years (Score:3, Informative)
I'm doing this already, it works bloody well. I have my OS and Programs installed on a 120GB SSD, which sits around 50% utilisation and use NTFS junctions (aka symlinks) to map storage for stuff that doesn't need superfast seek speed (aka data) onto a group of 1.5TB drives. It takes a little management, so isn't quite ready for the average user yet - but you do effectively get something like 5TB of online disk space combined with SSD performance.
Interestingly, i've found on Windows 7 that by running OS/Programs from the SSD that the contention issues you would normally get on a spinning disk are mitigated a great deal - and there's no noticeable hit with having the entire user profile (including junk like web cache). The system is booted in 5 seconds after finishing its POST, and the desktop is snappy right from the get go.
As to the hard disks as backup, it works pretty good. At work we have maybe 1.2TB in a full backup - we do a weekly full backup and incrementals onto LTO tapes, and a second weekly backup onto a consumer grade SATA 1.5TB drive in a USB cradle. The SATA drives are taken offsite in case the office burns down, £80 + carrying a few hundred grams around makes for really cheap and fast data transfer.
Re:...Or an arms race (Score:3, Informative)
Sorry to burst your joke, but in 2000, CPUs had only just hit 1 GHz.
Re:...Or an arms race (Score:2, Informative)
Sorry to burst your joke, but in 2000, CPUs had only just hit 1 GHz.
It's not so inaccurate... according to Wikipedia [wikipedia.org], November 2000 saw the release of 1.5 GHz CPUs, and 2 GHz P4s were out in August 2001. Even 3 GHz was out in Aug
Re:...Or an arms race (Score:4, Informative)
Re:In 5 years (Score:3, Informative)
I've had a few linux and bsd boxen run off of CF cards, but even the fast ones are so bloody slow compared to even a slow harddrive. I suppose the lack of a buffer might be part of the problem - or do CF cards have small buffers of SRAM?
iowait drives me crazy.
Re:...Or an arms race (Score:3, Informative)
Oh, I meant the vestigial PCI (not PCI-e) slot on most full size ATX boards.
Re:Not so fast. (Score:4, Informative)
- Wear leveling used on flash-based SSDs...
Oops! Wear leveling is done on HDDs too. And it isn't a disadvantage: it is a solution.
Wear-leveling is done for bad sectors on a HDD, not as standard practice.
- More expensive, lower capacity
Don't need to address that, since that is the topic of the article...
- Asymmetric read vs. write performance
Oops! Platter drives have this problem too!
Wrong. I just ran a benchmark and saw widely varying performance based on sector size and sequential vs random, but the reads and writes were the same speeds. I tested all 3 of my HDDs. If this is true, [citation needed]
- Requires TRIM
Solved.
- Limited lifetimes
Funny, that's considered the major downside to platter drives. Anyway, this is the same as the first point.
- Performance of SSDs degrades with use.
Solved. See TRIM. Note that this is also a problem on platter drives.
These two are related. SSDs may last longer than HDDs for certain use environments (space?) and types (maybe even a typical user) but they're definitely not a given. Try to defrag a MLC drive a few times and it'll be dead in a week (yes I know you don't need to defrag a SSD, but there are processes that can mimic it). And every HDD I've owned (dozens) for the past 15 years is still spinning, though I suppose I may be lucky. And a SSD will have cells go bad as a consequence of the technology, while a bad sector on a HDD is a fault. If a HDD goes bad, it's due to mechanical failure of the supporting systems, not degradation of the media itself.
--
Look, SSDs are great. I love coding on one because compiles are wicked fast. But they are not ready to replace HDDs. And they're not the panacea you're making them out to be. Like literally everything else, you pick the right tool for the job. I'll keep my backups and my movies and music and 11 days of recorded TV on my 2TBs of spinning media, and my OS, applications, and code on a smaller 128GB SSD.
Re:...Or an arms race (Score:1, Informative)
This is weird... after reading your post, I just took a look at my local computer retailer's website, it just happend to be added a few days ago (sorry, German, but obvious)
http://www.kmelektronik.de/shop/index.php?show=product_info&ArtNr=22670