Follow Slashdot stories on Twitter

 



Forgot your password?
typodupeerror
×
Data Storage

Intel Unveils SSDs With 6Gbit/Sec Throughput 197

CWmike writes "Intel announced a new line of solid-state drives (SSDs) on Monday that are based on the serial ATA (SATA) 3.0 specification, which doubles I/O throughput compared to previous generation SSDs. Using the SATA 3.0 specs, Intel's new 510 Series gets 6Gbit/sec. performance and thus can take full advantage of the company's transition to higher speed 'Thunderbolt' SATA bus interfaces on the recently introduced second generation Intel Core processor platforms. Supporting data transfers of up to 500MB/sec, the Intel SSD 510 doubles the sequential read speeds and more than triples the sequential write speeds of Intel's SATA 2.0 SSDs. The drives offer sequential write speeds of up to 315MB/sec."
This discussion has been archived. No new comments can be posted.

Intel Unveils SSDs With 6Gbit/Sec Throughput

Comments Filter:
  • on a (rich) consumer SSD. But, while I'm loving all the Marvell / Sandforce / Intel hypersonic speed-worthiness, how about a decently fast, really affordable solid state drive? How much longer will these be 20x the per GB cost of a HDD?

    • We saw prices of the previous generation drop from $4/GB to under $2/GB while waiting for SATA 3.0 adoption. Now those same SATA 2.0 SSD's will drop much further... by years end it will be a great time to pick up older SATA 2.0 SSD's
      • by haruchai ( 17472 )

        Having been a somewhat early adopter of SSDs, I got bitten a couple of time by the JMicron bug - a pox on them. I've decided I won't buy another SSD until I can get a 128 GB drive for $100

        • Comment removed based on user account deletion
          • Having been a somewhat early adopter of SSDs, I got bitten a couple of time by the JMicron bug

            Everything I've read so far suggests that if you are buying SSDs you want to go with Intel.

            Note that this new Intel SSD is the first Intel-branded SSD that uses a non-Intel controller. It uses the same Marvell controller [techreport.com] used in the well-regarded Crucial RealSSD C300.

            I've also read about Intel's great combo of performance and robustness, but that reputation is mostly a result of Intel's controllers. JMicron, a manufacturer of SSD controllers, got its buggy reputation from early JMicron-based SSDs. Marvell's controller performance has been proven in many reviews, but Intel's "endorsement" gives

            • by Lennie ( 16154 )

              "mostly a result of Intel's controllers"

              Actually, the controller is just the hardware. It is the software running on it which makes it smart. Intel owns the software I would think.

            • by jbolden ( 176878 )

              Thanks for that tip. I've been buying computer and printer memory from crucial but it was great to red the review on A&D and see how well their drives scored.

    • by TheLink ( 130905 )
      But this new Intel SSD uses a Marvell controller - the same as one used by Crucial's C300. Why should people pay a higher price for an SSD with the same controller? The uninformative article says nothing about that - no proper benchmarks or tests (random reads/writes, sequential/read writes, max latency of a random read/write transaction that interrupts those random/sequential read writes, read/writes of high/mid/low compressible data). Such an article should be pretty useless for real Slashdot nerd.

      As for
    • by Neil Boekend ( 1854906 ) on Tuesday March 01, 2011 @03:08AM (#35345672)
      IMHO SSD's should be used as "something in between your HDD and your memory". As long as it's big enough to contain the OS of your choice and all the programs it's big enough. Your MP3's and movies do not require the high throughput. In a desktop this means there should be 2 disks: an ssd for speed and a hdd for capacity. In a laptop there should be an SSD on the mini-PCI-e and an HDD on sata OR, if you must (for space reasons), an SSD on the mobo and an HDD on sata. This would optimize both capacity and speed, while keeping costs (relatively) down.
      The cost of an SSD is paid back by the speed, not the capacity. What I find strange is that shops list SSD's by EUR/GB instead of EUR * s/MB. The speed is the defining attribute, not the capacity.
      • by thue ( 121682 )

        > Your MP3's and movies do not require the high throughput.

        And more importantly, your MP3's and movies do not require the random reads and writes which is an SSD's greatest strengths.

        • Well, the *sequential* read/write speeds of this new drive also dominate spinning hard drives by a fantastic amount. Just a few years ago I was striping hard drives to hit 50 MB/s for a video capture device; ONE of these does over 300 MB/s, and they fit in a laptop! "SSD's: they're not just for IOPS any more."
      • by Lennie ( 16154 )

        That is what ZFS with L2ARC ('level 2'-cache) does, it uses the SSD as a cache for the slower but bigger disks.

        On Linux a fairly new development called bcache does something similair

      • For a server deployment, I'd agree, but on a laptop the fewer moving parts, the better. Now if only we could get rid of the fans...
      • >>IMHO SSD's should be used as "something in between your HDD and your memory"

        You're talking about what is essentially Vista/Win7's readyboost feature, using flash as a cache between RAM and HDD.

        You could do this now. Just buy an external SSD and plug it in.

        • No. I am talking about splitting the stuff we put on the harddisk now manually. Not as a cache, but the only version of the OS and the most important software.
          Having said that: Yes, I do think cache is a good use of SSD's. Just not the best use imaginable.
      • The speed of these devices on SATA3 (500 MB/sec) is almost that of PC133 DRAM (570MB/sec). While you could make that argument for a modern system, anything spec'd to run on a P3/1st gen Athlon "should" be able to use the HD as raw memory with similar performance to those machines.

        • Unlikely. If it's sitting on a plain jane PCi bus, the max throughput is only 133 MB/s (32 bits * 33 MHz). And the latency is likely 10 to 100 times higher.

      • IMHO SSD's should be used as "something in between your HDD and your memory"

        Unfortunately, only ZFS and Windows Vista or later with Readyboost and Superfetch will actually DO this without application support. There is an experimental dm-cache module for Linux which does this (it's a block-level cache) but it does not work on contemporary kernels and appears to have been abandoned.

        Your idea has a little merit, though, because hybrid drives already exist. Of course, they have even poorer driver support.

    • HDD tech is established... it's like complaining that 1.5 TB disks cost $100 when a backup tape the same size costs $50. You're paying for speed and essentially today's device helps pay the companies to make tomorrow's devices as well.

      The thing I really want to see is better syncing of devices. Really I don't NEED more than 64GB or 128GB on a laptop, certainly not for my day job... and I'm one of the biggest users of HDD storage at my company. If there was a good, solid way to sync to a 1TB drive easily an

      • >>HDD tech is established... it's like complaining that 1.5 TB disks cost $100 when a backup tape the same size costs $50.

        Except two or three high speed HDDs in a RAID configuration will outperform a SSD on most tasks, and cost about 10x less.

        I keep looking at SSDs, but their price and performance just aren't where they need to be to buy one other than... just 'cause they're cool. If they even had a 512GB model available for $200 (which is twice as much as I paid for my three 500GB drives 6 years ago)

        • by jbolden ( 176878 )

          No a mediocre SSD will crush a very expensive RAID-5 in read tests. If you go up to something like RAID-15 across 30 drives or so, yeah the raid will be better. SSD really is much faster otherwise no one would be all that excited about paying 20x the cost.

        • by dc29A ( 636871 ) *

          Except two or three high speed HDDs in a RAID configuration will outperform a SSD on most tasks, and cost about 10x less.

          Of course, however ...
          (1) SSDs own platter drivers in random read/write. What people do on their machines? Mostly random reads/writes!
          (2) SSDs consume less power, are dead silent, have no moving parts.
          (3) You can't cram three 15k SCSI drives into a MacBook. Well ... maybe you could but it would look very funny and be way less portable.

    • by jbolden ( 176878 )

      -- How much longer will these be 20x the per GB cost of a HDD?

      They spread will come down but there will be a spread until long after no one is using HDD. http://en.wikipedia.org/wiki/File:Disruptivetechnology.gif [wikipedia.org]

      I'd agree with your 20x right now at say the 256 g - 500 g levels. But lets say when SSD is around 3t the spread might only be 10. At 20t the spread might only be 5. And maybe at 20t people start to switch in mass. This at first drives the spread back up close to 10 but within 2 years, the spr

  • by Chuck Chunder ( 21021 ) on Monday February 28, 2011 @11:25PM (#35344758) Journal
    Somebody is confused. Thunderbolt is DisplayPort and PCIe bundled together.

    The SATA 3 ports on Cougar Point platform have nothing to do with Thunderbolt.
    • I believe it's intended to be used for external hard drives as well. Lacie and WD have signed on to develop compatible hardware - I imagined at the time it would be in that vein.

      Is there any reason you can't transfer hard drive data over PCIe?

    • by syousef ( 465911 )

      Somebody is confused. .

      With Intel's naming (lack of) conventions I'm surprised anyone ISN'T confused.

  • I know this problem has (probably) been satisfactorily addressed but if one were to use such a super fast drive for an application that had extremely heavy usage (swap space for the OS or an program like Photoshop) wouldn't it cause those sectors to be read/written to many many times very quickly? Doesn't each "cell" have a limited number off times it can be accessed before it fails? (on the order of 100,000 i think). And wouldn't that case the drive to fail (sooner rather than later because it is so fast

    • Re:Wear usage? (Score:4, Informative)

      by pz ( 113803 ) on Monday February 28, 2011 @11:39PM (#35344818) Journal

      Again, I'm sure the SSD drive manufacturers have looked at this problem very closely, I'm just concerned that's all.

      So, look up the specs, then. Current write cycles are over 1,000,000 per cell. Modern wear-leveling algorithms combined with extra blocks and ECC mean that it's more likely that some other component will fail before your SSD will.

      Besides, if you were really concerned, and not just trolling, wouldn't you have the same issues with your hard drive, too? Doubly so in a laptop?

      • Current write cycles are over 1,000,000 per cell.

        Far from it. Depending on the litho, at 25nm for instance you're down to 3,000 program/erase cycles. And even at 34nm, you're still little better than 10k cycles. The overprovisioning and ECC required at these scales is massive.

        But yes, studies have been done and it takes an industrial strength workload to kill an SSD. If one of these is in your home machine, you likely won't kill it. If you think you might, then you should already have practices in place to

        • I'm paranoid... generally redundant backups over raid redundancy as Ive had raid1 with both drives dead relatively close together, one time before I had a replacement, my nas box has a spare next to it. Though my intel ssd died att 11 months in my desktop, I still won't go back for my boot+os ...
        • Re:Wear usage? (Score:5, Informative)

          by TheEyes ( 1686556 ) on Tuesday March 01, 2011 @01:49AM (#35345444)

          But yes, studies have been done and it takes an industrial strength workload to kill an SSD. If one of these is in your home machine, you likely won't kill it. If you think you might, then you should already have practices in place to deal with disk failure.

          Just as important to note is the failure mode for flash memory is for it to become read-only; in other words, it simply becomes impossible to delete what is written on your drive, which is a perfect reminder to get a new one. Given that this sad event will be nearly ten years from now, it should be dirt cheap to buy a replacement drive.

          When you do, though, don't forget to remove the metal shell on the old drive and cook it in the microwave for a minute or two to destroy your old data. It's not like you're going to be able to sell the drive used anyway.

          • Just as important to note is the failure mode for flash memory is for it to become read-only;

            Are you sure? I was under the impression that it worked like EPROM, where the bits were set high by the erase cycle and data was written by grounding the bits which needed to be zero.
            That being the case, it's more likely that the data would be corrupted (since it would fail to set bits anymore) which is actually the sort of thing one of my old USB keys started to do.

            'Course, there might well be logic in the controller to detect this and put the drive into read-only mode when it runs out of non-defective

          • by klui ( 457783 )

            No, it doesn't work that way. Each cell contains a charge that is the pattern for a single or multiple bit. If wear leveling doesn't move the cell to another while the drive is turned on, the charges will leak and your pattern will either be all ones, all zeroes, or somewhere in between. The directory structure will also be corrupted. Yeah, you won't be able to write new data, but you probably won't reliably read what's already written.

        • If one of these is in your home machine, you likely won't kill it. If you think you might, then you should already have practices in place to deal with disk failure.

          How about a vicious piece of malware? Could a piece of code be written to circumvent the wear-leveling algorithm and carpet-bomb your SSD with repetitive writes so that it's worthless overnight? Could be a real PITA in cases like the Macbook AIR where the SSD drive is built into the mobo. It's not a case of just paying for a new SSD to replace.

          • The same piece of malware would destroy the contents of an HDD, since it overwrites all sectors. How often do you see that?
          • by dargaud ( 518470 )
            Yeah, I wrote some bootloader code recently to write a kernel to a flash without wear leveling (very low level embedded stuff) and I was really nervous the first time I ran it: if I'd goofed the code and written to the same cell at each iteration, it would have taken about one second to fry the flash !
          • From what I've read, it would take constantly overwriting the entire drive continuously for years to cause it to fail.

          • by Yvan256 ( 722131 )

            I don't know about the first model(s), but the latest MacBook Air uses SSD modules [fosketts.net], it's not built into the motherboard.

      • by Kjella ( 173770 )

        So, look up the specs, then. Current write cycles are over 1,000,000 per cell. Modern wear-leveling algorithms combined with extra blocks and ECC mean that it's more likely that some other component will fail before your SSD will.

        Are you looking at MTBF numbers or something? Expensive 34nm flash has 10,000/cell, cheap 34nm 5000/cell and 25nm is down to 3000/cell.

        And yes, you can kill an SSD that way, already done it. Granted, I tortured it in pretty much every way possible by running torrents and Freenet 24/7 and keeping it 90%+ full all the time. It died after about 1.5 years with 7000 writes/cell average, 15000 highest.

        Fortunately for me I still managed to sneak it in as a warranty repair, even though they aren't supposed to do t

      • So, look up the specs, then. Current write cycles are over 1,000,000 per cell.

        That's only almost true (it's about 100.000 reqrite cycles per cell) for SLC cells. SSDs unfortunately use MLC cells with a rewrite cycle number of 5000.

        But hey, never miss an opportunity to call someone a troll, right?

        • Too bad Newegg lists 17 SLC SSDs. They are also insanely expensive at $10/GB. What was that about trolling?

    • This is why I believe in hybrid setups. The other replies to this are talking about mere millions of writes. Do you know how many writes the swap file gets every hour you use your computer? I too would be worried about a pure SSD setup. If you check out the how-to guides on installing SSD's in your machine, they almost all mention to set the swap on a non-SSD and to move your home directory off it too. (Home directory contains such things as temp web files!)
      • Insofar as actual swap (as in paging physical ram in and out) goes, it depends heavily on the amount of ram the machine has and the type of work load. Mostly-read workloads with data sets in memory that are too large can be cached in swap without screwing up a SSD. A large dirty data set, however, will cause continuous heavy writing to swap space and that can shorten the life of SSD-based swap very easily.

        A system which needs to page something out will stall badly if the pageout activity is rate-limited,

      • This is why I believe in hybrid setups. The other replies to this are talking about mere millions of writes.

        There are no replies talking about mere millions of writes. The one reply that did mention 1 million was talking about per cell and is also very wrong. Furthermore, we do not speak of write limits with SSD's. We speak of erase limits. We should learn the difference before being critical. SSD's write in 4K Pages but erase in Blocks. OCZ is currently using 128 Pages per Block, so Blocks are 512K on their devices.

        Do you know how many writes the swap file gets every hour you use your computer?

        You are attempting to invoke a nebulous unknown to support your argument for hybrids. When you k

      • Re:Wear usage? (Score:4, Interesting)

        by TheRaven64 ( 641858 ) on Tuesday March 01, 2011 @05:46AM (#35346144) Journal

        Do you know how many writes the swap file gets every hour you use your computer

        Not specifically, but my OS - like pretty much any modern OS - gives me a disk I/O total since the last reboot. I last rebooted 11 days ago. Since then, I've written just under 50GB, and read 27.6GB. I don't think reads wear out flash, so we'll ignore the second number. That works out at 1660GB written per year, if the last 11 days have been representative. Assuming perfect weak levelling, that's 6.4 rewrite cycles per year for a 256GB SSD. A drive that can 'only' handle 3,000 rewrites will therefore wear out after about 450 years. If it does a tenth as well as that, then it will last for almost as long as hard drives have existed.

        The advantage of hybrids is the same as the advantage of CPU cache - you get almost the same performance as a very large SSD for a much lower price. Given infinite funds, you'd build a computer that had gigabytes of SRAM, but it's much more expensive than DRAM, so you only have a few megabytes of SRAM and a lot of DRAM, which gives you almost as good performance but costs vastly less (the relative complexity means that the lower bound for SRAM is about six times the price per bit of DRAM - in practice it's higher). If someone else is paying, I'd take the pure SSD solution. If I had to balance price against performance, then ZFS's L2ARC is currently likely to be the best solution (unless it's a laptop, where space is an issue).

        It's worth noting that Oracle is currently looking at using rotating disks as L3ARC, with tapes as the persistent storage. Writes remain fast, because ZFS is copy-on-write, so every write to the tape is just an append operation, but L3 cache misses are very expensive (you need to seek the tape, which can take several seconds, rather than several milliseconds) - for write-heavy workloads, it can give very good performance per dollar.

    • Hi! Its called ram. Its where you store data that you plan to access and manipulate a lot before you save it.
    • by jbolden ( 176878 )

      Its a large number of writes not reads. Something like swap space gets allocated and even after it no longer reflects memory it stays empty for "a while". As long as "a while" is an hour or more on average you are fine.

    • Swapping to flash is dumb. But more importantly, with modern memory prices, swapping is dumb. You're not going to end up in swap unless an application misbehaves, and then the swapping is going to bring your system to a standstill. Better to run out of memory and let the OOM system kill processes. You can get 4GB SODIMMs now. Swap? That's over.

  • ought to be severely reduced if you can pass the info off at those data rates to a similarly fast external drive.

    Then if you want to archive to a "slow" spinning hard drive, the external SSD could supply the data at the slower rate of the HD

    • It depends: You could use a Raid 5/6 NAS and have high disk speeds again. Something like this [softpedia.com]. The max speed of a thunderbolt connection is 800MB/s so, assuming they are WD black 2TB 3GB/s drives, you can get an average of 109*5 (raid 5 with 6 disks) = 545 MB/s (host to disk speeds. Host to buffer doesn't matter).
      That's, effectively, 5 disks to go slightly over their read speed. Then again, you'd get 10 TB (=9.09 TiB) on the external enclosure, while the biggest of these will be 250 GB.
  • With some lazy writing to solid-state-chips.. :D

    Yeah, I am dreaming.. Sigh!! :-/

    • The DDRDrive X1 almost fits your design. It's not on the memory bus, but on the PCI-e bus as a storage device. Bit pricey though per gigabyte.

  • What happened to the 3rd generation SSDs that Intel was supposed to release this month? They were supposed to be using 25nm flash and offer roughly twice the space for the same price as the G2 drives. Using a new controller and upgrading things a bit seems a poor substitute for that.
  • IOPS?

    It's important to know. [blogspot.com]

  • Now that 6GBit/s SATA ports are becoming commonly available on motherboards it's only natural that SSDs follow. Normal HDs can't really take advantage of a 6Gb/s port but SSDs can. These high speed ports will make port multiplier enclosures more useful as well.

    There's certainly a lot of use for this sort of thing. SSDs can already replace far more expensive DRAM (and the far more expensive motherboards needed to support lots of DRAM) for numerous problem sets, including mostly-read database accesses. Th

    • With the limited write cycles for SSD DRAM will not go away. Could you imagine your RAM to crash permanently after a long weekend of gaming?
  • It's been almost 30 months since the X-25e was launched and now the max 64GB capacity looks pathetic next to the competition so when will Intel launch their next generation SLC based SSD?
  • by 0111 1110 ( 518466 ) on Tuesday March 01, 2011 @03:01AM (#35345654)

    Sandforce has already announced [anandtech.com] its new sata3 controller. On paper it looks like it will have much faster sequential writes than Intel, but it sounds like it will also have a shorter lifetime and shorter data retention times due to the use of 25nm NAND. Intel is wisely sticking with 34nm. It may be more expensive to manufacture, but is superior tech. I can only hope that OCZ changes their mind and decides to at least offer a more expensive 34nm version. OCZ won't be shipping their Vertex 3 drives until Q2 so Intel will have a big head start in the market.

    The NAND industry seems to be doing its best to encourage ignorance on the disadvantages of smaller process sizes from the consumer POV and the ignorance seems to be widespread. Getting the facts on this issue can be a bit difficult. Here is a good thread on the topic.
    http://forums.anandtech.com/showthread.php?t=2142742 [anandtech.com]

    The following post sums it up better than I could. Note his point about data retention times as well. That is a point that is often ignored when the focus is solely on write cycles.

    As flash cells are shrunk, they become less good. This is a fundamental feature of the technology. The overall volume of the cell becomes smaller, so less electrons can be stored in the cell (so the signal picked up by the electronics is weaker and less clear, so you get a higher error rate) and the insulating barriers around the cell must be made thinner, in order to save space - allowing the electrons to leak out of the cell more easily (reducing power off data retention time). The thinner insulation also wears out more quickly (reducing life cycles)

    It's difficult to define a 'fundamantal' limit for flash, because it may be possible to work around poor performance, and as yet unknown new manufacturing techniques and semiconductor materials may be developed. However, it has been suggested in the scientific literature that 18-22 nm, is the realistic limit. Beyond that, the performance/reliability/lifespan of the flash would be too poor, no matter how much wear levelling, and how sophisticated the ECC codes were.

    Enterprise grade SSD flash, will need higher specifications than flash for toy cameras. Enterprise applications are unlikely to tolerate 18 nm flash with 100 write cycles and one lost sector per 100 GB of data stored. However, this probably would be acceptable for toys or throwaway devices.

    Some more coverage of the topic:
    http://techon.nikkeibp.co.jp/article/HONSHI/20090528/170920/ [nikkeibp.co.jp]

    NAND Flash memory quality is also beginning to drop. Chips manufactured using 90nm-generation technology in 2004-05, for example, were assured for about 100,000 rewrites and data retention of about a decade. As multi-level architecture and smaller geometry are introduced, quality is showing a sharp decline. The 30nm 2-bit/cell chips expected to enter volume production in 2009-10 may well end up with a rewrite assurance of no more than 3,000 cycles, and a data retention time of about a year. The first 3-bit/cell chips are hitting the market now, with only a few hundred rewrites.

    http://hardforum.com/showthread.php?t=1502663 [hardforum.com]

    Flash memory works by trapping electrons. Over time these electrons leak away, until the charge is too small for the data to be read any more. With smaller feature sizes (34 nm instead of 45 or 65 nm) this leakage is more significant and fewer electrons can be stored per bit, thus the time during which the stored value can be maintained is decreased.

    http://www.corsair.com/blog/force25nm/ [corsair.com]

    • Thanks for the info :)
  • by Eunuchswear ( 210685 ) on Tuesday March 01, 2011 @04:15AM (#35345890) Journal

    Give us fucking SAS already.

  • I have an Intel X25-M SSD in my Mac Pro and use an OCZ Vertex 2 in my Macbook. Both work very well, although the OCZ feels faster. One thing to watch out for though, is the way firmware updates are handled: Intel offers a bootable ISO image that will update the firmware very easily and without any hassles whatsoever.

    OCZ on the other hand only offers an .EXE tool (32bit only!) that needs an Internet connection and only works if your SSD has an MBR partition style and at least one NTFS formatted partition o

  • The Intel X25 series was so much better than the alternatives when it came out, I was expecting more from its successor. The 510 doesn't even use an Intel-developed processor. It uses one from Marvel, which will eventually be used by other SSD manufacturers as well. And the 510 is already slower than drives based on Sandforce's newest chips, such as OCZ's Vertex 3. Lame, Intel.

Beware of all enterprises that require new clothes, and not rather a new wearer of clothes. -- Henry David Thoreau

Working...