Comments Locked

83 Comments

Back to Article

  • Kitohru - Thursday, September 10, 2009 - link

    Does OS X Snow Leopard have trim support, and if not any word from apple about that or the like?
  • JakFrost - Sunday, August 30, 2009 - link

    I've bought this drive a few months ago due to the Storage articles here including the original SSD Anthology article.

    I hope that the issues with TRIM with Intel firmware and Intel Storage drivers is clarified and documented better because right now it is quite confusing.

    I already use Windows 7 RC 64-bit so far with latest Intel Matrix Storage drivers 8.9.0.1023 published on Intel's website.

    However, now I'm strongly considering selling my G1 drive on eBay and buy a G2 for about the same price to ensure that I get TRIM support in the future because it seems like Intel has forsaken the G1 users.
  • Hank Scorpion - Thursday, August 27, 2009 - link

    Anand,

    Keen follower of your website, and your reviews, LOVED SSD anthology and am champing at the bit for your next SSD update.

    Can i ask a quick one please,

    im buying a new PC once Windows 7 comes out in October. And am going to buy a 256GB SSD as my main drive. thanks to you and PcPer's SSD decoder ring, ive decided that the OCZ Vertex or Intel SSD is for me (basically anything with an Intel or Indilinx controller) i was ebbing and flowing between the Vertex and the Summit until i saw your warnings posted just recently about avoiding the Samsung...

    which one should i get? and also cost is a big one, anything with the Indilinx sounds like a great get, BUT i also see Corsair/G-Skill/Patriot all with a similar product but with differing prices.

    Should i go for the cheapest option (G-Skill) but suffer when it comes to firmware updates/support? or go the extra $150 AU to go OCZ as they seem to provide a better support with Firmware updates???

    Please let me know what you think!! as i dont want to pay a lot of money and get something that i wont be happy with.

    Im purely a gamer, and dont do video editing or anything like that....

    Thanks
    Hank!
  • ekerazha - Wednesday, August 26, 2009 - link

    On 24th has been released a new version (1.3) of the Intel SSD firmware, but there's no changelog. Is anybody aware of the changes?
  • mmaenpaa - Wednesday, August 26, 2009 - link

    It seems that it only affects X25-E (50nm) products. Others saty at the previous version.

    Quote (from releasenotes)

    X25-E (50nm product) Revision History
    Date Revision Description
    25 August 2009 045C8850 This firmware revision is for X25-E products only and has several continuous improvement optimizations intended to provide the best possible user experience with the Intel SSD

    markku
  • mmaenpaa - Tuesday, August 25, 2009 - link

    Hi,

    Just wanted to warn if anybody is going to try upgrading with Acer notebook.

    I just installed this SSD (X-25M G2, 80G) in new Acer TM6293 about 2 weeks ago. And I must say this is probably the biggest improvement in speedup feel I have seen in 5 years in this business. Very snappy indeed.

    As it happens I noticed the warning about the "bios password" feature after the notebook went to the customer.

    I took it in for quick fix (last thursday) and of course after firmware update the notebook does not recognize the SSD anymore. SSD itself seems to be working as other machines recognize SSD OK. Naturally (lucky me) I did make a full image before and now customer has an WD Scorpio Black 250GB mechanical drive in his new notebook. Boy it feels so slow compared to SSD. Even customer noticed it which is not always the case with not so advenced users.

    Now I have been waiting for Intel or Acer to contact me. I guess both emails got lost in the customer support system.

    As it happens both Intel SSD and Acer have the latest official firmware / bios so I am stuck for the time being.

    Any hot contacts with Intel (or Acer), anyone?

    Markku
  • StuR - Tuesday, August 25, 2009 - link

    I've downloaded and read the product datasheets and manuals and I don't see any published information from Intel showing the support for the TRIM command on any of their current SSD drives. How is a consumer supposed to intelligently purchase this technology when a lot of important information comes only from a Blog?
  • MrPoletski - Tuesday, August 25, 2009 - link

    I never realised Anand was a Mac fan..

    I'm pretty sure that's a little apple symbol in the top corner of your 'here is the article text' image ;)
  • Black Jacque - Tuesday, August 25, 2009 - link

    The table in the article SuperTalent UltraDrive GX 1711 is showing bandwidth, not the labeled 4KB Random Write IOPS. IOPS would be an integer value.
  • strafejumper - Monday, August 24, 2009 - link

    seen some talk about the ion platform on media player websites
    seems an ion pc is starting to come down in price to the point it is close to the same price as a dedicated media player such as the popcorn hour.

    They seem to suggest you have to install linux on it and XBMC to get hardware acceleration
  • NeBlackCat - Monday, August 24, 2009 - link

    I'd like to see some analysis of which is best for software development. It isn't clear to me whether compiling a large project, say the Linux kernel, would favour the small write performance of the former (which I suspect) or not.
  • eganvarley - Monday, August 24, 2009 - link

    Next thursday this Linux Weekly News article will be available for all (currently it's behind a pay-wall) : http://lwn.net/Articles/347511/">http://lwn.net/Articles/347511/

    In this article there is an critical enquiry about the TRIM command. It will be interesting for Anand to comment on this article. An extract :

    "At the ATA protocol level, a discard request is implemented by a TRIM command sent to the device. For reasons unknown to your editor, the protocol committee designed TRIM as a non-queued command. That means that, before sending a TRIM command to the device, the block layer must first wait for all outstanding I/O operations on that device to complete; no further operations can be started until the TRIM command completes. So every TRIM operation stalls the request queue."
  • leexgx - Monday, August 24, 2009 - link

    The way you posted is correct way , It should be doing it after any other i/o any way that is correct the trim command is sent to the drive its then up to the drive when to do the trim command when the drive is idle then it do clean up (the act of sending trim command does not mean it has to do it, it waits or that's how it should be implmented why it does not need to be qued)

    But I get the point that if trim command is used disk queing can not be used (or it seems that way )
  • eganvarley - Tuesday, August 25, 2009 - link

    If you read the LWN article you will see that the TRIM command on the ATA protocol is different than the TRIM command on the SCSI protocol. It's much, much better on SCSI.
  • techvslife - Monday, August 24, 2009 - link

    I'm suprised this hasn't received discussion here, but the 1711 driver definitely causes massive data loss with Windows 7 x64 when TRIM is enabled (so using the Intel drivers, which block TRIM, will not show the problem, but the MS drivers which allow TRIM will show the bug).

    http://forum.crucial.com/t5/Solid-State-Drives-SSD...">http://forum.crucial.com/t5/Solid-State...-SSD-sup...

    http://www.ocztechnologyforum.com/forum/showthread...">http://www.ocztechnologyforum.com/forum/showthread...
  • leexgx - Monday, August 24, 2009 - link

    Its beta at the moment, the 1.31 firmware best to use as it has self heal (needs 15% free for it to work or 20gb free (120gb verson))
  • techvslife - Monday, August 24, 2009 - link

    It's beta at ocz (they caught the problems just as they were about to release it), but 1711 has been released as final by crucial and by supertalent, and possibly others. Therefore, make sure NOT to install 1711-based firmware if you have an SSD drive using an Indilinx barefoot controller (it might not be labeled as beta!).
  • Visual - Monday, August 24, 2009 - link

    Why exactly do SSDs slow down with use, and how does TRIM help with that?

    Have I understood correctly that it is only the write speeds that slow down, not the read speeds? And the reason is that the drive can write faster to completely blank "cells" but is slower when writing to already used cells because it needs to first erase them?

    Then, this TRIM thing is essentially just erasing a block to prepare it for faster write in the future? But I still don't see how does TRIM support help any, as when the time comes to write some stuff, doing TRIM and then write should still take as long as without an explicit TRIM... Doing TRIM right after a delete also seems like it would not save any time, as it will make deletes slower...

    Or is TRIM supposed to be used not exactly before/after a write, but scheduled in the HDD's free time, sort of like defragmentation for the old magnetic drives?

    I frankly doubt that I've understood the concept, and I would appreciate some explanation in simple words - in the comments here or even in the final article, as I am sure a lot of other readers will benefit from it too.
  • Visual - Monday, August 24, 2009 - link

    Oh sorry I forgot one more question:
    Can someone explain about SSDs block sizes, how to find what they are, how they relate to filesystem block sizes, RAID clusters block sizes, alignment between them, possible impact on performance and such stuff?
  • Sunraycer - Monday, August 24, 2009 - link

    Hi Anand, I also am still interested in Ion.

    I am running an Intel D945GCLF2 with Atom 330 as an HTPC. I'm thinking up upgrading later in the year for Blu-Ray and HDMI. I've almost decided to go with a Core 2, but still wish the Atom could do what I want.

    Right now with the D945GCLF2 I can just get by with running full screen ABC/FOX shows (VP6?) on my TV at WXGA (720p). But full screen flash (Hulu/Comcast) doesn't play. DVD playback is fine. Watching NTSC and ATSC (OTA or ClearQAM) is fine (ATSC barely works) (Hauppauge 1600). Recording NTSC is fine, but ATSC is all choppy.

    An insight into how ION will handle these cases and Blu-Ray playback, beyond the great articles you've already done would be great.
  • Pirks - Sunday, August 23, 2009 - link

    Why Pages and not Word?
  • Tutor69 - Sunday, August 23, 2009 - link

    Hey Anand, will Intel update the first generation X25-M with trim functionality? I'll be quite upset if they won't.
  • StuR - Tuesday, August 25, 2009 - link

    Ditto. I'm getting tired of buying SSD's only to find out they're lacking important functionality (not mentioned by the vendor of course). I don't see any reason the X-25G1's can't support TRIM. And I expect that the Intel Matrix AHCI driver should start to pass the command in future releases too!

    Why hasn't Intel been more upfront about all of this? Why do we have to get this information from a second-hand source?
  • cliffman - Sunday, August 23, 2009 - link

    I would like to see the zotac mini itx board reviewed. I have been wondering how it compares to an atx board in speed. Some review sites say it has comparable performance while others show it being 5% slower. Also any information on core i5 mini-itx would be cool to read about.
  • StormyParis - Monday, August 24, 2009 - link

    5% is comparable. I challenge you to spot the difference during anything but benchmarks (or d**k size comparisons).
  • strikeback03 - Tuesday, August 25, 2009 - link

    Well, if you were transcoding a lot of movies, that 5% could definitely add up to a lot more time spent.
  • Zab00 - Sunday, August 23, 2009 - link

    I'm looking forward to any new SSD reviews, Anands seem to be the most thoroughly done of the tech sites.
    About SSD review, up to now I couldn't find any reviews on Toshibas products (up to 512GB). Aren't they interesting?

    In preparation of buying a SSD drive I realized that my only one year old PC with a Intel G31 (ICH7) doesn't support AHCI. As far as I know that means no NCQ available and also TRIM won't work.
    Is there a big deviation from the numbers in the reviews to expect when running in the IDE emulation?
  • graysky - Sunday, August 23, 2009 - link

    According to both wikipedia and a reply in this thread, the TRIM command has been implemented in Linux since kernel version 2.6.28 yet I can find no mention of it in the kernel docs (http://www.kernel.org/doc/menuconfig/x86.html)">http://www.kernel.org/doc/menuconfig/x86.html).

    1) Does the kernel auto detect and use TRIM?
    2) Does one need to modprobe a module to use it?
    3) How can one verify that TRIM is actually in effect under Linux?

    I am on waiting to purchase the 80 GB Intel G2 SSD for a Linux system using 2.6.30, but I do not want an SSD wo/ TRIM support.

    Thanks!
  • Fietsventje - Sunday, August 23, 2009 - link

    Hello Anand,

    Concerning what I would find interesting in an Ion platform article would be the network performance.

    I think an Ion board would be perfect to serve as a base for a very basic server. Using it as a file server board (despite the limited number of SATA-ports) would require it to be able to achieve at least 50 MB/s during file transfers.

    It would be nice to see these things measured and maybe even compared to Intel-based Atom boards.
  • strikeback03 - Monday, August 24, 2009 - link

    Would you really need the video enhancements of Ion for a server?
  • Fietsventje - Monday, August 24, 2009 - link

    Nope, I wouldn't, but small differences in the ethernet controller efficiency could mean quite a difference in performance since the Atom hasn't got a lot of processing power to spare ... And since the nVidia MCH is more powerful than Intel's offerings, it could offer just those improvements needed to get 'decent' gigabit network performance.
  • StormyParis - Sunday, August 23, 2009 - link

    Like many PC enthusiasts, I'm not at the stage where I want to choose WHICH SSD to buy, but rather wondering WHETHER I should buy one at all. I'm currently running off a 1.5 gig 7002.11, which was about half the price of a 64 megs SSD.. and I'm not suffering, or I'm not aware of it.

    So please, include a regular, mechanical HD(7.200 rpm) in the tests, and make the tests somewhat intuitive (boot times, level load times...).

    Thanks !

    Olivier

    PS and please explain what TRIM actually is: is it transparent, or do you have to run it once a week (more ? less ?), and how long does it take ? Is there a utility that tells you if should TRIM, like Win's defrag does ?
  • leexgx - Monday, August 24, 2009 - link

    If you have up todate firmware on on ssd most vertex based ssds will self heal the samsung bases ssds do this as well (corsair p128)

    You need to keep 15% free on the ssd for it to work correcty, if you have windows 7 it supports trim command and will not matter how full the disk gets (but no ssd has firmware support for it yet only in beta form but will be comming soon)

    Ocz norm get firmware updates first (best looking at the forums ignore tips for setting up drive corectly as they take it to far)
  • Ben90 - Sunday, August 23, 2009 - link

    Im not very good at explaining things like trim, read one of the earlier AT reviews where they find out about the write degrading over time if u want to understand what it does, and why its needed. But basically its invisible to the user as long as you are on an operating system that supports it
  • Badkarma - Saturday, August 22, 2009 - link

    I don't recall if you touched upon advanced deinterlacing for 1080i material. Does the Ion do motion or vector adaptive deinterlacing for 1080i? How about 1080i x264?
  • Beno - Saturday, August 22, 2009 - link

    can somebody tell me how to secure erase the intel drive?
  • semo - Saturday, August 22, 2009 - link

    http://www.anandtech.com/storage/showdoc.aspx?i=35...">http://www.anandtech.com/storage/showdoc.aspx?i=35...
  • rfnanand - Friday, August 21, 2009 - link

    I registered on the forums to ask only one question:
    WHY CAN'T I FIND ANY X25-M G2s ??
  • vailr - Saturday, August 22, 2009 - link

    A certain podcast (dated Aug. 20) from pcper.com mentioned that Newegg said they should have some stock of the Intel G2 SSD's by about Aug. 28.
  • yacoub - Friday, August 21, 2009 - link

    Temps, wattage, and sound levels are always appreciated.
  • iwodo - Friday, August 21, 2009 - link

    Well, what i like to see in SSD and ION. Basically the relate to Future.

    How are we ( or Intel ) is going to Scale SSD's performance? Indilix currently uses 2 - 4 Channel, Intel uses 10. But I guess using more channel would require more space, and it seems the current 10 channel 2.5" Intel using is already Max out by 10 Chip on a single Side. Using Newer OFNI 2.0 Chip will push the Raw Speed up to SATA 3.0 Max. But Random Read and Write, IOPS won't improve much at all without Getting More Channel chips like the Fusion IO.

    Are we seeing going to hit performance wall soon with SSD?

    ION - It may be wise to collect some Data now for HD decoding CPU usage. I read else where it is using much lower CPU usage then the upcoming Intel Core i3 with IGP.

    What about the Future of ION? There are news with MCP 89 coming, I have been expecting Nvidia to use PCI-Express 2.0 to collect with their IGP. ( Which is just a GPU with Southbridge ) But how will it work with Core i3 with two GPU.
    ( And on the subject why is Intel Forcing us IGP when we dont want it, i just want a Cheap Westmere without IGP )

    Looking forward to your article...
  • Gutcheck2009 - Friday, August 21, 2009 - link

    So I bought a Patriot Torqx 64GB based on your article in CPU. I can still return it, does it support TRIM? I must say it is very fast.
  • MadMan007 - Friday, August 21, 2009 - link

    Yeah so touch interface is great for some things but tactile controls needn't be abandoned or seriously marginalized. I'm just afraid that with the 'coolness' of Apple touch iStuff that companies will feel forced to go touch as much as possible in order to not appear old fashioned.
  • marks70 - Friday, August 21, 2009 - link

    I don't see how it's evenly remotely acceptable for a drive to lose the amount of performance these SSD drives do after such limited use. Shouldn't this have been a major consideration when designing this technology? I know they're still faster than your typical hard drive even after they slow down, but should they really require this much intervention to keep them up to speed?

    The amount of "work" required to maintain these drives, in addition to their price, makes me questions if they are really ready for prime time.
  • sheh - Friday, August 21, 2009 - link

    I hope the situation with HDD->SSD doesn't mirror what happened with CRT->LCD, a new technology superceding an older one before it has equaled or bettered it in all aspects.
  • MadMan007 - Friday, August 21, 2009 - link

    I don't see that happening because storage is much more of a 'background' thing than a monitor. You don't interact with it directly per se, you interact with programs that in turn use the storage. As long as they put out the same over the SATA interface the underlying details don't matter versus monitors where the underlying technology directly creats the user experience. And SSDs are already well on the way to being better all around than HDs aside from space.

    Mechanical drives will have a $/GB advantage for quite a while as well. It will be a good while until there are $200 2TB SSDs ;) plus mechanical drives might keep gaining density although it seems to have slowed a bit after the explosion thanks to perp recording. If it turns out that all mechanical drives are best at is massive space so be it. Maybe a return to 5.25" drives? lol
  • Rebel44 - Friday, August 21, 2009 - link

    Thats 13MB in 4K random writes - almost all HDDs have less than 1MB in 4K random writes. :)
  • JarredWalton - Friday, August 21, 2009 - link

    Yup. It's pretty simple, really. For a 7200RPM drive doing 4K random transfers:
    Average seek time = ~12ms
    Seeks per 1000ms = 83.3
    Data transferred randomly in 1000ms = .326MiB

    Switch to a faster 10K RPM HDD and you get:
    Average seek time = ~8ms
    Seeks per 1000ms = 125
    Data transferred randomly in 1000ms = .488MiB

    Or a super fast 15K SCSI setup:
    Average seek time = 4.3ms
    Seeks per 1000ms = 232.6
    Data transferred randomly in 1000ms = .908MB

    Basically, as long as you have to reposition the drive heads there is no possible chance for conventional HDDs to come anywhere near SSD speeds when doing random transfers. Unfortunately, there are a lot of instances when you have to do more or less random transfers. Defragmenting an HDD might get you up to 5MB/s on pseudo-random file accesses (i.e. "normal" use), but unless you're doing file transfers HDDs are really a bottleneck.
  • glugglug - Monday, August 24, 2009 - link

    Your "seek time" calculations are just calculating the time it takes for the disk to spin around (which is part of *access* time, NOT *seek* time). Seek time is the time for the armature to move from one track to another. To get access time, which is what you are really interested in, the 2 numbers need to be added together.

    And the average time to spin to the right sector (once the appropriate track is located) is half of what you are quoting - needing to spin *all* the way around is worst case.
  • JarredWalton - Monday, August 24, 2009 - link

    What I'm quoting is more or less approximate to average access time - half the rotational latency, plus a seek time that ends up being somewhere between the best-case and worst-case. The point isn't to be 100% exact, but to show how random access seriously handicaps performance of HDDs. If you were doing larger file accesses, the transfer rates are greatly improved - it's the small 4K random access pattern that's a killer for *any* conventional HDD. Even if you could get an HDD to 1ms average, you'd still be looking at 4MB/s compared to SSDs that are already hitting three times that rate.
  • bobvodka - Friday, August 21, 2009 - link

    That really isn't the case here.

    I'm currently copying my Steam folder (around 100gig) from my F1 spinpoint to a OCZ 250gig SSD. The transfer rate peaked at 90MB/sec during large files and then dropped and held at 75MB/sec when it started copying smaller, fragmented, files and the drive was haing to work harder to seek.

    Down to 60MB/sec now for .ogg files and the like with the hard drive working like mad.

    As a contrast I've NEVER seen transfer rates that fast with a HDD, even a 10,000RPM drive.

    The only 'downside' of SSDs right now is the price:size issue. But the price is dropping as time goes on and early adopters create a bigger market.

    My biggest comment when it comes to SSDs is unless you can afford to start replacing HDDs all over the place with them then avoid them, because they really are that impressive when you get used to them.

    As a side note, the current BETA Firmware from OCZ (1.42B) does have a couple of issues. While the beta 1.4 is working fine on my laptop, my desktop has had issues with file corruption which have taken a few days to solve. However between the TRIM support and the GC support in the firmware the final release should be pretty sweet when they have ironed out any issue.
  • MadMan007 - Sunday, August 23, 2009 - link

    I hate to burst your bubble but I get nearly those speeds (70+MB/s) even over GigE with the source and destination drives all being WD Caviar Blacks. Maybe I'm misunderstanding what you're saying though.
  • rajaiitm - Friday, August 21, 2009 - link

    Hey Anand! you should look at this page .. http://vr-zone.com/articles/ssd-stackup/6842-7.htm...
    I thought those IOMeter plots were quite informative than just presenting 4KB read/write performance. You should probably have something similar. Its quite interesting to see how different ssds behave differently over the file size range. But I don't know which range is more important and some insight with that would help.

    Sandforce! Any info about those guys? News is that they rocked the show at the Flash summit.

    P.S. Thanks a ton for the SSD Anthology.
  • Zorlac - Friday, August 21, 2009 - link

    Anand -

    Are you saying Intel G2 SSD firmware has TRIM enabled already?

    And if so, then WTH would Intel not have a MSM driver that supports TRIM? Have they told you when to expect the driver update?

    Thanks for keeping us up to date with your findings!! :)

    Zorlac
  • StuR - Tuesday, August 25, 2009 - link

    I'm sure you're correct but before spending hundreds of dollars on new SSD's, it would be nice to see all this (TRIM support) in a published Intel document (X-25M G2). I can't find it. Did I miss something?
  • DigitlDrug - Friday, August 21, 2009 - link

    Hi Anand,

    Thanks for your request for input - your thoroughness in researching the underlying technologies of a product is phenomenal.

    When discussing the Ion platform can you
    1) I think I already know the answer, but it may be helpful for others -> Can you go into detail regarding the implementation for codec acceleration? For example, if H.264 decoding is accelerated, does a specific decode path need to be taken to take advantage of that acceleration? Or is the acceleration performed at the lowest level by the driver, and therefor any player - Media Player, Adobe Falsh (Hulu), etc. - automatically takes advantage of the acceleration.

    2) What level of linux support are we currently seeing with Ion re. Video acceleration? (pertinent when using an HTPC linux distribution). Has NVidia devulged plans for future driver support. Obviously they would never say "Linux Who?" but maybe you have a helpful contact . . . . ; >)

    3) Is the accelerated decode pipeline flexible enough to handle future codecs, and are they likely to be supported via the Unified Driver model? Obviously the atom lacks the horse power to handle decode on the latest codecs, so mpc users looking for "Relatively" future proof players would be relying heavily on the Ion to pickup the slack.

    Re. SSD's - can you include a discussion on the lower end Indilinx models from OCZ? They appear to be introducing different flavors of drive based on the controller, and all reviews thus far have looked at differences in speed, etc. I'm curious to know if they are also using different versions of the controller (in features), and any gotchas beyond speed (future firmware/feature support, longevity). Where is the cost savings coming from?

    Thanks again for all the great work.

    - Jordan
  • Pandamonium - Saturday, August 22, 2009 - link

    I'm still very interested in the Ion coverage. I think Tweaktown or somebody had a rough review of the ASRock BD-330. They claimed to get a stable overclock that was a bit higher than what Anand was able to get out of the Zotac, and IIRC, the GPU clock was slightly higher. Since it looks like the Zotac ION-A can't handle Hulu, I was wondering if ASRock's implementation fared any better. As far as I can tell, the options for a small and relatively cool-running HTPC are an Ion system (might be too slow), a Dell Studio Hybrid (still uses GMA), or a Mac Mini (lacks blu-ray option).

    As far as SSDs, I'd like to see some data on the X18-M or other 1.8" SSDs. I suspect performance is comparable to the X25-M, but I wanted to know the battery life impact of upgrading from a 1.8" HDD to a 1.8" SSD. From what I've read, Intel has some enhanced firmware that reduces power consumption whereas the other manufacturers do not. I'll probably end up with an X18-M G2 anyway, but I'd like the data if it's available. (My XT2 only takes 1.8" drives)
  • chemist1 - Friday, August 21, 2009 - link

    Also, for those of us with slightly older Core 2 Duo macs (e.g., my early 2008 MBP), which are limited to 150 MB/s for the SATA interface, is their any *practical* lessening in performance (vs. w/ a 300 MB/s interface)? It's possible there is not since, as you've pointed out, it's 4K random reads and writes that are important, and these are <150MB/s with current SSDs. [Alternately, do you know if a firmware adjustment might be able to bring those pre-2009 macs up to 300 MB/s? The Intel ICH8-M AHCI chipset used for the SATA interface on my MBP is, I think, in principle, able to handle 300 MB/s.]
  • Anand Lal Shimpi - Saturday, August 22, 2009 - link

    Nearly every single drive I've tested can read at 250 - 260MB/s. You would theoretically lose some read/copy performance there. If Apple hasn't enabled it yet, I don't think there's any hope for Apple enabling it in the future.

    That being said, even limited to 150MB/s, you're still *far* better off with a good SSD than with a hard drive. Especially a 2.5" notebook hard drive.

    Buy an Indilinx or Intel SSD. Try it for a week, then switch back to your old hard drive. You'll forget all about the 150MB/s limit and be begging to put the SSD back in :)

    Take care,
    Anand
  • chemist1 - Friday, August 21, 2009 - link

    Anand, do you know if trim will be supported in snow leopard?
  • Anand Lal Shimpi - Saturday, August 22, 2009 - link

    TRIM is currently not supported in Snow Leopard. It wouldn't be difficult to add it though.

    If Apple would stop shipping crap SSDs and take the technology seriously then we might actually get it in an OS update. I'd expect that it won't take long though.

    Take care,
    Anand
  • kolunda - Sunday, August 23, 2009 - link

    Anand,

    I've been wondering this for a while but I don't currently own an SSD to see if this would potentially work. Would using Disk Utility in OS X to erase free disk space in 10.x be equivalent to performing a TRIM command. Would this boost performance of a worn SSD or is this actually writing to the disk and not improving the performance at all? If not, is there any equivalent way under OS X to do this?

    -kolunda
  • WillR - Wednesday, August 26, 2009 - link

    Short answer: No.

    TRIM and Zero'ing or wiping data are completely different functions. TRIM doesn't overwrite the actual data contained on the drive, it overwrites the literal data of "where information is currently stored". Wiping a drive is like going through an array and setting every element to zero or null and TRIM is like taking the pointer to an array and setting it to null. The data is still technically there, but you've told the program/OS that you no longer care about it. Thus it no longer has to maintain it which is what causes the slow down of used SSDs.
  • sprockkets - Friday, August 21, 2009 - link

    How about using the mini-PCIe slot on the Ion boards for a fast SSD drive instead of wireless?
  • Eeqmcsq - Friday, August 21, 2009 - link

    is TRIM support in Ubuntu and I'll be a happy SSD camper.
  • GourdFreeMan - Friday, August 21, 2009 - link

    That is entirely dependent on your kernel version. The Linux kernel has supported the draft version of TRIM since 2.6.28. Expect to update your kernel and SSD firmware after the specification for TRIM is finalized.
  • acolona - Friday, August 21, 2009 - link

    forgot to ask one last thing.

    I hear everyone say ssd's slow down as you approach their storage limit. How much are we talking...like 2-5%, or like 25%.

    Reason I am asking, is my boot partition usually hovers around 50 gigs....so am I shooting myself in the foot by getting a ~60GB ssd? Do I really NEED a 120Gb?
  • Anand Lal Shimpi - Saturday, August 22, 2009 - link

    It really depends on the architecture of the controller. Intel drives drop a lot in performance, Indilinx drives drop less. It also means that Intel stands to gain more from TRIM...

    If you use 50GB, I'd personally opt for the Intel 80GB X25-M. It's nicely in between the 60GB/120GB pricing and gives you more than enough "breathing room".

    Take care,
    Anand
  • icrf - Saturday, August 22, 2009 - link

    I thought it was quite a bit. Maybe not 25%, but certainly 10. From the numbers posted above, though, it looks like a hell of a lot more than that.

    The kicker has always been that it beats mechanical drives by such a large margin that a 50% slow-down still means it's several times faster than a mechanical drive on the random tests.
  • jimhsu - Saturday, August 22, 2009 - link

    From daily use of the X-25M G2, I find that seq writes, not random writes, are most adversely affected by the lack of free space. When I had less than 5G on the drive, I saw seq writes plunge to less than 30MB/s. Performance seems to be back up in the 70MB/s range after clearing up 15G of space.
  • jimhsu - Saturday, August 22, 2009 - link

    PSS And can someone give an explanation why SSDs slow down with more data filled, regardless of TRIM availability? I thought the point of TRIM was to enable the SSD to recognize certain blocks as free, seeing as an SSD that has been written at least once needs to keep track of all its blocks, empty or not. If so, why does it matter if its real data or deleted data?
  • Anand Lal Shimpi - Saturday, August 22, 2009 - link

    The Intel SSD treats all unused area as free space. The more free space you have, the lower your write amplification resulting in better performance. As your free space goes away, either by using the drive or by actually filling it with valid data, the frequency of block-cleaning (read modify writes) goes up. Your write amplification goes up, performance goes down.

    TRIM makes sure that if you're not filling the drive with valid data, that its performance remains as close to new as possible. This is very helpful.

    I'll explain this in great detail in the next SSD article :)

    Take care,
    Anand
  • MrPoletski - Tuesday, August 25, 2009 - link

    Sounds to me like the solution to this issue would be to do background cleaning of the disk. They say you don't need to defrag an SSD... well is that really true?

    somehow, you've ended up with a usage pattern of the first 4 kb used, the next 4kb free, next 4 used, next 4 free and so on, all the way to the end of the disk. Horribly inefficient storage means and an extreme example of what you disk might end up looking like. It also means that every single overwrite needs to be a read-erase-write.

    so what if you had a little engine running in the background that re-organises data on the SSD to free up individual blocks completely, cramming most of the data down into single, fully used, block size chunks.

    on top of that, why not have your file system point to TWO places for the same data. So you copied that solo 4kb from one block into a partially populated block filling it completely. Why delete the old data? Instead you could make the disk aware that the data exists in two places and decide which to modify (destroying the other copy) in the event of a overwrite and which to forget about in the event of something else needing the space?

    you decide this based on which route would require the most work from the drive.
  • jimhsu - Sunday, August 23, 2009 - link

    Perhaps my question is better rephrased like this:

    How does the performance of a used X-25M w/o TRIM, with 40GB free space compare to one with TRIM, but with say only 5 GB of free space? From what you say, the 2nd drive should be faster, because (ideally) the number of free blocks for the 2nd drive should be greater than the first. Maybe we can't answer that right now because there is no firmware with TRIM out for Intel, but it's something to think about.
  • jimhsu - Saturday, August 22, 2009 - link

    To be clear: The drive is obviously struggling, because disk queue lengths are high (5-10). This was tested with the Intel Matrix Controller driver.

    PS Also about the Matrix Controller driver - for some reason on ICH9R, Intel disables NCQ functionality for AHCI on Windows 7. I have not totally confirmed this, but AS SSD 64 thread /4k random read results without Matrix Controller are 3 TIMES higher than with the Matrix Controller.
  • acolona - Friday, August 21, 2009 - link

    I have been waiting on your article to make my decision...any idea on an ETA? seeing all this SSD goodness is killing!

    I have basically been waiting to see if you have an opinion on an intel G2 vs OCZ vertex vs OCZ vertex turbo.

    I mainly use my machine for gaming, and audio/video duty...ripping dvd's, a LOT of video transcoding, and then multitasking for work.

    If money didn't matter between these 3 drives, and you had to buy tomorrow, what would you suggest? What is the main tradeoff between an intel and a indilinx drive?
  • AbRASiON - Wednesday, August 26, 2009 - link

    I agree with this ETA request, we're dying here with credit cards waiting.
  • semo - Saturday, August 22, 2009 - link

    Intel's main advantage right now is that even without TRIM it performs very well when worn compared to the others. Also, nothing can beat it in random writes at 4kb and it joins other SSDs at saturating SATA2's read bandwidth
  • Anand Lal Shimpi - Saturday, August 22, 2009 - link

    Intel vs. Indilinx really boils down to this: random write performance (Intel) vs. sequential write performance (Indilinx).

    In most single application benchmarks, there's no measurable difference between the two. There are a few cases where the sequential write advantages of Indilinx help a lot, but they seem to be rare (outside of pure sequential writes to the drive).

    Ultimately Intel's drives offer lower latency for a random write. You're looking at 0.3 - 0.5ms for Intel vs ~1ms worst case scenario for Indilinx. Both are an order of magnitude faster than a hard drive. Theoretically the Intel drive should "feel" faster in normal use, but that's a tough one to quantify.

    Indilinx had gotten so good that I even swapped out my personal X25-M for an OCZ Vertex just to make sure I wasn't recommending something I wasn't comfortable with. The experience has been wonderful; it feels slightly slower but it's difficult to tell if that's real or not. The benchmarks indicate it is but I'm not sure if my experience testing is what is making me hypersensitive to performance or not.

    You can't go wrong with either drive. My pick continues to be Intel but the pricing works out in a nice way; simply pick the capacity you want and you'll find the right solution.

    If all you need is 60GB, go Indilinx. If you need 80GB, go Intel. 120GB? Indilinx. 160GB? Intel. If price doesn't matter? Intel.

    Just stay away from the Samsung based drives.

    Take care,
    Anand
  • techvslife - Tuesday, August 25, 2009 - link

    Unfortunately, the new 1711 firmware for Indilinx barefoot controller drives is causing massive data corruption on many Windows 7 systems (now discussed on ocz vertex, supertalent, & crucial support forums). The cause currently is thought to be the interaction of TRIM with ACPI on some motherboards.
  • erikejw - Wednesday, August 26, 2009 - link

    it seems bad but it is a beta so noone that was scared of his data should have installed it

    they will solve it before any proper release
  • techvslife - Wednesday, August 26, 2009 - link

    Unfortunately, 1711 was NOT beta. It was beta only at some makers who caught the bugs only a day before release. Some SSD makers had already released to the general public as final.
  • erikejw - Tuesday, August 25, 2009 - link

    Any news on when Vertex(or any other Indilinx) will go 32nm,34nm?
    At that point I am sure we will see decent price reductions.

    I'd like to also see used drives in all benches since that is the performance that 95%+ of users will have and hence is of most importance.

Log in

Don't have an account? Sign up now