Previous 1 2 3 Next 33 Replies Latest reply: Jun 6, 2013 5:06 PM by Hamper
Hamper Level 1 Level 1 (0 points)

I am looking into getting an SSD and am constantly reminded that now warranties are typically capped on data allowance averages e.g. 40GB per day. With the new SSD I am to get a whole new computer and that means access to advanced interfaces that I've not had before. I have a number of HDDs with FW800 and eSATA (okay, a Mac Mini doesn't have that) which I plan to transfer the data from to new HDDs via the likes of Thunderbolt.

 

My concern is that exchanging one HDD of 500GB to another will impact on the SSD in the main computer? I have been told that under Windows this does not occur, that the only data affected are the external HDDs themselves.


MacBook (13-inch Late 2007), Mac OS X (10.7.3), 4GB 667 MHz DDR2 SDRAM
  • William Lloyd Level 7 Level 7 (20,905 points)

    I wouldn't worry about it.  Reading all the data to copy it over should be fine.  SSDs do wear-leveling and other things so their lifespans are in the "many years" range.

  • Hamper Level 1 Level 1 (0 points)

    Well I tend to keep my computers for the five year range and the warranty on the M500 SSD is only three years, this is a new concept for me, I obviously like most, coming from the HDD world. I remember there being a software upgrade for MacBooks in OS X which would let the user know when their power chord had to be replaced, does this extend to SSDs?

     

    I also wanted to know the actual answer as to whether the data touches the SSD as it would have a great impact because I have many hundreds of GBs to transfer and sometimes something can go wrong and I would have to start over again.

  • FatMac>MacPro Level 4 Level 4 (3,390 points)

    Hamper wrote:


    ...I also wanted to know the actual answer as to whether the data touches the SSD as it would have a great impact because I have many hundreds of GBs to transfer and sometimes something can go wrong and I would have to start over again.

    I'm no expert so take the following in that light: you may be able to test your concern with your current setup. If your Mac is in the middle of the transfer chain rather than the recipient, I'd expect that the data being transferred would pass through in RAM without having to be written. But if it was, it'd likely be going to your Mac's virtual memory. In Activity Monitor, the System Memory button shows the VM size and Page ins and Page outs. Page outs are what's being written temporarily to your internal storage, so if it's zero, the data is really just passing through in RAM. At the same time, the Mac OS "gulp" size for reading and writing may not adjust for the available RAM, and since you've only got 4 GB currently, the test may not be valid for your new Mac. But if it stays zero anyway, then more RAM would likely speed things up but the SSD shouldn't be touched.

  • Hamper Level 1 Level 1 (0 points)

    Hmm, interesting theory. I cannot try it as my MacBook screen is kaput and the HDD is damaged, plus all the storage I have is maxed out, I really am on the cusp. I am actially using Window (Vista) right now on a machine older than my MacBook, something like six to seven years old (Chrome works fine!).

     

    However I think I'll contact some of the guys who perform hardware tests, this sounds like a question they could answer and I'll get back to you.

     

    Thanks!

  • FatMac>MacPro Level 4 Level 4 (3,390 points)

    Hamper wrote:

     

    ...However I think I'll contact some of the guys who perform hardware tests, this sounds like a question they could answer and I'll get back to you...

    Or I'll get back to you. Figured I could test your problem and my theory to see what happens.

     

    Source - Toshiba 3TB external. Destination - Kingston SSD in an OWC external enclosure. Both attached via USB 3.0 to rMBP 15" 2013 with 16GB RAM. Transferred 25 files totaling 990 MB while watching Activity Monitor's page outs. There were none. And the transfer went really fast. Tried that again, this time with a 6+GB Debian Mint VM and, again, no page outs. Took about a minute.

     

    One thing you can consider doing to extend the life of the SSD is to partition the drive with a partition smaller than the rated capacity; drive manufacturers already do that in what's called "over-provisioning" and that unused space is sometimes what gets written to in the garbage collection (GC) process to support wear-leveling.

     

    SSD's are a brave new world for old guys.

  • Hamper Level 1 Level 1 (0 points)

    Ah thanks, great speeds you're getting there, Retina too, I'm living in the stone age here, big time.

     

    I was holding out for Thunderbolt but the peripherals are just too much, I mean, a 4TB HDD is cheap right now but if I want it in an enclosure allowing TB, forget it, tripple the price! USB 3 is on the Mac Mini though, so I can do FW800 to that I suppose, but SSD is what I'm really looking forward to but I didn't want to have to outlive it any sooner than I had to.

     

    I put myself about a bit to find the information you provided out, so I'll be able to feed those results back here, but good to read the positivie results so far, I half expected/hope they would reflect this.

     

    I am interested in this partitioning you are talking about because I know OS X doesn't do 3rd party TRIM support and am also investigating 3rd party apps to enable it or whether newer SSd's, like the M500 from Crucial I am interested in, work well enough on their own?

     

    I am also going to have to do some tricks with the system as there arw two drive in a Mac Mini and I wanted to keep them separate: an SSD for main use, while the other is used for backing up via Time Machine, all in the box, as I don't have access to a Time Capsule. So this mini partition you are speaking of, I wasn't aware of this precaution, would that entail a partition of say 980GB as the SSD is 990GB or do I have to shave more off?

     

    Thanks again.

  • FatMac>MacPro Level 4 Level 4 (3,390 points)

    I mentioned the "do-it-yourself" over-provisioning because it's my understanding that the enterprise level SSD's build that in to a greater degree than consumer SSD's because price is less of a concern. TRIM support for 3rd party SSD's can be turned on using this; I've never tried it because my OWC SSD internal uses a RAID configuration and TRIM doesn't work with RAID.

     

    Thunderbolt connections are going to offer greater speed than anything else but the bottleneck will be the drive you're using, not the connection when it's Thunderbolt, USB 3.0 or SATA III. FWIW, if I'm backing up to a Toshiba 3TB USB 3.0 external from an SSD, I can get large-file transfer speeds of over 130 MB/sec when it's writing to the first partition on the drive (and I think the drive is a "green" drive since it runs cool, though I've never taken one apart). If it's to the inner partitions and/or the files are small, there's a noticeable slowdown.

     

    The Crucial M500 does seem to be a good choice. And if you look at the "Performance Consistency" page in that review, you'll see mention of 25% spare area; that's the "do-it-yourself" over-provisioning.

  • Hamper Level 1 Level 1 (0 points)

    Thanks, I would have otherwise have missed that valuable snippet of info, but do I have to give 25% of the HDD away, I am alreay at 500GB and this SSD is only 960, at 768GB I might be struggling some.

     

    I'm still waiting on Crucial to get back to me on TRIM, but thanks for the link.

     

    I am eagre to see if the "laboratory conditions" all those years ago when USB2 came out can be replicated when plugging it into a USB3 port to see just how fast it can go, I should achieve the speeds they always promissed, right

  • FatMac>MacPro Level 4 Level 4 (3,390 points)

    Hamper wrote:

     

    ...do I have to give 25% of the HDD away, I am alreay at 500GB and this SSD is only 960, at 768GB I might be struggling some...

     

    ...I am eagre to see if the "laboratory conditions" all those years ago when USB2 came out can be replicated when plugging it into a USB3 port to see just how fast it can go, I should achieve the speeds they always promissed, right

    You don't actually have to give any of it away since the drive does come with some over-provisioning built in. But that clever adjustable table in the review shows that the extra spare room improves performance under heavy use so it's a trade-off. Also, an SSD is not immune from bogging down when it gets nearly full.

     

    I'm not sure what you mean by "can be replicated when plugging it into a USB3 port to see just how fast it can go." Obviously, a USB 2.0 device can't go any faster than the USB 2.0 spec even if it's plugged into a USB 3.0 port, but I don't think that's what you meant. I just tried copying a virtual machine from the internal SSD to Kingston via USB 3.0 and it peaked at about 260 MB/s onto an almost full SSD. It wasn't continually that fast but I had to try it twice because the first, smaller VM went so fast I almost missed the transfer rate. The USB 3.0 spec calls for almost twice that rate but that's the connection's capability, not the devices' speeds. And in this case, the Kingston had been heavily used so GC would likely be an issue, not just raw transfer rate.

     

    Nothing is ever simple; you heard it here first.

  • Hamper Level 1 Level 1 (0 points)

    What kind of allowance then do you think I should give, 50GB, 100GB?

     

    I meant that whenever a new standard comes out there are the real world rates and the laboratory conditions, for example with 802.11ac. However I would like to see if the laboratory conditions or theoretical speeds could be achieved when all hurdles are removed, which should come about with USB 2.0 in USB3.

     

    Sources reveal to me that SSDs should not be used for virtual memory, that it should be disabled to begin with, this was in relation to the test you took. But I think that virtual memory was on the RAM itself, not the SSD, right? Was I wrong to think it was a 40GB allowance for the whole SSD itself (per day), rather than per block?

  • Hamper Level 1 Level 1 (0 points)

    "Cerb" here came up with a different result to your own:

     

    "Page out refers only to swap. They are only indicative of going to the swap files if you're low on RAM (unlike Windows, OS X treats swap as emergency extra space only). They ill only occur if you're running low on RAM.

     

    Page ins don't matter (I'm pretty sure they're just soft faults, but I'm no OSX wiz).

     

    You can't use either to determine what you're asking.

     

    As to what you are asking, the SSD may get written to when copying those files, but the writes to the SSD when doing so will be negligible. They will be log additions and maybe lock files and such, or the occasional search index. The amount of writes from such activities is noise."

     

    ref: http://forums.anandtech.com/showthread.php?p=35087050&posted=1#post35087050

  • Hamper Level 1 Level 1 (0 points)

    I recieved a reply from one of the reviewers of the M500 and he said:

     

    "If you're transferring from, say, an external drive on FireWire to another drive on Thunderbolt, then no, the data won't be hitting the Mac's internal drive at all.

     

    Even with the default RAM configuration of 4GB or 8GB in modern Macs, there should be little significant page outs to the internal drive anyway with normal OS X usage, and any virtual disk space taken will be used more for applications' caching rather than the data stream from one drive to another."

     

     

    So it seems concrete that the SSD would be sparred.

  • Csound1 Level 8 Level 8 (41,280 points)

    Hamper wrote:

     


    "Cerb" here came up with a different result to your own:

     

    "Page out refers only to swap. They are only indicative of going to the swap files if you're low on RAM (unlike Windows, OS X treats swap as emergency extra space only). They ill only occur if you're running low on RAM.

     

    Page ins don't matter (I'm pretty sure they're just soft faults, but I'm no OSX wiz).

     

    You can't use either to determine what you're asking.

    When you open a file or start an app you take it from the HDD to Ram, that's a Page In.

     

    When there is no more available Ram and the system needs to make room for a new request it will swap from Ram back to disk, that's a Page Out.

     

    Lots of Page Outs means more Ram is desirable,

  • FatMac>MacPro Level 4 Level 4 (3,390 points)

    Hamper wrote:


    ...However I would like to see if the laboratory conditions or theoretical speeds could be achieved when all hurdles are removed, which should come about with USB 2.0 in USB3.

     

    Sources reveal to me that SSDs should not be used for virtual memory, that it should be disabled to begin with, this was in relation to the test you took. But I think that virtual memory was on the RAM itself, not the SSD, right?...

    As I understand it, USB 3.0 is entirely separate from USB 2.0 in the sense that physically separate channels are used to carry USB 3.0 traffic. So it's not as if anything extra has been given to USB 2.0 in the process. Compatibility of devices and connections is maintained by having both sets of channels built into the USB 3.0 cables and connections, which is how Apple can provide one type of USB socket that can handle both types of connections (PC's I'm familiar with have separate 3.0 and 2.0 USB sockets [though I suspect 2.0 devices will work in the 3.0 sockets], and avoids the issues I've read about in the Communities where Apple's sockets would run at 2.0 speeds with 3.0 devices if 2.0 devices were plugged in first - I haven't run into that myself, which would suggest that the newer versions of the rMBP got more than just a speed bump).

     

    Memory is normally RAM, but if enough isn't installed for what's being done, some of what's in RAM will be temporarily written to memory which is "virtual" (i.e., to disk) to make room in the "real" memory for more current stuff, as has already been mentioned. My thought of using Disk Utility's Page outs is based on the idea that any transfer of data from one external to another will be passing through RAM without making any permanent copy on the internal drive (programming which would require an additional write-then-read step in the middle would be terribly inefficient); however, if the amount of data read in each read/write cycle exceeded the real RAM available, some would temporarily go to virtual memory (i.e., to disk) as a Page out. The response from one of the M500 reviewers suggests that the OS is smart enough to tailor the data gulp taken in each read/write cycle to the available RAM and avoid hitting the drive. That would also tend to disagree with Cerb's comment that Page outs can't be used to determine what you're asking.

     

    As Csound1 says, the solution to heavy use of virtual memory is more RAM. I don't know if it's possible to disable virtual memory usage if you're using an SSD and it doesn't seem advisable anyway. Virtual memory is the OS solution to system demands that exceed the available resources (RAM) and if VM usage was denied, what else could the computer do?

Previous 1 2 3 Next