Vienna Symphonic Library Forum
Forum Statistics

181,613 users have contributed to 42,178 threads and 254,543 posts.

In the past 24 hours, we have 4 new thread(s), 12 new post(s) and 40 new user(s).

  • Internal Hardrives - Seagate Barracude vs Western Digital

    Hi, I am currently looking into getting more internal drives for a new 8 core Mac Pro. I am in between the seagate barracuda 7200.11 750GB and the Western Digital Caviar Black series, specs seem pretty siimilar. Has anybody tried either of them and would like to share experience, good or bad with them? My plan is to get 2 and RAID 0 them for streaming. Thanks, Thor

  • rather than the 7200.11 i'd get the ES 2 series from seagate if the intended use is sample streaming,

    can't comment on the WD Caviar Black, but the description allows to assume they are also intended for professional use.

    christian


    and remember: only a CRAY can run an endless loop in just three seconds.
  • That's interesting, can you explain why you'd prefer the ES.2 over a 1.0 or 1.5 TB 7200.11 drive?  From all of the perf specs I've seen, the 7200.11 actually does better than the ES.2 in most cases (for example, http://techreport.com/articles.x/15730/1).  It seems like the vibration sensors on the ES.2 are the main difference between the two.  I can see opting for the ES.2 drives if they were being densely stacked in a server rack or something, but I haven't seen a reason from the stuff I've read so far to pick them over the consumer models for streaming samples.  I'm sure I could be missing something, though.

    - Chris


  • Christian,

    thanks for your help. I have heard that the seagates get quite hot, especially the ES 2 series. I am planning on having 4 drives in my mac pro so heat could be an issue. Could be hear say, maybe its not actually true. Have you used those drives personally?

    Also, size wise - for streaming is it better to get a smaller drive alla 3* 500 rather then 3* 1TB ?

    Thanks!


  • the ES / ES.2 series are intended to be used in servers (permanent access) whereas the 7200.11 is for desktop (higher average idle time), and sample streaming is very similar to the former.

     

    from the specs it seems the black caviar is using less power (8.4 W) than the ES (10.42 - 11.16 W) and therefore would get less hot, but WD doesn't specify the percentage of operating/seeking/idle time ...

     

    the airflow in a macPro is excellent and we did not notice with the 1 TB ES.2 in a macPro getting unusually hot

     

    data transfer rate is specified very similar (106 / 105 MB/s) what i noticed to result in *real life* in 78 MB/s for an 1 TB ES.2 what equals to 460+ sample streams ...

    allow 50% overhead for seeking, allow average 2 streams per instrument (eg. A/B cell) results in 115 tracks streaming from a single disk.

     

    remember: leave ~25% free space on each disk as the system requirements indicate what means in detail eg. a super package should be spread across 2 disks and of course one would spread the most intensively used collectons carefully (say strings on one, brass & woodwinds on the other)

    christian


    and remember: only a CRAY can run an endless loop in just three seconds.
  • Thanks Christian. Now here is the next question then...would it be better to spread the library out between 2 drives or RAID0 the drives. I haven't tried RAID0 myself but always wonder if it would increase voice count over seperate drives. I remember a test in sound on sound showed that RAID o didn't help as much as one would think for audio. Also, for anybody that is interested in the HD subject, check this link out... http://www.tomshardware.com/reviews/hdd-terabyte-1tb,2077.html

  • raid 0 (striping) has pros and cons ... for storing sample content i would not do so (neither in hardware nor in software) ... why:

    - loosing one disk means loosing the content of both (twice the time needed to restore everything)

    - reading sample content is another type of file acces than say video streaming ... we have to read tiny portions (~ 16 KB) out of large files (~ 512 MB) what means every type of caching is contra productive (well, we can't avoid the caching of the operating system, but we can avoid the one from the disk and some raid controller) ... caches tend to assume the next portion of a file would be needed next which is not true with sample streaming for the overwhelming number of read-cycles.

    - with raid 0 the system has always to wait until both drives are in the right position to read the data which is spread across them

    - again for little data portions (here 16 KB) this makes no sense at all since raid 0 cluster size is 64 KB in most cases.

     

    tom's hardware guide is always a great resource of facts .... juts keep always in mind we do not have some common desktop application, but an almost realtime server-like behaviour, very similar to extensive database access

    christian


    and remember: only a CRAY can run an endless loop in just three seconds.
  • last edited
    last edited

    @cm said:

    type of caching is contra productive
     

    So does this also count for the drive cache so 16mb is actually better than 32mb?

    This is really interesting, I don't think anybody has ever answered this better than you have christian. Thanks!

    Too bad the raptors come in such small sizes...


  • Christian, I went with the ES 2 1 TB drives. I did an inital benchmark test using EXS and only got about 400 voices out of it. Then I realized I hadn't removed the Tab that limits it to 1.5GB/sec, after I removed it I didn't notice much difference. Any ideas of what could cause the bottleneck. Some people report 1000 voices out of a drive like that. What do you think?

    I run logic 8 , 128 Buffer, EXS to slow and extensive setting...


  •  hmmm, voices (in EXS) is not always identical to sample streams (from harddisk) ... the 400 (stereo-voices?) you mention are fitting pretty well to my estimated 460 from above ....

    i would try to analyze a precise definition of the reported 1000 voices (maybe mono?) and all the lefts and rights of those reports ...

    setting of the harddisk interface to 1.5 Gbit/sec is definiely not the bottleneck ...

    41.000 Hz x 48 bit = 1.97 Mbit/sec ... 1.5 Gbit/sec / 1.97 Mbit/sec = 838.000 streams .... clearly miles away from anything the harddisk can physically deliver ... you wouldn't get even close to 1.5 Gbit/sec with the most expensive flash disks

    christian (calculation corrected)


    and remember: only a CRAY can run an endless loop in just three seconds.
  • last edited
    last edited

    @Another User said:

    41.000 Hz x 48 bit = 1.97 Mbit/sec ... 1.5 Gbit/sec / 1.97 Mbit/sec = 838.000 streams .... clearly miles away from anything the harddisk can physically deliver ... you wouldn't get even close to 1.5 Gbit/sec with the most expensive flash disks

    hm, yeah. Thanks Christian