Sign in with
Sign up | Sign in

Benchmark Results: Throughput And Streaming

Three Generations Compared: Why Storage Density Matters
By

Performance increases for each generation have been almost linear (2007, 2008, 2009). However, even the newest Deskstar 7K1000.C doesn’t take a top position in the current hard drive arena. Samsung’s Spinpoint F3 delivers between 136 and 150 MB/s peak throughput, depending on the particular model. Seagate’s Barracuda XT reaches almost 140 MB/s, and the WD Caviar Black or RE4 drives at 2TB are a bit above 140 MB/s.

Read and write throughput doesn’t differ much.

The streaming test confirms the performance findings on h2benchw.

Display all 6 comments.
This thread is closed for comments
  • -9 Hide
    joex444 , February 27, 2010 5:24 AM
    So fewer platters = less energy and higher transfer rates and higher latency.

    One could guess that with elementary physics. In fact, it is obvious this should be the case. Way to earn your money, asshat.
  • -3 Hide
    LaloFG , February 27, 2010 7:51 PM
    Nice article, I'm tempting to buy a pair of this hard disk.

    It's true that high density equals high failure ratio? It is too risky use these disk (with high density) to hold data compared with old disk?

    (I have two 500GB HDs from ~2006 and still they are working flawless right now)
  • -2 Hide
    davidhbrown , February 28, 2010 11:23 AM
    You write that fewer platters yield better performance. I can certainly see why that would lower cost. But given that each surface has its own r/w head, shouldn't drive manufacturers be able to use the platters in parallel, sort of like RAID 0, to boost performance? Is this already done to some degree (maybe the two surfaces work in parallel), or is there some reason beyond controller cost and lack of market demand preventing this.
  • -1 Hide
    candide08 , February 28, 2010 12:40 PM
    Perhaps it should have been written that "for the same capacity (GB) fewer platters yield better performance." This would be because the data density (bits per sector/track) would be higher. "Smaller" bits take less time to read - with newer generation disk heads.

    There are also huge differences in capacity between generations of disk RW heads and newer heads tend to used with disks that have fewer platters, at least initially.

    Some RAID controllers do exactly as you say. The theory is that if the data is stored on 'cylinders' (same sector different platter) the disk arm will need to move less to transfer the data, saving time.

    This is classic main-frame disk array architecture, now easily available on high-end PC disks, some SAN's and some raid controllers. The more disk heads the faster the performance with these arrays.

  • 0 Hide
    wuzy , March 2, 2010 3:41 PM
    7200rpm 1TB drives are best suited for nearline storage or worktation usage if put into RAID array, where performance at shallow depth queue is more essential and is not heavily loaded 24/7. So the conclusion of this article makes sense, it's about $/GB while still maintaining 7200rpm needed for decent IOps (where 5400rpm is suited for backup arrays) and in this case highest possible density platter is needed.

    Article provides good data for historical references.
  • 0 Hide
    anamaniac , March 3, 2010 10:28 AM
    wuzyArticle provides good data for historical references.

    It does. =)
    Now, how about another comparison when they release a single platter 1TB 7200RPM 3.5" HDD?
    I'm going to stick with my 1TB Samsung and 500GB 7200x10's for now, but Hitachi, you certainly do have me considering you for my next big purchase.