I was recently contemplating the purchase of a new workstation, since my current system is about two years old. After some research, I ended up choosing NOT to buy a new system just yet, for a few reasons. Read on for an explanation of why, and how a small upgrade kicked up my system performance by two-thirds.

The main reason for the purchase delay has to do with SATA-III. While this new disk connection standard supports up to 6 Gb/sec transfer, and SATA-II is 3 Gb/sec, the reality is that right now the 6 Gb/sec speed is unusable in real life. No traditional hard drive comes close to saturating the SATA-II bandwidth, and they will NEVER get into SATA-III transfer speed due to the physical limitation of the spinning platters. And even the new solid state drives (SSDs) barely fill up the 3 Gb/sec pipe, and come nowhere close to 6 Gb/sec. So spending a bunch of money on a SATA-III motherboard and SATA-III drives is pointless right now.

Over the next 6 months or so, this will change. New SSDs will come out that have faster transfer, and more space, for decreasing price. But until an SSD comes out that actually uses the new bandwidth limit, SATA-III is little more than a marketing gimmick. Once they finally get to the SATA-III speed then I will reconsider.

In the meantime, what I did instead was drop about $279 on a new 600 Gb Western Digital Velociraptor drive. These spin at 10,000 RPM, compared to the normal 7,200 RPM, and this is the fastest non-SSD hard drive on the market. They are roughly 66% faster than a normal drive, basically approaching standard drives in RAID-0. Also, at 600 Gb in size, it can hold my existing system partition (which is about 400 Gb). It came with a program called Arconis TrueImage, so I used it to clone my existing 1 Tb system drive onto the 600 Gb Velociraptor, then pull out the old system drive and put the new drive in its place. Windows 7 boots from the new drive without knowing the difference. The cloning process took about an hour and worked perfectly. So I'm now running off of the 600 Gb drive.

Speed-wise, it's a big difference. I did a few real-world tests:

  • I timed the system boot time from POST until the desktop was fully loaded. By fully loaded I don't just mean seeing the desktop, but having ALL desktop widgets loaded, ALL system tray icons loaded, and until the CPU activity drops back to idle usage. It went from 180 seconds before to 110 seconds after. Result: 63% improvement.
  • I timed launching and building large Eclipse workspace. It went from 56 seconds before to 35 seconds after, a 60% improvement.

So basically, for $279 I've boosted my system speed by nearly two-thirds. That's pretty crazy, considering that swapping the CPU for a faster i7 or going from 1333 to 1600 RAM would probably only generate a 5% or 10% increase, and would cost way more than $279.

The surprising lesson is apparently: do NOT underestimate the impact that a really fast hard drive will have! This should be a nice boost to carry me forward until the SSDs get faster, bigger and cheaper.

Comments Comments (5) | del.ico.us del.icio.us | Digg It! Digg It! | Linking Blogs Linking Blogs | 3733 Views

Comments

  • # Posted By Daniel D. | 12/6/10 1:28 PM

    Disk are still the lowest common denominator in speed for Computers right now. So make sense that buying a faster drive will improve things greatly. When access time and transfer rates are in general a factor of 10 or more less then the rest of the system easy to get significant improvement. Love to see some of the tech that is out there in the lab(s) right now for disks that work at ram (not flash ram, but normal system memory) speeds and is stateful. When these get into production and priced into an afordable range (for servers at least). This will change the way we work with data. You can see some of this with RAM based Databases (fast fast and did I memtion fast).

    I am about to upgrade the disk in my laptop (need more space) and also hope to get more performance at the same time strech another 2 years or so out of it.

  • # Posted By jason olmsted | 12/6/10 4:24 PM

    For notebook performance, you might look at the Momentus XT - a hybrid of spinning disks and ssd. In some tests it even outperforms the Velociraptor (and for a single drive system, it is likely the best tradeoff of cost, speed and space): http://www.storagereview.com/seagate_momentus_xt_r...

    Additionally, for the desktop, you might consider using an SSD as a scratch disk (for system swap file or even application swap; photoshop for instance) in addition to the raptor. In this case, smaller SSDs are getting cheaper and the relatively small volume isn't an issue.

  • # Posted By Peter Doyle | 12/7/10 1:16 AM

    Hi,

    I think your ideas and comments are sound. Surely your maths through the article are not though?

    A 70 second reduction over 180 seconds is around a 38% reduction not 66%?

    Sorry to pick holes but it does make a fair difference to the basis of the post, unless I have completely misread?

    Cheers
    Peter

  • # Posted By Brian Kotek | 12/7/10 3:19 AM

    Interesting. First, let me show what I did to get my numbers:

    180/110=1.63. And 56/35=1.6.

    Put another way, you'd have to multiply 110 by 1.63 to get to the original value of 180. And likewise, you'd have to multiply 35 by 1.6 to get to the original 56. Which is where I got 63% and 60%, respectively. So 180 is 63% more time than 110. (http://www.percent-change.com/index.php?y1=110&...)

    Or put yet another way, if the speed had doubled in both cases (a 100% improvement), the new values would have been 90 and 28.

    Which seems to make perfect sense to me. But after looking it up, switching the numbers gives the value you mentioned (http://www.percent-change.com/index.php?y1=180&...).

    So it looks like you may be right, but it still seems odd to me. Maybe I'm just tired and it will make more sense tomorrow heh. Or maybe not...math was never my forte. ;-)

  • # Posted By Jason Fisher | 12/7/10 7:44 AM

    Brian,

    I think you're both right: to make up for a 50% reduction, you have to have a 100% increase. It's just a matter of looking top-down (which Peter is doing) or bottom-up (which you're doing).

    same numbers, "different" percentages:
    100 > 50 = 50% reduction
    50 > 100 = 100% increase