Cooling Some Jets

Add Your Comments

There’s an initial review of the i-RAM over at Anandtech.

No doubt the results will be very disappointing to those with unreal expectations (actually, they’re somewhat worse than I expected.)

The price tag is also a bit disappointing, the initial models will cost not $50 or $80, but $150.

Future Fixes?

There are some reasons for iRAM’s relatively disappointing results:

  • Throughput is bottlenecked by the SATA 1 bus speed.
  • Software/hardware is not properly optimized for high throughput
  • Some software is so well optimized to use hard drive and Windows caches that they’re mostly being read from memory, anyway.
  • For some operations, it just takes the CPU a good deal of time to get the work done. It’s not just a matter of file transfer.

    It’s important to understand that it’s not just one item that’s the problem, and that some items just aren’t very fixable, no matter what you do to, say, throughput speed.

    The first item is the only one that could be “fixed” to some degree fairly quickly. A SATA 2 interface would double the throughput, but this still would be a theoretical bottleneck (only a multi-channel PCI-E slot would get rid of most of that), and even that wouldn’t help much if factors two through four heavily influence a situation.

    Per optimizing software/hardware, you have the problem of doing a lot of work to benefit few people, so that won’t be on too many front burners. I suspect it’s more likely that a hack will come before any “official” release for major programs.

    Items three and four are inherently not very fixable.

    The Real Test

    While the limited number of tests run with the i-RAM are certainly good enough to tell the average person whether or not the card is worth it for them, what needs to be done is to figure out to what degree future versions of this technology will offer improvement, and what just isn’t improveable.

    For instance, for other reasons, you might not want to use a software-controlled RAMDisk in real life, but that would give a better (though not perfect) idea of how much SATA represents a bottleneck. One could also test large motherboard configurations with big Windows caches to see what improvement one might get there.

    So I’ll probably get an i-RAM sooner or later, but the testing I’ll do will be mostly aimed at determining that. It won’t be “news you can use” right away, but in the long run, that ought to give us a better idea of what can be reasonably expected from this.

    Ed

  • Leave a Reply

    Your email address will not be published. Required fields are marked *