Would an additional hard drive improve performance on my machine?

June 26, 2004 11:34 PM

Would installing a new (smallish) hard drive in my machine and dedicating it solely to the Windows swap file (i.e., instead of having it on C:) improve performance?
posted by oissubke to computers & internet (13 answers total)

Move the pagefile off the disk that holds your system and boot partitions to another fast and dedicated hard disk. If you do put the file elsewhere, you should leave a small amount on C: - an initial size of 2MB with a Maximum of 50 is suitable - so it can be used in emergency. Without this, the system is inclined to ignore the settings and either have no page file at all (and complain) or make a very large one indeed on the C: drive. (1)

Paging and Registry file fragmentation can be one of the leading causes of performance degradation related to file fragmentation in a system. (2)

Both of the pages cited as source materials are worth a full read, but the short answer to your question is yes.
posted by sequential at 12:01 AM on June 27, 2004

... as long as it's a fast(ish) hard drive.
posted by Space Coyote at 1:12 AM on June 27, 2004

what's fast-ish these days?
posted by scarabic at 1:15 AM on June 27, 2004

Well, a 8MB cache 7200RPM drive should suit you well, if you don't want to blow cash on a really fast model. The 8mb cache really helps because of the quick "burst" speed. Prolonged data pulling makes the 8mb drive as fast as the 2mb drive. If you don't want to blow money, just get one of these.

Notice the size difference. If you have SATA, think about getting the raptor if you're interested in faster windows performance.
posted by Keyser Soze at 2:03 AM on June 27, 2004

Yes, it probably would, but I assume you're doing this because your RAM is already maxed out? If not - add as much RAM as you can.
posted by danhon at 2:05 AM on June 27, 2004

You'd be better off getting a similarly-sized hard drive and RAIDing them, if you have the ability.
posted by Civil_Disobedient at 4:15 AM on June 27, 2004

I am about to go to bed, but does anyone remember that website that showed you how to edit the registry to put the windows kernel in the RAM? That would definately make a difference.
posted by Keyser Soze at 4:46 AM on June 27, 2004

Fuck page files. Get more RAM.

I am about to go to bed, but does anyone remember that website that showed you how to edit the registry to put the windows kernel in the RAM? That would definately make a difference.

No, but I remember TweakXP could do it.
posted by angry modem at 8:17 AM on June 27, 2004

Even with a gig of RAM, moving my page file to another drive helped performance.
posted by frenetic at 8:56 AM on June 27, 2004

ArsTechnica has some Win2K tweaks for improving virtual memory performance.
posted by kindall at 9:03 AM on June 27, 2004

I'm telling you. RAID is the way to go.
posted by Civil_Disobedient at 12:01 PM on June 27, 2004

Civil_Disobedient: I'm really not sure RAID for performance is a great recommendation nowadays.

From Storage Review:
3. RAID helps multi-user applications far more than it does single-user scenarios. The enthusiasm of the power user community combined with the marketing apparatus of firms catering to such crowds has led to an extraordinarily erroneous belief that striping data across two or more drives yields significant performance benefits for the majority of non-server uses. This could not be farther from the truth! Non-server use, even in heavy multitasking situations, generates lower-depth, highly-localized access patterns where read-ahead and write-back strategies dominate. Theory has told those willing to listen that striping does not yield significant performance benefits. Some time ago, a controlled, empirical test backed what theory suggested. Doubts still lingered- irrationally, many believed that results would somehow be different if the array was based off of an SATA or SCSI interface. As shown above, the results are the same. Save your time, money and data- leave RAID for the servers!

Much better off directing your money to a SATA controller/mb and a good SATA drive with a cache.
posted by malphigian at 12:46 PM on June 27, 2004

Short answer: Yes.

Long answer: Yes. No matter how much memory you have, the kernel will dip into swap and with good reason. For most workloads -- those with moderate I/O and a preference for latency over throughput -- it's better to swap out unused pages of memory and use the freed core for disk buffering. Given that, your goal is twofold: reduce the latency of paging out, and reduce the latency of normal disk activity.

You do this by putting swap on a separate spindle from all other disk activity, ideally on a bus with low to no congestion. This means that normal, everyday I/O -- the work you want your system to do -- isn't impeded by the latency of seeking the disk back and forth between swap storage and filesystem storage.

Modern disk drives and busses can do tricks like having the drive disconnect from the bus while it's moving the head around, essentially saying "this is going to take a while, I'll come back when I'm done." You can take advantage of this with swap on a separate spindle.

Disk seeks are really, really slow. The fewer of them required to complete an operation, the faster that operation can be done. This is where the value of filesystem defragmentation comes from, incidentally. If you have to seek 125 times to do some swapping, and 300 times to get some work done such as launching an application, you can either wait on one drive to do 425 seeks or wait on one drive to do 300 while the other drive does 125 at the same time. At an average of 10 milliseconds each, removing these little delays really improves the responsiveness of a system, which you'll "feel" as performance.

So put swap on a disk all by itself or at least on a very infrequently used disk if at all possible.
posted by majick at 5:02 PM on June 27, 2004