ace101

Seek Times

Recommended Posts

Simple equations..

* Higher RPM = Lower Seek Times

* Higher Platter Density = Better STR

* Bigger Cache = Slightly Better Performance

Overall, we can't win them all.. Also the overall drive firmware coding also affects some performance characteristics. 10Krpm and 15Krpm drives rules with the best seek times. However, not nescessarily will win in overall performance category. A good combo is needed.. higher rpm, higher density platter and bigger cache would be "ideal". Not forgetting the hardware setup as well.. ;)

Higher RPM = Lower Seek Times

No, SEEK time remains the same, access time goes down.

Seek time=time from request until head is ready for read (unaffected by RPM)

Latency=time for disk area to spin under head for read (completely related to drive RPM)

Access time=seek time+latency.

Share this post


Link to post
Share on other sites

Don't forget the importance of firmware - properly optimised firmware can easily make up for different cache sizes, access times and STR - though these basic mechanical factors do still have an influence.

Also, while faster seek times and higher RPM tend to go together, that's just because it makes sense to combine a higher spin speed with a faster actuator, since you're already shifting the balance away from price and towards performance. They don't directly affect one another.

For a long while, the Raptor was faster than the 15K SCSI drives available at the time when it came to desktop tasks. The reason? The firmware on those 15K SCSI drives was optimised for server workloads, while the Raptor's firmware was better tuned for desktop tasks.

Share this post


Link to post
Share on other sites

I really tend to believe that faster spinning hdds still hold a grasp on the hdd market for high-end. If i had to choose, between 7.2k models. Speed really wouldn't be that big of an issue, cuz a points of a millisecond don't make that much. But when one speaks of hearly halfing the time needed to get to my data, WOW; now there's an improvement!!!...

SCSA

Share this post


Link to post
Share on other sites

Spin rate and seek times are meaningful only when the data requested is NOT in the HDDs cache. Caching algorithms that focus on increasing performance with desktop workloads are not present in server-optimized drives (like 15krpm drives).

15krpm drives with about half the latency of 7200rpm drives and faster seeking ability does not match the performance of a desktop drive that is highly optimized for desktop environment (such as 7K500). That is simply because regardles how fast the 15krpm seeks, 7200rpm drive doesn't necessarily have to seek at all! Bursting data out of cache at the speed limited only by the interface and without any rotational latency always beats any mechanical seeking. Not even a 20krpm or a fictional 30krpm drive could beat the speed of transistor-based RAM.

scsa, if spin rate is really that meaningful, does that mean a solid state drive would have inferior performance to any mechanical drive as the spin rate of SSD is 0rpm?

Share this post


Link to post
Share on other sites
Spin rate and seek times are meaningful only when the data requested is NOT in the HDDs cache. Caching algorithms that focus on increasing performance with desktop workloads are not present in server-optimized drives (like 15krpm drives).

15krpm drives with about half the latency of 7200rpm drives and faster seeking ability does not match the performance of a desktop drive that is highly optimized for desktop environment (such as 7K500). That is simply because regardles how fast the 15krpm seeks, 7200rpm drive doesn't necessarily have to seek at all! Bursting data out of cache at the speed limited only by the interface and without any rotational latency always beats any mechanical seeking. Not even a 20krpm or a fictional 30krpm drive could beat the speed of transistor-based RAM.

scsa, if spin rate is really that meaningful, does that mean a solid state drive would have inferior performance to any mechanical drive as the spin rate of SSD is 0rpm?

Good points - but your probably wasting your time ;) SCSA has almost a thousand posts on this board since '03 and in his last point - states that doubling the rpm will cut the time required to get his data in half. I couldn't even guess how many times and in how many different ways this myth has been crushed and obliterated - some people will just hang on to their ideas regardless I guess.

Anyways, I'm off to play some Farcry on my 6 disk raid 0 made from original 9 gig 10K Atlas drives - (that's like 60K rpm - and that means it RulZ)

Share this post


Link to post
Share on other sites

"Feel the power"... but use proper hearing protection.

Like having a nuke under one's possession but not knowing how to arm it. Useless. But others alike will probably get jealous.

Share this post


Link to post
Share on other sites
I really tend to believe that faster spinning hdds still hold a grasp on the hdd market for high-end. If i had to choose, between 7.2k models. Speed really wouldn't be that big of an issue, cuz a points of a millisecond don't make that much. But when one speaks of hearly halfing the time needed to get to my data, WOW; now there's an improvement!!!...

SCSA

Decimals of a ms do matter when it comes to seek time, at least they matter more than is obvious. When the arm has to swing back and forth several hundred or thousand times to read files, 8ms VS 8.6ms would make a difference, but again, people probably still wouldn't notice it.

Share this post


Link to post
Share on other sites
Simple equations..

* Higher RPM = Lower Seek Times

* Higher Platter Density = Better STR

* Bigger Cache = Slightly Better Performance

Overall, we can't win them all.. Also the overall drive firmware coding also affects some performance characteristics. 10Krpm and 15Krpm drives rules with the best seek times. However, not nescessarily will win in overall performance category. A good combo is needed.. higher rpm, higher density platter and bigger cache would be "ideal". Not forgetting the hardware setup as well.. ;)

Then why does a high RPM, low density drive like WD Raptor outperforms most other drives?

Share this post


Link to post
Share on other sites

Most is key, it appears that the raptors speed makes up for the very few shortcomings that it has. The faster the drive, the further apart the data bits are on the platter itself. So the new Hitachi 500 gigger is able to beat the raptor in some things that require data to be closer together. Where as if the disk has to go out and seek data across the platter, the raptor wins hands down!

There is a link that compares the raptor to the new Hitachi! Notice that if you run a file server, the raptor wins hands down. But in some of the desktop comparisons, it doesn't do as well! I personally would pick (for my O/S drive) the 74 raptor over the new hitachi. The nice thing about the hitachi is that it is huge and somewhat fast!!!...

SCSA

Share this post


Link to post
Share on other sites

The way I interpret the main points in this thread is: high RPM reduces access time in worst case scenarios, provided all else is equal. The problem is that all else is clearly not equal, and a large disk holding the same amount of data as a faster small disk will more greatly enjoy the benefits of locality, since the heads need not seek as far. The SR DriveMarks effectively "short-stroke" larger drives with a small partition, though the more full a disk becomes, the more likely it is the seek distances will be long and even have to seek clear across the disk in such a "worst case scenario." Eugene pointed out that 25% of the time was spent on seeking past 8GB--on that small partition, so it may be much worse on a full 500GB partition. And I mentioned previously that some older disks have rather poor track-to-track seek times (despite having very fast average seek times) so are not able to take full advantage of the increased locality.

The analogy I'd like to mention is with games' minimum FPS compared with average FPS. A gamer is clearly far more concerned with the minimum frame rate because that is when the system is least responsive and they get killed. That is, even though the average frame rate may be very high, occasional choppiness is very noticeable, and it gives the perception of slowness, making improvements in minimum FPS advantageous even when at the expense of average FPS. While the IPEAK charts show that in the clear majority of the time, seek distances are short and average seek times are low, when a rare worst case scenario is encountered and many seeks across the disk are required at once, the disk will seem relatively unresponsive (eg less "snappy") even though these periods are not long enough to significantly impact the average time. Note the DriveMarks are measures of averages too, though the Server DriveMarks at lighter I/O depths may well describe these brief periods.

Cacheing at both the disk hardware level and the OS of course greatly impacts subsequent accesses to the same data and thus disk performance. But because most single-user accesses are so localized, the firmware algorithms may be optimized to read ahead a set amount on the disk, expecting it may be needed after a seek across the disk and making sure it will be available whilst seeking back. This way, even data that has not previously been used may be cached! The OS cache can be *huge*, but requires CPU cycles to run--which may explain why a fast disk seems to make a more noticeable difference in a slow computer than a fast one. The disk may thus "outrun" the OS cache in a heavily loaded slow computer (or the OS may recognize things are CPU-bound and preempt memory for itself, shutting off the cache) but is always slower than the cache in a fast computer. Actually the only time I really notice a 15k rpm disk is much faster than a modern 7200rpm disk on a fast computer is when first opening a folder with thousands of thumbnails. Some activities are simply not helped by cache.

Bottom line is it's not average seek time that matters at all. You have to pick which benchmarks you think most closely resemble how you'll use the disk.

Share this post


Link to post
Share on other sites
Most is key, it appears that the raptors speed makes up for the very few shortcomings that it has. The faster the drive, the further apart the data bits are on the platter itself. So the new Hitachi 500 gigger is able to beat the raptor in some things that require data to be closer together. Where as if the disk has to go out and seek data across the platter, the raptor wins hands down!

There is a link that compares the raptor to the new Hitachi! Notice that if you run a file server, the raptor wins hands down. But in some of the desktop comparisons, it doesn't do as well! I personally would pick (for my O/S drive) the 74 raptor over the new hitachi. The nice thing about the hitachi is that it is huge and somewhat fast!!!...

SCSA

What about the Seagate 7200.9? It has denser platters, would it not be as fast/faster than the Hitachi? Though as far as I know, the Seagate has longer seek times. Even so, the 7200.9 wouldn't be a slouch. My 160GB one is damn fast. Really dense platter with only 2 heads, seek times are under 9ms.

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now