Partitioning for Win 7 - speed issues

[gtvr]

Does someone have a definitive answer as to if I partition my HD for a Win 7 install, which partition will be the faster one?

I have a WD 750GB Caviar Black. I will probably make a 100GB OS partition, 650GB data/etc partition. Clean install of Win7 (64). So I start the installer, make a 100GB partition - does that default to the (outside?) fastest part of the disk?

 

[TheViper]

When you create the partition for the OS, it will ask where you want to put it; beginning or end of the drive. You want the beginning of the unallocated space.

In truth though, the difference will be so minor, it would require synthetic benchmarking to notice.

 

[sminlal]

Yes, the first partition on the drive is the fastest one, it's located at the outermost tracks where the transfer rate is highest. But the seek time is no different, and for booting and starting programs that's typically what's most important.

Beware - if you put two partitions on one drive and use the second partition heavily, you'll probably end up with a system that's slower than if you just leave everything in a single partition.

 

[gtvr]

sminlal - why do you say that you would end up with a slower system that way? I did find an interesting site which suggests that if you make a smaller partition for your OS and apps, you keep the disk heads in that part of the drive so seek times stay low.

 

[sminlal]

I said you'd end up with a slower system if you access the second partition heavily. That would mean you're forcing the access head to move back and forth between the two partitions, and that would slow things down.

If you had only one partition, then most of the files would be situated as close a possible to the outside edge of the disk (at least initially). Putting them into a second partition forces them into more distanct tracks. Unless your first partition is full, there's going to be some empty tracks between the files in the first and second partitions, and the heads will have to spend more time moving across that empty space as they alternate accessing between the two sets of files.

But if you only access files in the second partition occasionally, or especially if you leave the rest of the disk empty (using only the space in one small partition at the start of the disk), then you will see a performance benefit. Don't expect miracles, though - the longest component of access time is the rotational delay of the disk, and using smaller partitions doesn't speed that up at all.

 

[tmike]

>> In truth though, the difference will be so minor, it would require synthetic benchmarking to notice.

Hardly. The performance difference between outer and inner cylinders on large spindles can be seen quite readily.

 

[TheViper]

>> In truth though, the difference will be so minor, it would require synthetic benchmarking to notice.

Hardly. The performance difference between outer and inner cylinders on large spindles can be seen quite readily.

I meant in regards to just using 1 partition for the whole thing.

 

[sminlal]

>> In truth though, the difference will be so minor, it would require synthetic benchmarking to notice.
Hardly. The performance difference between outer and inner cylinders on large spindles can be seen quite readily.

The biggest difference in performance is due to the increased transfer rate which is a result of having more data stored on the outermost tracks. That will show up very clearly on a transfer rate benchmark. But access times aren't very much better on the outermost tracks because most of the access time is caused by the rotational delay, and ALL of the tracks on a drive have the same rotational delay. For most real-world use such as booting and loading applications, access time is a lot more important than transfer rate. So for most people there's not really going to be nearly as much difference in perceived performance as a transfer rate benchmark would suggest.

 

[cjl]

Most of the access time is caused by seek time, and this can be significantly improved by partitioning. Most 7200rpm drives have somewhere between a 12 and 14ms access time. Of this, 4.16ms (regardless of model) is rotational latency. The other 8-10ms is seek time, and this can be significantly improved if the drive is "short-stroked" (in essence, if only the outer portions of the drive are used to shorten the maximum seek length).

 

[TheViper]

But an 8-10 ms reduction is seek time is only a significant improvement in synthetic benchmarks. You'd never notice that in daily use.

 

[cjl]

Actually, you would notice it. If you could shave 4-5ms off a Caviar Black's access time, it would essentially turn it into a velociraptor.

Look at it this way. If you can shave 4ms off, every single random access takes 33% less time. Every one. That's a noticeable change.