New hdd, speed improvement?

Associate
Joined
15 Sep 2008
Posts
7
I need to get another hard disk for backups & I've been looking for a way to speed up things recently, particularly boot time.

Are there any drives that are likely to give me a noticeable improvement in speed if I were to make it my OS drive?

My current setup is:
GA-M59-S5
Athlon X2 4200+
Nv 7900GS
3Gb PC6400
Seagate 7200.10 320gb
Vista Ultimate
 
Boot times are down to random access + sustained transfer.

The sustained transfer on newer drives such as the 7200.11 and Samsung F1's are more than 50% higher than the 7200.10.

This translates into snappier load times by a considerable margain. With Vista Ultimate, I would definitely recommend turning on ACHI if that is an option.
 
In a typical case, startup time is most influenced by the disk's access time. This is directly affected by the rpm of the disk. Expect 8.3ms from 7,200rpm, 6ms from 10,000rpm (e.g. WD Velociraptors), 4ms from 15,000rpm (high end SAS disks) and 0.1ms for SSD.

You can, however shift this from being access time limited to being read speed limited by optimizing the layout of the files on the disk. Modern disks have large caches and read ahead on every read because the most likely data to be required next is the data immediately following the data that has just been read. This means that fragmentation makes read ahead "miss", and the next block of data has to be found elsewhere on a disk, which incurs the seek-time penalty. If you have a decent defragmentation program / disk optimizer, it will typically install a low lever driver that logs the order in which the files are accessed at startup. It can then re-order the files on disk in such a way that all the files accessed at startup are ordered sequentially right after one another on the disk.

So, for example, on an fragmented disk, if startup requires 1000 files to be read, at 8ms seek time each, it will take 8 seconds to find all the files, on top of the time required to load them from the medium. The 8s overhead also assumes each file is not fragmented in itself, otherwise each fragment will incur the additional seek penalty. On the other hand, if all the files are laid out in the correct order, they can effectively be read in a single operation without seeking, and read-ahead on the disk's cache will always score a "hit". This will get OS start-up time in the same ball-park as an SSD would yield.

Unfortunately, I'm not sure what (if any) defragmentation programs support this form of file re-ordering (and logging that is required to find out what order the files are accessed in) nowdays. Norton used to be able to do it back in the Win9x days, but I've not really used Windows for any significant length of time since 1996, so I have no idea if the tools today are as useful as they were back then.
 
Get a VelociRaptor. F1 drives are good but no match for the VR. Even the 'old' Raptors out perform a F1 320gb for os. I went to a 320gb F1 from a 74gb Raptor thinking I would get a improvement in boot up but in tests the F1 takes a full 10 seconds longer than my old tech Raptor to get to the same point.
 
I would say avoid sammy F1-320 as a boot drive. Got one last week and very disappointed with it so far nowhere near as quick as my other Sammy F1-750 (250GB platters + 32mb cache). Some guy even comments on this in the reviews section on the product page!
 
It's also worth looking at the sustained transfer rates on the manufacturer web sites. Not all disks are the same, and the biggest available disks (rpm being equal) will generally be faster, because they'll have more platters, which means more heads. The sustained transfer rates are directly proportional to the number of heads on the disk (whereas the access time is generally only a factor of rpm). This is why my 500GB Samsung disks only manage about half the sustained throughput of my 1TB Seagate disks.

There is, of course, a downside. In theory, more platters/heads means more parts that can fail and render the disk unusable.
 
Hmm, interesting.
Not sure whether to go for a WD Raptor or just get a cheap backup drive for now and look at SSD's in the future.
 
You should probably be aware that SSDs tend to be very bad at write speeds because they have to overwrite the entire bank of flash for each write. There is commercial software available that works around this in a clever way, but it is quite expensive. A log-based file system (logfs for Linux, IIRC) also goes a long way toward getting most out of an SSD, but sadly, if you use Windows, you're pretty much out of luck.

Basically: SSDs are great for read speed and access time, bue quite slow for writes, especially if you have a lot of small writes, unless you have a file system that works around the problem.
 
But 320GB uses one platter, and 640GB uses 2, thus 640GB is 2x faster. Similarly, 750GB and 500GB use same platters, so 750GB is 50% faster than 500GB.
 
i had this trouble about a week ago raplaced an old WD 200gb with a 640gb f1, computer runs smoother and is alittle bit faster, its worth getting a new drive..
 
But 320GB uses one platter, and 640GB uses 2, thus 640GB is 2x faster. Similarly, 750GB and 500GB use same platters, so 750GB is 50% faster than 500GB.

you're hugely mistaken

the only way you can use 2x and 50% when talking about those drives is the difference in size, not drive speed
 
Not true. The number of heads (which is usually 2x the number of platters, but it could be less with single-sided platters) does determine the speed as much as the platter density. Heads are effectively linked up in a RAID0 array, the analogy is pretty much spot on. Thus, a drive with 2 heads will be about 2x faster on sustained transfer than a similar drive with 1 (rpm and platter density being the same).
 
Not true. The number of heads (which is usually 2x the number of platters, but it could be less with single-sided platters) does determine the speed as much as the platter density. Heads are effectively linked up in a RAID0 array, the analogy is pretty much spot on. Thus, a drive with 2 heads will be about 2x faster on sustained transfer than a similar drive with 1 (rpm and platter density being the same).

any chance that you can prove that? reason i say is because from various reviews on the net and my own personal testing i have drives like the 150gb raptor and a 74gb raptor, only difference is 150gb raptor has 2 platters and its hdtune transfer rate graphs are identical to the 74gb model.

same for my seagate 500gb and 750gb drives, both drives use same sized platters but more platters in the 750gb model, hdtune graph again shows no performance advantage for the 750gb drive which has more platters than the 500gb one.

same with the samsung f1 drives iv had, the 500gb and the 750gb use 250gb platters, the 750gb model uses 3 platters and 500gb uses 2 platters, hdtune graphs in performance is the same.

gordan it could be your looking at drives which dont have the same platter density which is why your mistaken with the performance graphs. you need to read up a lot more on this subject to get a grasp on hard drive basics.

also another possible negative for drives with more platters is that there are more moving parts inside so possibility for failure could be higher on drive with more platters.
 
Back
Top Bottom