The big XP defragmenter test

[poutnik]

I upgraded to a 7200 RPM 500GB drive for my notebook. This drive is increatable fast: ~130 MB/s sequentially read and write, ~50 MB/s random read and write with one 4K cluster.

Hm, interesting. 

I cannot imaging myself a disk so good in random 4kB cluster access, unless it is SSD or results are measure of caching efficiency.

For my 3 years old 7200 RPM 250GB SATA2 AHCI I got  AFAIK
about 80 MB/s sequential,
about 30 MB/s for random 512kB
about 0.3 MB/s ( 0.6 with NCQ ) for random 4kB.
( CrystalDiskInfo Benchmark )

This IMHO match better typical sequential speed and average access times.

[BloodySword]

Ooups. I twisted the lines of CrytalHD mark. xD Of course it was not a 4K cluster, it was 512KB random access.

But it is really faster. It has desktop 3.5 Inch niveau!

4K Cluster access was around 0,9 MB/s for read and write.

Drive: Hitachi TravelStar 7K500 500GB

[poutnik]

Interesting review with benchmarks of the disk is here, with some lower numbers.
http://www.storagereview.com/hitachi_travelstar_7k500_review

[BloodySword]

I can not comprehend the CrytalDiskMark result. I get MUCH higher values. Perhaps there were some Windows writes or reads while the 500MB test, this is the reason why I used a 50MB test to avoid interrupting by other running programs.

It could be that the WD is faster but I'm not keen with WD, I hade very bad experiences in the past.

It is MUCH faster than the old 5K320 drive, but quite louder but it saves power.

Access times unter 4 ms is physically not possible with 7200 RPM.

[poutnik]

I guess the disk caching settings can play role too, especially for data size comparable to disk cache.
AFAIK there are 3 levels in Vista+ :
1) Usual write disk caching on OS level
2) Disk write caching
3) "Disk write caching even if PC is told it is already written", not recommended without secondary power source.
    ( english term not handy now )

Sure, 4 ms is avg rotation delay,
by 4.2 they had in mind probably avg seek latency.

[BloodySword]

AFAIK CrytalDiskMark does not use any caches. I don't know 100%.

[poutnik]

AFAIK CrytalDiskMark does not use any caches. I don't know 100%.

The disk driver/firmware settings can be set so, that in effect
the disk drive cache can ignore requests to bypass disk write cache,
deciding on its own when it is written.
It is the last 3rd check box in disk cache policy settings in device manager,   sometimes dimmed.
When I temporarily activated it, the benchmark improved, but not dramatically.

[cesaralvareza65]

If free is paramount then use what the OS provides. There are a few that wanted absolute control over defragmenting such as optimizing big or small files at front or rear of drive or by placing more accessed filess at rear or front of the drive. But then again how could a defragmenter quess what would be best for you. For Windows, http://www.theeldergeek.com/disk_defragmenter_utility.htm writes:
"For the majority of users, the Disk Defragmenter Utility included with XP is sufficient to keep the hard drives in relatively good condition, but it's actually what is known as a Lite or slightly crippled version of Diskeeper, a product made by Executive Software."

[USer]

If free is paramount then use what the OS provides. There are a few that wanted absolute control over defragmenting (...)
But then again how could a defragmenter quess what would be best for you. For Windows, http://www.theeldergeek.com/disk_defragmenter_utility.htm writes:
"For the majority of users, the Disk Defragmenter Utility included with XP is sufficient to keep the hard drives in relatively good condition, but it's actually what is known as a Lite or slightly crippled version of Diskeeper(...)"

I agree -- mostly.
Actually, a defragmenter cannot "guess" much. It can defragment files, and use the access logs which the OS provides (if it does), and probably follow some simple rules which its developer gave it, like
"Place directories including MFT first" ,
"Place recently accessed files next" ,
"Push everything which hasn't been accessed for 90 days to the end of the partition" and so on.
However, I had good results with a two-partition approach:
Partition #1: XP and selected programs only (those which are loaded at boot time)
Partition #2: most programs, games, data, documents, backups, whathaveyou.
At boot time, partition #2 is hardly ever accessed. And because #1 is so tight, there is next to no head movement. I have a dozen games installed, and still boot time in the 40's without defragging.
Of course, after I had everything installed, I defragged the whole mother with a "directories only" script, ie the "data disk monthly" except that there are only 3 zones: Mft, dirs, everything else. That cut down boot time to ~35 seconds, on a 230GiB Seagata Sata.
A more thorough defragging run with "data disk monthly" shaved off another 5 seconds, and it was a better result than "system disk monthly", even on C:.

However, I gave up on the idea of "sort by name". They tend to shuffle way too many data. What use is it to sacrifice ~30 minutes for a mere 1 to 2 seconds game launch time?
If one of the games is noisy on the disk, I modify the script like that:
MFT, dirs, D:\Games\<Game>, recently accessed files, non-spacehog files, spacehogs.
That'll sort out most of the HD noises nicely, and save one-third on load times, give or take.
Sort by name does the same, since the full path is accounted for, but it does go overboard IMO.

If anybody suffers from performance degradation within days, I'd try to introduce / increase gaps near the front, so that new temp files don't have to be created near the end. I don't suspect that defragmented but re-fragmented files are to blame.
~USer

[erudite]

Your XP defragmenter test is amazing.  Thanks a bunch.  Maybe one day you could do a windows 7 one?  Then again I guess the results wouldn't differ much.

[Darlis]

Thanks.
Yes, a Windows 7 version is planned but it may take while, though.