Harddisk overview updated

H

hjreggel

Weaksauce
Joined
Sep 23, 2005
Messages
96
Hi,

I managed to update my harddisk overview once again. Now I'll try to do more than just two update cycles per year. For those who don't read the comments: I only list the biggest or fastest models.

http://www.hjreggel.net/hdtechdat/

I also added some graphs, especially for the WD Raptor WD1500ADFD, the Fujitsu MAX3073, and several 2.5" models. I only show the actual sustained read rate, no pretty-printed graphs. This is fast as it can get, no more - but less ;-)

Test Data SCSI
Test Data 3.5"
Test Data 2.5"

I have plans for other sections, but not enough spare time. Just a few comments:

Fujitsu MAX3073 vs. WD 1500ADFD
- About 4MB/s higher sustained data rate
- About 2/3 of the access time
- More noise, more heat
- Benchmarks report less overall perfomance for the Fujitsu on the Adaptec 29160 versus the WD Raptor on S-ATA I.

I have PATA to SATA and SATA to PATA converters, stacking both results in almost no loss with the old WD Raptor WD740GD. I did not test with the WD1500ADFD, yet.

My CardBus S-ATA has about 81MB/s read, but only 17.5MB/s write.
My CardBus U2W SCSI has about 63MB/s read, but only 18.3MB/s write.
I dont't know whether this is caused by the card or the Notebook chipset.

Hans-Jürgen
 
You know what would be quite nice: If the table in "Technical data" had links to the tests, so that I could click on "WD1500ADFD" and get to the chart in your "Test Data" area. Otherwise, at least some specifications in the "Test Data" area would be nice. I just don't know enough about 2.5" drives to recognize the HM160JC as a 5400 rpm drive.
 
Hi,

drizzt81 said:
You know what would be quite nice: If the table in "Technical data" had links to the tests, so that I could click on "WD1500ADFD" and get to the chart in your "Test Data" area.
Click to expand...

Well, i'll think about that. Maybe I add a link so that clicking the entry opens the graph in a new window. I just won't get the chance to get hold of enough disks...

drizzt81 said:
Otherwise, at least some specifications in the "Test Data" area would be nice. I just don't know enough about 2.5" drives to recognize the HM160JC as a 5400 rpm drive.
Click to expand...

Excellent point. I already thought about writing more comments, but now I added a "quick feature list" below the heading.

BTW: The main tables are veeery wide, so it's best to print on Legal/A3 landscape. But I tried to put the important stuff in the leftmost columns.

Hans-Jürgen
 
Hi,

I updated the test results for the Samsung HM160JC 2.5" 160GB disk and added the test results for the Samsung HD501LJ 2.5" 500GB together with an analysis of the test results:
HM160JC 2.5" 160GB
HD501LJ 2.5" 500GB
HD501LJ Analysis

The Samsung HD400LD 3.5" 400GB will be next...

Hans-Jürgen

P.S.: I just noticed that the pages look weird on Opera, and that Netscape seems to have some JavaScript errors. I'll try to fix this as soon as possible...
 
Hi,

I just finished evaluating the test results from the Samsung Spinpoint T133 400GB PATA harddisk.

Test Details
Analysis

The two Samsung disks HD501LJ and HD400LD remind me of the rule "You get what you pay for" :mad:

I also added a page where you can play around with three different calculations related to harddisks:

Harddisk Calculator

The most impressive example is calculating the data rate for a 10,000RPM 75MB/s harddisk with 4ms average seek when reading 100 files of 0.1MB. And then trying to improve the resulting data rate by doubling the sustained data rate assuming RAID-0 (striping).

Hans-Jürgen
 
Hi,

I finished the first tests with the
WD5000AAKS.

I also did a research on the model numbers. Although the WD5000KS and WD5000AAKS are different models, it looks like it's just old/new numbering scheme where the additional "AA" just means 3.5" Desktop Caviar Series.

Hans-Jürgen
 
hjreggel said:
Hi,

I finished the first tests with the
WD5000AAKS.

I also did a research on the model numbers. Although the WD5000KS and WD5000AAKS are different models, it looks like it's just old/new numbering scheme where the additional "AA" just means 3.5" Desktop Caviar Series.

Hans-Jürgen
Click to expand...

The little I could ever find on the AAKS is that it is 3 platters where the KS is 4 platters. Glad to see those numbers :)
 
Since I was once told "there are no stupid questions":
Why are some of the graphs very "fat" (like the AAKS) while others (740GD, Desktar 500GB) are rather skinny? At first I assumed it was due to drive size (500GB implying more samples), but when comparing the 500GB Deskstar to the WD 500GB model, it is clear that the former has less "jitter", especially when considering that the range of the Deskstar graph is smaller (40MiB/s vs 55MiB/s). Also, what program are you using to generate these graphs? Do I understand that the AAKS is no a "perpendicular recording" drive?
 
Hi,

drizzt81 said:
Since I was once told "there are no stupid questions"
Click to expand...

No problem. I should not post here if I did not want to deal with questions.

drizzt81 said:
Why are some of the graphs very "fat" (like the AAKS) while others (740GD, Desktar 500GB) are rather skinny? At first I assumed it was due to drive size (500GB implying more samples), but when comparing the 500GB Deskstar to the WD 500GB model, it is clear that the former has less "jitter", especially when considering that the range of the Deskstar graph is smaller (40MiB/s vs 55MiB/s).
Click to expand...

Even if there are more samples involved, samples with equal data rate should lead to a straight line. The amount of jitter shows the different data rates for the samples. A harddisk has no "true" sustained data rate. Data is read track by track across all heads, an then cylinder by cylinder. Current harddisks have up to about 1600 sectors per track, that's 800kB per track. Even with 5 platters, that's 10 tracks per cylinder, i.e. 8MB per cylinder. The default sample size for the tests is 32MB, so the influence of head and cylinder switches should not have impact on individual samples.

EDIT: If you take a look at the "Analysis" section, you can see graphs with a detailed views where the individual samples are marked with dots.

There might be several reasons for jitter: If the jitter is symmetrical to the average, this might be a timing problem between internal data transfer from disk to buffer and external data transfer from buffer to host. If it is more like a drop from the maximum, it's most likely missed head or cylinder switches, where the controller misses the start of the next track and waits one full revolution. A combination of both might be a missed track start, but the rest of the track gets cached until the start is read, which means that the controller uses the cache to compensate the missed switches.

I suppose that the WD5000AAKS case might be the latter, because the jitter is too big for simple timing problems.

drizzt81 said:
Also, what program are you using to generate these graphs? Do I understand that the AAKS is no a "perpendicular recording" drive?
Click to expand...

The graphs themselves are screenshots from MS Excel: One column with the data position (GB) the other with the data rate (MB/s), then an XY Scatter graph, straight lines. The source data is taken with a program that I originally wrote for my flash card tests, but at some point I noticed that I can use it for harddisk benchmarks, too. For several reasons, I can't hand out that software... But you could take any data from benchmarks that let you export the plain figures.

Hans-Jürgen
 
You must log in or register to reply here.
Back
Top