Personal computing discussed

Moderators: renee, morphine, Steel

 
Aruna
Gerbil
Topic Author
Posts: 73
Joined: Tue Feb 19, 2002 7:00 pm

Mon Mar 18, 2002 3:06 pm

So i finally did some PCmark2002, HDtach, sandra2002 testing, and I've encountered something strange.

I've got a 18GB 10K RPM U160 SCSI drive, and a 30GB 7200RPM ATA100 drive.

The ATA100 drive is besting the Ultra160 drive! Tell me there's something wrong with my U160 drive! :sad: Some benchmark numbers: PCmark2002 puts the SCSI drive at 555, and puts the ATA drive at 650.

Has anyone else experienced this?
 
SecretSquirrel
Minister of Gerbil Affairs
Posts: 2726
Joined: Tue Jan 01, 2002 7:00 pm
Location: North DFW suburb...
Contact:

Mon Mar 18, 2002 4:27 pm

Have you checked all the normal things: driver updates, cables too long, bad termination, SCSI bus speed settings, etc.

I do have another theory, but it's rather complex, and I need to know the drive models (as specifically as possible) for each drive to further it.
 
Aruna
Gerbil
Topic Author
Posts: 73
Joined: Tue Feb 19, 2002 7:00 pm

Mon Mar 18, 2002 10:22 pm

Welp.. The SCSI drive gets recognized as an ultra160 drive, running on the latest bios version of the Adaptec 29160 SCSI controller. It IS the only device on an 8 device U160 LVD cable. SCSI HD is a Fujitsu MAG3182MP, 18GB 10K. IDE HD is IBM DTLA- 307030, 30GB 7200rpm, connected to IDE0. It is the only device on the channel as well, and it gets identified as an ATA100 drive, connected with an ATA100 compliant cable. Any suggestions or feedback on why this could be happening?
 
Steel
Global Moderator
Posts: 2330
Joined: Wed Dec 26, 2001 7:00 pm

Tue Mar 19, 2002 12:20 am

Well, just because a drive is SCSI or spins at 10,000 RPM doesn't mean it's always going to be faster than an IDE drive. That Fujitsu drive is a pretty old one and even in its day it trailed just about every other 10K RPM drive at the time. I ran a comparison between the U2W version of the Fujistu drive with the IBM 34GXP in Storage Review's database (the 75GXP was reviewed on a different testbed). Even the older 34GXP beats the Fujitsu in most of the Winbench tests, the 75GXP would only do better.

About the benchmarks: Sandra has never been considered very reliable as a hard disk benchmark and PCMark is way too new to take it's results seriously yet. Winbench 99 provides pretty consistant results and most hard drive makers still use it as the measure of thier drive's performance.
 
SecretSquirrel
Minister of Gerbil Affairs
Posts: 2726
Joined: Tue Jan 01, 2002 7:00 pm
Location: North DFW suburb...
Contact:

Tue Mar 19, 2002 12:31 am

Ok, here is part of the difference. The Fujitsu drive has a maximum transfer rate of 80MB/s. The IBM has a maximum transfer rate of 100MB/s. Both those rate are of course from on drive cache to the controller. I think the rest of it comes down to areal density. The IBM drive uses two platters with a max density of 11.0Gbits/sq-in. The Fujitsu has five platters with a max density of 3.7Gbits/sq-in.

When you cram more data into a smaller area, you make up for the fact that it may be moving slightly slower. The IBM drive has a max sustained transfer rate of 55.5MB/s and minimum of 37MB/s, whereas the Fujitsu has a max rate of 44.68MB/s and minimum of 29.45MB/s.

A difference 9MB/s sustained tranfser rate and 20MB/s burst transfer rate, both in favor of the IBM are enough to cause that difference in score.

edit ---

As a note, the 75GXP data sheet doesn't specify whether 37MB/s is minimum or maximum for the sustained rate. The max media rate is 444Mb/s. I took that a devided by 8 (bits to bytes) and got 55.5MB/s. After further checking it looks like the conversion ratio from media tranfer rate to sustained tranfer rate is 12 to 1 instead of 8 to 1. That would change the numbers above so that the SCSI drive would win in sustained transfer rates by about 20%.

I'm still curious about Fujitsu's specifying 80MB/s as the max data rate, since it is an Ultra-160 drive. Perhaps that is still enough to make the difference in the benchmarks you are running.

<font size=-1>[ This Message was edited by: SecretSquirrel on 2002-03-18 23:57 ]</font>
 
Aruna
Gerbil
Topic Author
Posts: 73
Joined: Tue Feb 19, 2002 7:00 pm

Tue Mar 19, 2002 1:13 am

Ah ha.. That makes much more sense to me now.. I'm not comparing apples to apples. Just did a little test on HDtach 2.61. Not sure if it's "benchmark worthy", but I thought I should share.


<IMG SRC=http://www.vfxvancouver.com/images/techreport/u160ata.jpg>
 
Steel
Global Moderator
Posts: 2330
Joined: Wed Dec 26, 2001 7:00 pm

Tue Mar 19, 2002 10:36 am

The IBM graph looks right, mirrors what I've gotten using Winbench on mine.
Typically high RPM, low seek SCSI drives do better in server environments where there are tons of random accesses. ATA drives are optimized for desktop use where most of the seeks are localized and the cache can do its work. Check out the drive sorts on SR's newest testbed (use the same link as before). You'll find that in most of the desktop benchmarks the 10K RPM SCSI drives get their ass handed to them by the higher end WD ATA drives.
 
SecretSquirrel
Minister of Gerbil Affairs
Posts: 2726
Joined: Tue Jan 01, 2002 7:00 pm
Location: North DFW suburb...
Contact:

Tue Mar 19, 2002 2:15 pm

Yup, your HDTach graphs pretty much proved my theory. I learned something last night while pouring over drive spec sheets. It is nigh impossible to find meaningful numbers. The damn marketing people have got their hands in too many places.

Who is online

Users browsing this forum: No registered users and 1 guest
GZIP: On