//flex table opened by JP

Click to See Complete Forum and Search --> : SCSI Situation


Steve R Jones
01-25-2004, 11:50 AM
I’ve ventured into the land of SCSI after getting a good deal on some used parts. I did some tests.

TEST MACHINE
Operating System->Microsoft Windows XP Professional
CPU Type->Intel Pentium IIIE, 1000 MHz (7.5 x 133)
System Memory-> 512 MB (SDRAM) (PC133)
Video Adapter-> AG315-64 (64 MB)
Optical Drive->LG CD-ROM CRD-8482B (48x CD-ROM)

Disk Drive->Maxtor 52049H3 (20 GB, 7200 RPM, Ultra-ATA/100)

Rotational Speed->7200 RPM
Buffer Size->2 MB
Active UDMA Transfer Mode->UDMA 5 (ATA-100)
Buffer-to-Host Data Rate->100 MB/s

Partitions
C: (FAT32)->3890 MB (420 MB free)
D: (NTFS)->11687 MB (6485 MB free)
E: (NTFS)->3882 MB (3882 MB free)

Disk Drive->IBM IC35L036UWPR15-0 SCSI Disk Device (36 GB, 15000 RPM,
Ultra160 SCSI)

Rotational Speed->15000 RPM
Buffer Size->4 MB
Interface->Ultra160 SCSI
Buffer-to-Host Data Rate->160 MB/s

F: (NTFS)->17500 MB (17500 MB free)
G: (NTFS)->17500 MB (17500 MB free)

I used a cd with 578 Megs of data for ALL of the copying tests.

1. CD to IDE -> 10min 30sec
2. CD to SCCI -> 7min

3. SCSI to SCSI -> 5min
4. IDE to IDE-> 2min 30sec

5. IDE to SCSI -> 3min
6. SCSI to IDE -> 2min

#4 seems a bit weird. I’m going to install XP to the scsi and try some of these again.


Update

I'm now booting off the SCSI on the test system and running XP Pro.

When I run Sandra SiSoftware and do the File System Benchmark the drives comes in lower/slower than my #1 machine's 7200 ide drive. (wd drive w/8meg cache) Me thinks this isn't good.

Course, it probably doesn't help that two different utils from IBM tell me to send the drive back to the manufacture. The system seems to run like a champ. Just wondering if I’m getting the best bang for the buck.

Comments?

BipolarBill
01-25-2004, 12:02 PM
I have the same drive mounted on an LSIu160 adapter. It doesn't bench very well. The Seagate and Quantum competitors are much faster. My WD800JB IDE drive benches better in everything but access times. The IBM runs ~6ms vs. about 17ms fo the WD. I think that the access time is what makes the difference.

Still, in actual use, the IBM is much faster and reactive than an IDE drive. It's doesn't seem to bring the rest of the system to it's knees when doing huge transfers like the IDE drives do. It's noisier and runs hotter, but I like it despite those drawbacks.

Try using Aida32 and HDTach to bench:

http://www.aida32.hu/download/aida32pe_390.exe

http://www.simplisoftware.com/Public/Content/Pages/Products/Benchmarks/HdTach.html

In AIDA32, it's under "Plugin".

Still, I would be concerned about the diagnostics.

Steve R Jones
01-25-2004, 12:15 PM
Thanks. Will give it a shot.

sm8000
01-25-2004, 01:26 PM
I also have the same drive and card, but since it is by itself I'm not seeing the full potential yet.

At this page (http://www.heise.de/ct/ftp/ctsi.shtml) download h2bench.zip (for DOS, not h2benchw.zip) Boot into DOS and run it: h2bench -english -e -c 0 0 (where the last digit is the drive number).

Peter M
01-25-2004, 02:36 PM
Everyone who's using SCSI drives in Windows 2000 or XP please first run a low level benchmark to make sure that write accesses are actually buffered and cached. There have been various reports of Windows' own registry and configuration fsckups that hobbled SCSI drive performance quite effectively. Please research the web.

You should also check with your SCSI adapter's BIOS that the drive is actually running at proper interface speed of 160 or more MB/s.

The IBM SCSI drive should fly a lot faster than what you currently get.

Peter M
01-25-2004, 02:38 PM
Oh, h2bench is good for low level performance assertion. However, with SCSI drives, you need to load the SCSI adapter's DOS driver to enable U160 or higher interface speed.

With the LSI SCSI adapters, put

device=aspi8xx.sys

into your DOS boot drive's config.sys file. (And of course have aspi8xx.sys on that drive - download from www.lsilogic.com, DOS driver package.)

Philip1952
01-26-2004, 11:58 AM
Steve I did a little transfer to see if you were was close to mine. My cd-rom is SCSI based. I did not have a IDE drive worth transfering to. The one in this box is a very old Quantium. But I did transfer between three different makes of SCSI drives.

System spec's
Windows XP Pro on NTFS file system
Abit TH7-2
P4 1.9
512 meg memory
Tekram DC390U3W
Plex 40X SCSI cd=rom
Cheetah 15k.2 HDD "OS drive"
IBM 15k drive
Atlas 2 10k drive

Transfer file size was 623 meg. This was a OEM win 98 disk.

cd-rom to Cheetah drive = 10 min 30 seconds
Cheetah to IBM drive = 2 min 20 seconds
IBM to Atlas drive = 4 min

Now I copyed the same file back across to the drive before it in the test.

Atlas back to IBM = 4 min
IBM back to Cheetah = 2 min 50 seconds.

I am not sure why the copy back to the Cheetah drive was 30 seconds longer. I am writing it of to it being the OS drive.

sm8000
01-29-2004, 12:11 PM
I couldn't load LSI's DOS drivers, I got an error message. However I did install LSI's ASPI layer and my throughput improved by 20-30MB/s, in both h2bench and HDtach.

Peter M
01-29-2004, 01:27 PM
The ASPI layer _is_ the driver - the lowest layer. This is what you needed to load to get U160.

Steve R Jones
01-29-2004, 01:34 PM
I'll give that a shot this evening!!

Peter M
01-29-2004, 04:35 PM
Again, that's only for the raw throughput testing in DOS! Windows doesn't need an ASPI layer to be fast with HDDs.