Cheetah is not running as fast as it should

[YoYo]

:-?
I've got a Cheetah ST118202LC 10kRPM, which depending on whose words you believe, should be able to deftly run at between 28.8Mbyte/sec to 80Mbyte/sec, but doesn’t seem like wanting to go any faster than 18 or 19 Mbyte/sec. This Cheetah has previously been struck down by a virus but has since recovered and hopefully nursed back to health (basically by administering a default parameters reload). Are there any vital signs and bodily conditions I should have checked and monitored, but missed to do so?

Any valuable advice and assistance from our learned practitioners in this forum would be gratefully appreciated.

 

[Mercutio]

What are you using to benchmark? What kind of controller are you using? Do you have a Seagate drive or an OEM model? What else is on your cabling? What kind of cabling do you have?

Uh, we need a little more info.

 

[YoYo]

Hello,
The following is an X-ray of the viscera of the animal. Is everything where they should be?

########################
Page 08 - Caching parameters
########################
*Initiator Control (IC)= 0 (0x0)
Abort Pre-Fetch (ABPF)= 0 (0x0)
*Caching Analysis Permit (CAP)= 0 (0x0)
*Discontinuity (DISC)= 1 (0x1)
*Size Enable (SIZE)= 0 (0x0)
*Write Cache Enable (WCE)= 1 (0x1)
Multiplication Factor (MF)= 0 (0x0)
*Read Cache Disable (RCD)= 0 (0x0)
Demand Read Rentention Priority= 0 (0x0)
Write Rentention Priority= 0 (0x0)
Disable Pre-Fetch Transfer Length= 65535 (0xFFFF)
*Minimum Pre-Fetch= 0 (0x0)
*Maximum Pre-Fetch= 65535 (0xFFFF)
Maximum Pre-Fetch Ceiling= 65535 (0xFFFF)
*Force Sequential Write (FSW)= 1 (0x1)
*Logical block cache semgment size (LBCSS)= 0 (0x0)
*Disable Read-Ahead (DRA)= 0 (0x0)
*Number of Cache Segments= 3 (0x3)
*Cache Segment Size= 24576 (0x6000)
Non-Cache Segment Size= 0 (0x0)

'*' entries are changeable.


Other label particulars of this patient:
Drive: Seagate Cheetah 18G
Model: ST118202LC (detected by HA bios as SX118202LS)
Firmware: B801 (diagnosed by SCSI utility as B808)
Serial number: LK588967
Part number: 9J9006-422, Made in Singapore (=OEM)
Glist = 0
ETF defects = 1415

Host Adapter: Adaptec AIC7890AB
Host Adapter Bios: “589217rE9100(c)1999v2.20” shown on label; “Adaptec SCSI Bios v3.10 (c) 2001” displayed on post screen. (This adapter is also OEM)

Cabling: non-terminated 68-pin

Cable converter: 68-pin to 80-pin with a separate 4-pin power connector, and a 4-way jumper.

Other device: IMB LVD DXHS36D (may also be OEM)

Benchmark tools used: simple HDD tools downloaded.

 

[timwhit]

You should definately have a terminator on the cable. This could possibly be what your problem is.

 

[YoYo]

I had thought terminating at the HA should be sufficient, however tests using HD Tach v2.70 by www.simplisoftware.com gave the perplexing results below, so I’m still baffled.

IBM DXHS36D
test actions: random seek; read burst; read max
cable & HA terminated: 9.7ms; 27.7Mbyte/s; 20.4Mbyte/s
cable only terminated: 9.8ms; 27.7Mbyte/s; 20.3Mbyte/s
HA only terminated: 9.8ms; 43.1Mbyte/s; 20.4Mbyte/s

Cheetah ST118202LC
test actions:random seek; read burst; read max
cable & HA terminated: 8.2ms; 27.8Mbyte/s; 19.2Mbyte/s
cable only terminated: 8.4ms; 27.7Mbyte/s; 19.2Mbyte/s
HA only terminated: 8.6ms; 43.2Mbyte/s; 19.3Mbyte/s

 

[The JoJo]

The cable must be terminated at both ends, the controller is usually the terminator at one end, but then you must deffinitely have termination at the other end.

 

[sechs]

It must be *properly* terminated at both ends.

 

[YoYo]

The silver cable I used for the above-mentioned tests has four device connectors and (I think,) a passive terminator on it. Could this cable be specifically tuned for a 4-device setup, hence the drop in (read burst) performance in my 2-device arrangement test??

 

[cquinn]

Passive termination is not good enough. And that is probably
not a cable that is rated for the higher spec speeds you are
trying to reach.

 

[Santilli]

Cable converter: 68-pin to 80-pin with a separate 4-pin power connector, and a 4-way jumper.

This could also be the problem.

s