simon@barnum.enet.dec.com (Curiosier and Curiosier...) (09/14/89)
Yesterday I installed a 12 MHz 286 accelerator (Orchid's Tiny Turbo XTra)
into my 8 MHz XT. The Norton SI went to 10.5 from 1.7, Landmark shows
8.7 MHz. These tests are recommended by the accelerator manual for
installation verification and for comparison. So far so good. But running
some real application I noticed a considerable slow down from a disk (so
considerable, in fact, that it was practically intolerable). Ran SI again,
this time with a disk performance. It was 0.5 of XT! Before the accelerator
installation it was 2.
The disk drive is a 72 Mb Maxtor 2085. The low level format was done with
interleave factor 3. FDISK from Compaq MS-DOS 3.31 partitioned it into
two partitions, 1 (C:) and 2, 2 being logically divided into drives D: and
E:.
I ran a SPINTEST (shareware) on drive 1 only. With the original 8088
it was around 200,000 bit/sec. With the 286 I got only 30,000. An HDTEST
(also shareware) showed two drives, 0 and 1. Interleave measurement showed
that changing it from 3 to 5 would increase the performance by around 300%.
I did, and the SPINTEST indicated the rate of around 100,000. Running
an application I got a feeling that the speed was satisfactory.
I went back to the SPINTEST, to test drive 2. It bounced me back to DOS
with a message: "System error". I went to the HDTEST and tried to change
the interleave. HDTEST can do it without destroying the data. However,
it gave me an error message ("seek failure" for some cylinders, "format
error" for others). Eventually I aborted it.
Now with only disk 1 re-formatted the performance is at best acceptable,
but I would like to get the previous 200,000 bit per sec.
A few questions that I have:
1. Was it normal that the disk performance changed so drastically
because of an accelerator?
2. What is this System Error?
3. What should I do to fix it?
4. Is a radical measure necessary such as doing a low level format
again?
5. How can I change the interleave on the drive 2?
Any help will be appreciated.
---
Leo Simon simon@hpstek.enet.dec.com
--or-- ...!decwrl!hpstek.enet.dec.com!simon
--or-- simon%hpstek.enet.dec@decwrl.dec.comjearly@lehi3b15.csee.Lehigh.EDU (John Early) (09/16/89)
In article <456@ryn.esg.dec.com> simon@barnum.enet.dec.com (Curiosier and Curiosier...) writes: > >Yesterday I installed a 12 MHz 286 accelerator (Orchid's Tiny Turbo XTra) >into my 8 MHz XT. The Norton SI went to 10.5 from 1.7, Landmark shows >8.7 MHz. These tests are recommended by the accelerator manual for [stuff deleted] >this time with a disk performance. It was 0.5 of XT! Before the accelerator >installation it was 2. I had the same problem with my Seagate 251 when I chaged my 8088 to a V20... my disk went from 1.8 to 0.6! I backed up all 3 partitions and re-formatted with an interleave of 5 (which made sense, considering the slowdown--I must admit to lacking the software for diagnostics that you do) The speed returned to a 1.8. I would suggest re-formating the whole drive, unless you want to try Spinwrite(commercial), which has worked wonders for friends. >1. Was it normal that the disk performance changed so drastically >because of an accelerator? Probably _is_ normal, since the increased processor speed throws the I/O off. John Early jearly@lehi3b15.csee.lehigh.edu-- John Early jearly@lehi3b15.csee.lehigh.edu JPE1@Lehigh.Bitnet
mlord@bmers58.UUCP (Mark Lord) (09/16/89)
My guess is that the 286 accelerator is designed for a 4.77Mhz IBM PC bus, and therefore always assumes that that is what it is dealing with. Consequently, it does slower transfers than your original 8Mhz turbo XT. As a result of this, it misses the interleave on your hard drive (as I recall the original IBM-PC plus hard drive used 4:1 or 5:1 interleave). -Mark
simon@barnum.enet.dec.com (Curiosier and Curiosier...) (09/16/89)
In article <621@lehi3b15.csee.Lehigh.EDU>, jearly@lehi3b15.csee.Lehigh.EDU (John Early) writes... >In article <456@ryn.esg.dec.com> simon@barnum.enet.dec.com (Curiosier and Curiosier...) writes: >> >>Yesterday I installed a 12 MHz 286 accelerator (Orchid's Tiny Turbo XTra) >>into my 8 MHz XT. The Norton SI went to 10.5 from 1.7, Landmark shows >>8.7 MHz. These tests are recommended by the accelerator manual for >[stuff deleted] >>this time with a disk performance. It was 0.5 of XT! Before the accelerator >>installation it was 2. > >I had the same problem with my Seagate 251 when I chaged my 8088 to a V20... >my disk went from 1.8 to 0.6! I backed up all 3 partitions and re-formatted >with an interleave of 5 (which made sense, considering the slowdown--I must >admit to lacking the software for diagnostics that you do) The speed returned >to a 1.8. I would suggest re-formating the whole drive, unless you want to >try Spinwrite(commercial), which has worked wonders for friends. I did all of the above: backed up all three partitions, reformatted the drive with interleave of 5, and got the same result as I got when changing interleave without destroyung the date. With 8088 and interleave of 3 I had a factor of 2x XT, with 80286 and interleave of 3 it was 0.5, with interleave of 5 I got 1.5. Not the greatest but tolerable. But measuring with SPINTEST shows that the present throughput of 105K bit/sec is still well below previous 175K bit/sec with 8088. Is there a way to increase the disk performance? --- Leo Simon simon@hpstek.enet.dec.com --or-- ...!decwrl!hpstek.enet.dec.com!simon --or-- simon%hpstek.enet.dec@decwrl.dec.com
slimer@trsvax.UUCP (09/17/89)
The interleave factor has a drastic affect on the performance of your hardware. With a factor of three, this means that every third sector will be the continuous sector. This is done for those slow machines that cannot keep up with the hard disk. When a sector is read, by the time the machine is ready for another sector, the one it was supposed to read has already spun past the heads and you have to wait for it to come around again. By speeding up your performance of the CPU, you can read data faster from the hard disk. With the interleave so high at three(2 being normal usually), now the CPU is waiting on the drive to come around to the right sector. Perhaps an interleave of two would suffice, but to re-interleave a drive properly, you should re-low-level-format the drive with the new interleave specs. Maybe your performance degradation is due to this phenomenon (sp). **************************************************************************** * Thank You, texbell!letni!rwsys!trsvax!slimer * * Bill "Icon do windows!" - ComputerWorld * * George W. Pogue, 1300 Two Tandy, Fort Worth, TX. 76102 (817) 390-2871 * ****************************************************************************