Mask vs. MaxTransfer 1200 IDE

Mask vs. MaxTransfer 1200 IDE

Post by Jill How » Thu, 16 Apr 1998 04:00:00



I've gotten all sorts of conflicting answers to last night's posting.

  What are the correct settings for a Maxtor 25128 AT 2.5"  HD?  Can the
AMAX partition have different settings frmo the other partns?  Maxtor
could only tell me the drive as a 32K cache but they didn't know if that
would play into either Amiga 1200 setting.

  The HD had worked with Mask=0x1ffe0  and  MaxTransfer=0xffffff (the AMAX
partition had a Mask=0x7ffffffe, which I left alone).

  When I repartitioned it, HDToolBox gave a Mask of 0x7ffffffe to all the
Amiga partitions.  And I'm having programs freak out upon loading.

PS Here are the Maxtor specs...if that helps any...
===============================================================================
                      MAXTOR 25128A 2.5-Inch IDE DRIVE

  * 15ms average seek                             * 32K on-board cache
  * Embedded AT controller                        * High data transfer rate

Formatted: ................................. 128,235,520 bytes
Tracks per surface: ........................ 1,092
Data Zones: ................................ 4
Sectors per track: ......................... 30-60

Seek times* (ms)
     Track to track: ....................... 3
     Average: .............................. 15
     Maximum: .............................. 27

Average latency (ms) ....................... 8.4
Rotation speed ............................. 3,551 RPM
Controller overhead ........................ 1 ms

Data transfer rate (Mbytes per sec)
     To/from media ......................... 1.19-2.29
     To/from buffer (AT) ................... 8.0

Buffer Size: ............................... 32K

Jill

Please CC: reply... our news feed is bad.

 
 
 

Mask vs. MaxTransfer 1200 IDE

Post by Gene Hesket » Thu, 16 Apr 1998 04:00:00


Gene Heskett sends Greetings to Jill Howe;

Quote:

>I've gotten all sorts of conflicting answers to last night's posting.

No doubt many from people who have _never_ dealt with a Maxtor drive,
simply repeating "company policy" ad nauseum.  Until such time as they
have in fact dealt with such an elderly Maxtor drive, they should, with
all due respect for their expertise in other arena's, stay the hell out
of this one!

Quote:

>  What are the correct settings for a Maxtor 25128 AT 2.5"  HD?  Can the
>AMAX partition have different settings frmo the other partns?  Maxtor
>could only tell me the drive as a 32K cache but they didn't know if that
>would play into either Amiga 1200 setting.

And there exactly, is the answer Jill.  A 32k cache.  Translated to the
correct "MaxTransfer", this is 0x7fe0, one sector less than the cache
size.  Its also exactly what I said last night, but with the actual
figure now filled in.

The mask is probably a non-issue, but this MaxTransfer value must be
asserted, using hdtoolbox or other suitable disk rdb preparation
software, in _every partition_ of that drive, AMAX included.

Quote:

>  The HD had worked with Mask=0x1ffe0  and  MaxTransfer=0xffffff (the AMAX
>partition had a Mask=0x7ffffffe, which I left alone).

>  When I repartitioned it, HDToolBox gave a Mask of 0x7ffffffe to all the
>Amiga partitions.  And I'm having programs freak out upon loading.

And I repeat Jill, they are not being miss-loaded, they are being
miss-written in the first place, and no amount of re-reads will ever
succeed in getting a good read once the miss-write has trashed the file.
The *_only_* cure is to fix the maxtransfer, and *_then_* re-write the
file.  You may even have to start with a fresh logical format of each
partition that now has a trashed file in it.

Quote:

>PS Here are the Maxtor specs...if that helps any...
>============================================================================
>===
>                      MAXTOR 25128A 2.5-Inch IDE DRIVE
>  * 15ms average seek                             * 32K on-board cache
>  * Embedded AT controller                        * High data transfer rate
>Formatted: ................................. 128,235,520 bytes
>Tracks per surface: ........................ 1,092
>Data Zones: ................................ 4
>Sectors per track: ......................... 30-60
>Seek times* (ms)
>     Track to track: ....................... 3
>     Average: .............................. 15
>     Maximum: .............................. 27
>Average latency (ms) ....................... 8.4
>Rotation speed ............................. 3,551 RPM
>Controller overhead ........................ 1 ms
>Data transfer rate (Mbytes per sec)
>     To/from media ......................... 1.19-2.29
>     To/from buffer (AT) ................... 8.0
>Buffer Size: ............................... 32K

>Jill
>Please CC: reply... our news feed is bad.

Cheers, Gene
--
  Gene Heskett, CET, UHK       |Amiga A2k Zeus040 50 megs fast/2 megs chip



--

 
 
 

Mask vs. MaxTransfer 1200 IDE

Post by Ben Hutching » Fri, 17 Apr 1998 04:00:00



<snip>

Quote:>>  What are the correct settings for a Maxtor 25128 AT 2.5"  HD?  Can the
>>AMAX partition have different settings frmo the other partns?  Maxtor
>>could only tell me the drive as a 32K cache but they didn't know if that
>>would play into either Amiga 1200 setting.

>And there exactly, is the answer Jill.  A 32k cache.  Translated to the
>correct "MaxTransfer", this is 0x7fe0, one sector less than the cache
>size.  Its also exactly what I said last night, but with the actual
>figure now filled in.

<snip>

Since when has MaxTransfer had *anything* to do with the cache size?

--
Ben Hutchings -- should be revising | Jay Miner Society: http://www.jms.org/

It is a miracle that curiousity survives formal education. - Albert Einstein

 
 
 

Mask vs. MaxTransfer 1200 IDE

Post by Fred Wrigh » Fri, 17 Apr 1998 04:00:00



: Since when has MaxTransfer had *anything* to do with the cache size?

When the drive has firmware bugs.  A *correct* implementation doesn't
require the software to be aware of the cache size.

                                        Fred Wright

 
 
 

Mask vs. MaxTransfer 1200 IDE

Post by Fred Wrigh » Fri, 17 Apr 1998 04:00:00



: The mask is probably a non-issue, but this MaxTransfer value must be
: asserted, using hdtoolbox or other suitable disk rdb preparation
: software, in _every partition_ of that drive, AMAX included.

The mask might be mildly relevant given the large MaxTransfer.  Since
the mask she had restricts transfers to CHIP RAM only, the poorer
access time might reduce the probability of filling the buffer when
writing.  Perhaps this was someone's empirical "fix".

                                        Fred Wright

 
 
 

Mask vs. MaxTransfer 1200 IDE

Post by Wolfgang Hochwelle » Fri, 17 Apr 1998 04:00:00


Quote:> Since when has MaxTransfer had *anything* to do with the cache size?

Since ages - atleast for IDE disks.
Most ( if not all ) IDE disks are not able to handle
buffer ( cache ) overruns properly, that
is the reason.

Wolfgang Hochweller

 
 
 

Mask vs. MaxTransfer 1200 IDE

Post by Michael Sin » Fri, 17 Apr 1998 04:00:00



> > Since when has MaxTransfer had *anything* to do with the cache size?

> Since ages - atleast for IDE disks.
> Most ( if not all ) IDE disks are not able to handle
> buffer ( cache ) overruns properly, that
> is the reason.

No, that is not the reason.

The reason is that a number of IDE drives failed to correctly
handle one of two special cases.  To understand these cases,
you have to know how MaxTransfer is used.

In the IDE driver (and SCSI driver) the software will try to
ask the disk drive for more than one sector at a time whenever
possible.  This is important since it reduces command overhead
and also can get all of the sectors in a single revolution of
the disk (or n-revolutions if there are enough sectors)  This
makes things *much* faster than if you have to wait a revolution
per sector.

Now, in IDE, the sector count is an 8-bit value.  The spec
says that the value is from 1 to 256.  Ahh, but how do you get
256 from an 8-bit value?  Well, since it is silly to ask for
0 sectors, the spec says that 0 means 256 sectors.

Ok, so here is the first type of problem many drives had.  They
did not handle 0 correctly.  Either they failed to transfer
anything or the hung or they did some other random thing.
Some people we talked to from the drive makers even looked
at the drive microcontroller source code and told us this much
when we reported it to them.

Why was this not found without the Amiga?  Well, the PC was the
main user of drives and for a long time it did not do multi-sector
disk I/O.  When they did start doing that in the drive caching
software (including SmartDrive from MS) they still had this
strang 64K limitation - the segment size of the CPU *and* the
largest cluster size.  So, they never needed more than 128 sectors.

Even more interesting is the fact that 128-255 sometimes did not
work.  Why?  Well, those values, when in a signed-byte, are
negative.  If the software in the drive did that, they would also
cause confusing results.

Why were these not caught?  Well, even when the caching software
on the PCs started to do multi-sector transfers, most drives never
had cluster sizes over 32K.  In fact, only NT supports 64K cluster
sizes and you would only need this if you went over 2gig in a
single partition.  How many 2+ gig drives were around back then?
Very few...

Anyway, so, the safe setting was such that the Amiga, which did not
have cluster sizes nor a 64K segment size, was to never ask for
more than 127 sectors in a single IDE command.  That works out to
something just under 32K (128 sectors is 32K)

It is a shame that MaxTransfer was called what it was.  It is
a shame that people thought it was a transfer rate (well, setting
it too low would slow things down)  It is a shame that people
now think that it has anything to do with caches on the drives.
(Many of the drives back then only had 4K of cache, but later
had 256K of cache, in either case they still had one or both
of these bugs.)

--
Michael Sinz -- ex-Senior Amiga Systems Engineer - Commodore-Amiga
My place on the web -----> http://www.users.fast.net/~michael_sinz

 
 
 

Mask vs. MaxTransfer 1200 IDE

Post by Gene Hesket » Fri, 17 Apr 1998 04:00:00


Gene Heskett sends Greetings to Fred Wright;


>: The mask is probably a non-issue, but this MaxTransfer value must be
>: asserted, using hdtoolbox or other suitable disk rdb preparation
>: software, in _every partition_ of that drive, AMAX included.
>The mask might be mildly relevant given the large MaxTransfer.  Since
>the mask she had restricts transfers to CHIP RAM only, the poorer
>access time might reduce the probability of filling the buffer when
>writing.  Perhaps this was someone's empirical "fix".

I thought of that too Fred, but as it was the post shoulda had a 'long
post' warning.  And then I screwed the math, for a 32k cache, it should
have been 0x7e00, not 0x7fe0. But on the other hand, the driver would
have reduced it to 0x7e00 anyway, so even thats a non-issue.

But you are correct, with the buffers in chip, even that old slowpoke
drive might have been able to keep up!

Cheers, Gene
--
  Gene Heskett, CET, UHK       |Amiga A2k Zeus040 50 megs fast/2 megs chip



--

 
 
 

Mask vs. MaxTransfer 1200 IDE

Post by Jamie Kei » Fri, 17 Apr 1998 04:00:00


Hi.

Michael seems to have a strange talent to stop a discussion like this
dead. And long may it be so!
Oh, and thanks for continuing support for Enforcer.

Jamie Keir


<stuff>

> Michael Sinz -- ex-Senior Amiga Systems Engineer - Commodore-Amiga

Ahh. That'll be why, then.

--
  _______________________________________________________________
 /###############################################################\
+--------------------+---------------+----------------------------+
| Oracle 7     DBA   | Digital Unix  | Jamie Keir                 |

| PL*SQL             | AmigaOS 3     |                            |
| Unix Admin         | Linux         |                            |
| Network Support    | ADA           | ...now if I could just get |
|                    | C / C++       | them to *pay* me for it... |
| Alpha AXP          | M68040 Asm    |                            |
+--------------------+---------------+----------------------------+

 
 
 

Mask vs. MaxTransfer 1200 IDE

Post by Joanne Do » Sat, 18 Apr 1998 04:00:00





> <snip>
> >>  What are the correct settings for a Maxtor 25128 AT 2.5"  HD?  Can the
> >>AMAX partition have different settings frmo the other partns?  Maxtor
> >>could only tell me the drive as a 32K cache but they didn't know if that
> >>would play into either Amiga 1200 setting.

> >And there exactly, is the answer Jill.  A 32k cache.  Translated to the
> >correct "MaxTransfer", this is 0x7fe0, one sector less than the cache
> >size.  Its also exactly what I said last night, but with the actual
> >figure now filled in.
> <snip>

> Since when has MaxTransfer had *anything* to do with the cache size?

Never has but some people claim to see some significance to it. It MAY be
safest to plck the smaller of cache size or 130560 bytes to avoid both counter
problems and cache problems. On VERY rare VERY old IDEs some of the
intermal CPUs were 15 bit beasties. For then you want to use 65024 bytes.


 
 
 

Mask vs. MaxTransfer 1200 IDE

Post by Andrew Davidso » Sat, 18 Apr 1998 04:00:00



Quote:> : Since when has MaxTransfer had *anything* to do with the cache size?
> When the drive has firmware bugs.  A *correct* implementation doesn't
> require the software to be aware of the cache size.

These old Maxtor are infamous for the incredibly dodgy firmware.
Probably the worst drive I've seen is the Maxtor 71084, the 1Gb version of
the old 7000 series.  Its impossible to get it to work properly - a
complete disaster.

* Please note *new* E-Mail address! My old Demon address terminates soon! *
--


--

 
 
 

Mask vs. MaxTransfer 1200 IDE

Post by Gary Colvill » Sat, 18 Apr 1998 04:00:00



Quote:> These old Maxtor are infamous for the incredibly dodgy firmware.

Ever tried an old Conner drive? I have a pair of Conner CP3040 drives here
that will not even take an RDB. They just seem to be totally incapable of
working with an Amiga no matter what :(

Quote:> Probably the worst drive I've seen is the Maxtor 71084, the 1Gb
> version of the old 7000 series.  Its impossible to get it to work
> properly - a complete disaster.

The first hard drive I ever used on my A1200 was a Maxtor 7040-III. Lovely
drive it was, once you got round the MaxTransfer problem :)

--  _
 _ //    Gary Colville        AmigaOS Software Developer          

--
Monsoon - the E-Mail/News/Fido reader  [http://www.empire.u-net.com]
--

 
 
 

Mask vs. MaxTransfer 1200 IDE

Post by Joanne Do » Sun, 19 Apr 1998 04:00:00




> > These old Maxtor are infamous for the incredibly dodgy firmware.

> Ever tried an old Conner drive? I have a pair of Conner CP3040 drives here
> that will not even take an RDB. They just seem to be totally incapable of
> working with an Amiga no matter what :(

> > Probably the worst drive I've seen is the Maxtor 71084, the 1Gb
> > version of the old 7000 series.  Its impossible to get it to work
> > properly - a complete disaster.

> The first hard drive I ever used on my A1200 was a Maxtor 7040-III. Lovely
> drive it was, once you got round the MaxTransfer problem :)

It is not confined to IDE, either. Note that you CANNOT do a raw diskcopy
to a Quantum drive. If you write multiple blocks starting at block zero the ONLY
way I have found to bring the drive back to life is a low level format. They's
bad boys for some reason. Once you get past the first few blocks things work
just peachy, though.


    mfg these days. Asian knowhow and experience and management
    killed the quality of the product before the product line died.

 
 
 

Mask vs. MaxTransfer 1200 IDE

Post by Andrew Davidso » Sun, 19 Apr 1998 04:00:00



Quote:>> These old Maxtor are infamous for the incredibly dodgy firmware.
> Ever tried an old Conner drive? I have a pair of Conner CP3040 drives
> here that will not even take an RDB. They just seem to be totally
> incapable of working with an Amiga no matter what :(

I have a Conner here of that same era - a CP3000 40Mb job.  It won't even
run sideways.  I also have 6 IBM WDA-L42 drives, from the same era as the
Conner, and they also refuse to accept a RDB.  They make excellent door
stops, power supply testers, etc, etc.

Quote:> The first hard drive I ever used on my A1200 was a Maxtor 7040-III.
> Lovely drive it was, once you got round the MaxTransfer problem :)

Heh, heh.  As no doubt you know, I have great results from my old 7245,
even with its 500k/sec transfer rate.

* Please note *new* E-Mail address! My old Demon address terminates soon! *
--


--

 
 
 

1. A4000 IDE MaxTransfer and Mask settings?

I have two Western Digital 1.25 gig drives.. right now  I have
my maxtransfer set to 1xfe00, with a mask of 7ffffffe.
Can anyone tell me a  better (faster) maxtrasnfer to set
with HDToolBox?  I've tried other #'s but noticed no
differences in speed...

Thanks... apreciate any email..

--

----

 "Nurses heal the world"       -   A Mac Quadra 900 for $35.
                                    Only Amiga makes it possible...
                               -   "... and though I don't know
                                    where, and don't know when I'll
                                    find myself in love again, I
                                    promise : there will always be
                                    a little place no one will see,
                                    a tiny part, deep in my heart,
                                    that stays in love with you ..."

2. Help! Moving a cursor

3. Mask and MaxTransfer

4. CView and CDialog question

5. Removable SCSI Disks and Universal Mask/MaxTransfer Problems?

6. CFP: COLING-2002: 19th International Conf. on Computational Linguistics

7. Here we go again: Mask & MaxTransfer settings

8. w2k uses pdc ONLY as logon server!??!

9. MaxTransfer and Mask

10. Ancora su MaxTransfer e Mask

11. Mask/MaxTransfer, who does it?

12. A3000: which Mask/Maxtransfer values best?

13. MaxTransfer & Mask