Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]

Groups > comp.os.linux.hardware > #2570 > unrolled thread

Bad Hitachi disk.... how so from a pic?

Back to article view | Back to comp.os.linux.hardware

Contents

  Bad Hitachi disk.... how so from a pic? Jax <remove.bear.bottoms1@gmail.com> - 2014-12-06 14:44 +0000
    Re: Bad Hitachi disk.... how so from a pic? philo  <philo@privacy.net> - 2014-12-06 09:44 -0600
      Re: Bad Hitachi disk.... how so from a pic? notbob <notbob@nothome.com> - 2014-12-06 19:37 +0000
        Re: Bad Hitachi disk.... how so from a pic? philo  <philo@privacy.net> - 2014-12-06 14:54 -0600
        Re: Bad Hitachi disk.... how so from a pic? Jax <remove.bear.bottoms1@gmail.com> - 2014-12-06 22:50 +0000
          Re: Bad Hitachi disk.... how so from a pic? notbob <notbob@nothome.com> - 2014-12-06 23:13 +0000
        Re: Bad Hitachi disk.... how so from a pic? philo  <philo@privacy.net> - 2014-12-07 04:57 -0600
          Re: Bad Hitachi disk.... how so from a pic? Aragorn <thorongil@telenet.be.invalid> - 2014-12-07 16:52 +0100
            Re: Bad Hitachi disk.... how so from a pic? philo  <philo@privacy.net> - 2014-12-07 09:59 -0600
              Re: Bad Hitachi disk.... how so from a pic? Aragorn <thorongil@telenet.be.invalid> - 2014-12-07 17:10 +0100
                Re: Bad Hitachi disk.... how so from a pic? philo  <philo@privacy.net> - 2014-12-07 10:48 -0600
                  Re: Bad Hitachi disk.... how so from a pic? Aragorn <thorongil@telenet.be.invalid> - 2014-12-07 18:37 +0100
                    Re: Bad Hitachi disk.... how so from a pic? philo  <philo@privacy.net> - 2014-12-07 12:28 -0600
                    Re: Bad Hitachi disk.... how so from a pic? Pascal Hambourg <boite-a-spam@plouf.fr.eu.org> - 2014-12-07 21:46 +0100
                      Re: Bad Hitachi disk.... how so from a pic? Aragorn <thorongil@telenet.be.invalid> - 2014-12-07 22:27 +0100
                        Re: Bad Hitachi disk.... how so from a pic? Pascal Hambourg <boite-a-spam@plouf.fr.eu.org> - 2014-12-08 00:36 +0100
                          Re: Bad Hitachi disk.... how so from a pic? Richard Kettlewell <rjk@greenend.org.uk> - 2014-12-08 09:26 +0000
                            Re: Bad Hitachi disk.... how so from a pic? Aragorn <thorongil@telenet.be.invalid> - 2014-12-08 11:31 +0100
                              Re: Bad Hitachi disk.... how so from a pic? Richard Kettlewell <rjk@greenend.org.uk> - 2014-12-08 10:53 +0000
                                Re: Bad Hitachi disk.... how so from a pic? Aragorn <thorongil@telenet.be.invalid> - 2014-12-08 17:48 +0100
                                  Re: Bad Hitachi disk.... how so from a pic? Richard Kettlewell <rjk@greenend.org.uk> - 2014-12-08 17:28 +0000
                              Re: Bad Hitachi disk.... how so from a pic? Robert Nichols <SEE_SIGNATURE@localhost.localdomain.invalid> - 2014-12-08 08:43 -0600
                              Re: Bad Hitachi disk.... how so from a pic? Pascal Hambourg <boite-a-spam@plouf.fr.eu.org> - 2014-12-09 11:02 +0100
                              Re: Bad Hitachi disk.... how so from a pic? scott@alfter.diespammersdie.us (Scott Alfter) - 2014-12-09 23:35 +0000
                                Re: Bad Hitachi disk.... how so from a pic? philo  <philo@privacy.net> - 2014-12-09 17:43 -0600
                                  Re: Bad Hitachi disk.... how so from a pic? philo  <philo@privacy.net> - 2014-12-09 17:54 -0600
                                  Re: Bad Hitachi disk.... how so from a pic? scott@alfter.diespammersdie.us (Scott Alfter) - 2014-12-10 20:22 +0000
                                  Re: Bad Hitachi disk.... how so from a pic? scott@alfter.diespammersdie.us (Scott Alfter) - 2014-12-10 20:27 +0000
                                    Re: Bad Hitachi disk.... how so from a pic? philo  <philo@privacy.net> - 2014-12-10 16:40 -0600
                                Re: Bad Hitachi disk.... how so from a pic? Thomas Richter <thor@math.tu-berlin.de> - 2014-12-10 11:06 +0100
    Re: Bad Hitachi disk.... how so from a pic? Baho Utot <baho-utot@invalid.com> - 2014-12-06 13:39 -0500
      Re: Bad Hitachi disk.... how so from a pic? Jax <remove.bear.bottoms1@gmail.com> - 2014-12-06 22:51 +0000
    Re: Bad Hitachi disk.... how so from a pic? Johnny <johnny@invalid.net> - 2014-12-06 13:02 -0600
    Re: Bad Hitachi disk.... how so from a pic? Keith Keller <kkeller-usenet@wombat.san-francisco.ca.us> - 2014-12-06 16:04 -0800

Page 1 of 2  [1] 2  Next page →

#2570 — Bad Hitachi disk.... how so from a pic?

Someone I know who says he knows about computer hardware says
"Hitachi disks are sh*t". As proof he posted this picture of two
Hitachi disks in another thread here. 

   http://osama-is-dead.net/images/bad_hitachi.jpg 

Is he right? How can a picture of a disk can show it is bad? 

-- 
Jax        

[toc] | [next] | [standalone]

#2572

On 12/06/2014 08:44 AM, Jax wrote:
> Someone I know who says he knows about computer hardware says
> "Hitachi disks are sh*t". As proof he posted this picture of two
> Hitachi disks in another thread here.
>
>     http://osama-is-dead.net/images/bad_hitachi.jpg
>
> Is he right? How can a picture of a disk can show it is bad?
>




Hitachi manufactured the IBM Deskstar which was later nicknamed the 
"death star"


If it was one of that series I would definitely not use it.


The drive on the right is simply labeled "Hitachi" and since it's not 
from the Deskstar series is /possibly/ OK


The one on the left is from their newer HGST production
and from the data I saw are considered reliable.

[toc] | [prev] | [next] | [standalone]

#2575

On 2014-12-06, philo  <philo@privacy.net> wrote:

> Hitachi manufactured the IBM Deskstar which was later nicknamed the 
> "death star"
>
>
> If it was one of that series I would definitely not use it.

The "deathstar" was one of IBM's few failures.  It was the model 75GXP
Deskstar (6 capacity variants) and was only one of many Deskstars made
by pre-Hitachi IBM.  I had 2 or 3 earlier Deskstars and they always
worked flawlessly, but the mythology of those few "deathstars" were
devastating to IBM's hard earned hardware rep.

https://en.wikipedia.org/wiki/Hitachi_Deskstar

nb

[toc] | [prev] | [next] | [standalone]

#2576

On 12/06/2014 01:37 PM, notbob wrote:
> On 2014-12-06, philo  <philo@privacy.net> wrote:
>
>> Hitachi manufactured the IBM Deskstar which was later nicknamed the
>> "death star"
>>
>>
>> If it was one of that series I would definitely not use it.
>
> The "deathstar" was one of IBM's few failures.  It was the model 75GXP
> Deskstar (6 capacity variants) and was only one of many Deskstars made
> by pre-Hitachi IBM.  I had 2 or 3 earlier Deskstars and they always
> worked flawlessly, but the mythology of those few "deathstars" were
> devastating to IBM's hard earned hardware rep.
>
> https://en.wikipedia.org/wiki/Hitachi_Deskstar
>
> nb
>



I figure that by now all the bad ones should be long gone...though I 
might just have one in my junkbox somewhere. If I do, I'd never put it 
in a machine that was going to anyone.

[toc] | [prev] | [next] | [standalone]

#2577

notbob <notbob@nothome.com> wrote in
news:ceh481Frsl6U7@mid.individual.net: 

> On 2014-12-06, philo  <philo@privacy.net> wrote:
> 
>> Hitachi manufactured the IBM Deskstar which was later nicknamed
>> the "death star"
>>
>>
>> If it was one of that series I would definitely not use it.
> 
> The "deathstar" was one of IBM's few failures.  It was the model
> 75GXP Deskstar (6 capacity variants) and was only one of many
> Deskstars made by pre-Hitachi IBM.  I had 2 or 3 earlier Deskstars
> and they always worked flawlessly, but the mythology of those few
> "deathstars" were devastating to IBM's hard earned hardware rep.
> 
> https://en.wikipedia.org/wiki/Hitachi_Deskstar

Notbob..... what's so special about this pic that it can claim to
show the Hitachi drives are bad? 

http://osama-is-dead.net/images/bad_hitachi.jpg 

-- 
Jax        

[toc] | [prev] | [next] | [standalone]

#2579

On 2014-12-06, Jax <remove.bear.bottoms1@gmail.com> wrote:

> Notbob..... what's so special about this pic that it can claim to
> show the Hitachi drives are bad? 
>
> http://osama-is-dead.net/images/bad_hitachi.jpg 

The dented cover to the center spindle of HDD?

Got me.  They're both post-IBM, which it was IBM who created/sold the
"deathstar".  I notice both are 7200 rpm, the speed the deathstars
were attaining.  

nb

[toc] | [prev] | [next] | [standalone]

#2582

On 12/06/2014 01:37 PM, notbob wrote:
> On 2014-12-06, philo  <philo@privacy.net> wrote:
>
>> Hitachi manufactured the IBM Deskstar which was later nicknamed the
>> "death star"
>>
>>
>> If it was one of that series I would definitely not use it.
>
> The "deathstar" was one of IBM's few failures.  It was the model 75GXP
> Deskstar (6 capacity variants) and was only one of many Deskstars made
> by pre-Hitachi IBM.  I had 2 or 3 earlier Deskstars and they always
> worked flawlessly, but the mythology of those few "deathstars" were
> devastating to IBM's hard earned hardware rep.
>
> https://en.wikipedia.org/wiki/Hitachi_Deskstar
>
> nb
>



I had a look in my "junkbox" and there is a Hitachi Deskstar there.

It's not an IBM. Does anyone know if this could be one of the "bad" ones 
or was that strictly IBM?


It's got a very old backup on it and is not critical

[toc] | [prev] | [next] | [standalone]

#2584

On Sunday 07 December 2014 11:57, philo  conveyed the following to 
comp.os.linux.hardware...

> I had a look in my "junkbox" and there is a Hitachi Deskstar there.
> 
> It's not an IBM. Does anyone know if this could be one of the "bad"
> ones or was that strictly IBM?

To the best of my knowledge, the bad DeskStars were all manufactured by 
IBM, and IBM even corrected the design before they sold their hard disk 
drive technology to Hitachi.

For what it's worth, I have used the IBM UltraStar hard disks and their 
Hitachi UltraStar successors.  Now, those are enterprise-grade SCSI and 
SAS disks, so their build quality is much higher than that of consumer-
grade disks, but I've only ever had one of those ─ an IBM, not a Hitachi 
─ fail on me... after a *very, very* long time.  

All hard disks do eventually fail, so I don't think it was all that 
exceptional, especially given the mileage on that thing.

> It's got a very old backup on it and is not critical

It probably still works fine. ;-)

-- 
= Aragorn =

         http://www.linuxcounter.net - registrant #223157

[toc] | [prev] | [next] | [standalone]

#2585

On 12/07/2014 09:52 AM, Aragorn wrote:
> On Sunday 07 December 2014 11:57, philo  conveyed the following to
> comp.os.linux.hardware...
>
>> I had a look in my "junkbox" and there is a Hitachi Deskstar there.
>>
>> It's not an IBM. Does anyone know if this could be one of the "bad"
>> ones or was that strictly IBM?
>
> To the best of my knowledge, the bad DeskStars were all manufactured by
> IBM, and IBM even corrected the design before they sold their hard disk
> drive technology to Hitachi.
>
> For what it's worth, I have used the IBM UltraStar hard disks and their
> Hitachi UltraStar successors.  Now, those are enterprise-grade SCSI and
> SAS disks, so their build quality is much higher than that of consumer-
> grade disks, but I've only ever had one of those ─ an IBM, not a Hitachi
> ─ fail on me... after a *very, very* long time.
>
> All hard disks do eventually fail, so I don't think it was all that
> exceptional, especially given the mileage on that thing.
>
>> It's got a very old backup on it and is not critical
>
> It probably still works fine. ;-)
>



Thanks for the info.

BTW: I do have a few SCSI 'ultra-wides' but they are only 10 gig
and thus far have only been able to find one project where I could use it.

[toc] | [prev] | [next] | [standalone]

#2586

On Sunday 07 December 2014 16:59, philo  conveyed the following to 
comp.os.linux.hardware...

> On 12/07/2014 09:52 AM, Aragorn wrote:
>
>> On Sunday 07 December 2014 11:57, philo  conveyed the following to
>> comp.os.linux.hardware...
>>
>>> I had a look in my "junkbox" and there is a Hitachi Deskstar there.
>>>
>>> It's not an IBM. Does anyone know if this could be one of the "bad"
>>> ones or was that strictly IBM?
>>
>> To the best of my knowledge, the bad DeskStars were all manufactured
>> by IBM, and IBM even corrected the design before they sold their hard
>> disk drive technology to Hitachi.
>>
>> For what it's worth, I have used the IBM UltraStar hard disks and
>> their Hitachi UltraStar successors.  Now, those are enterprise-grade
>> SCSI and SAS disks, so their build quality is much higher than that
>> of consumer- grade disks, but I've only ever had one of those ─ an
>> IBM, not a Hitachi ─ fail on me... after a *very, very* long time.
>>
>> All hard disks do eventually fail, so I don't think it was all that
>> exceptional, especially given the mileage on that thing.
>>
>>> It's got a very old backup on it and is not critical
>>
>> It probably still works fine. ;-)
> 
> Thanks for the info.
> 
> BTW: I do have a few SCSI 'ultra-wides' but they are only 10 gig
> and thus far have only been able to find one project where I could use
> it.

I've got a whole box of anything from Ultra Wide to Ultra 320 lying 
around, and in capacities from 9 GiB to 36 GiB.  Some of those came out 
of a machine of my own, and others came out of a now defunct Dell 
PowerEdge server that we were using back when we were still running an 
IRC network.  The ones from the Dell were installed in hardware RAID 
configurations ─ RAID 5 and RAID 10.

Some of those disks may be defective but I don't know which ones.  My 
colleague lumped them all together into a single box.  There's about six 
or eight of them.  And they're all IBM or Hitachi UltraStars. ;-)

-- 
= Aragorn =

         http://www.linuxcounter.net - registrant #223157

[toc] | [prev] | [next] | [standalone]

#2587

On 12/07/2014 10:10 AM, Aragorn wrote:

<snip>
>>>
>>> It probably still works fine. ;-)
>>
>> Thanks for the info.
>>
>> BTW: I do have a few SCSI 'ultra-wides' but they are only 10 gig
>> and thus far have only been able to find one project where I could use
>> it.
>
> I've got a whole box of anything from Ultra Wide to Ultra 320 lying
> around, and in capacities from 9 GiB to 36 GiB.  Some of those came out
> of a machine of my own, and others came out of a now defunct Dell
> PowerEdge server that we were using back when we were still running an
> IRC network.  The ones from the Dell were installed in hardware RAID
> configurations ─ RAID 5 and RAID 10.
>
> Some of those disks may be defective but I don't know which ones.  My
> colleague lumped them all together into a single box.  There's about six
> or eight of them.  And they're all IBM or Hitachi UltraStars. ;-)
>



A lot of people in town know that I'll take just about any computer 
parts they do not want. No CRT's though.

Sometimes I come home and just find stuff on my porch.
That's how I got the ultra-wides and scsi controllers.


I tested all of them and took the magnets out of some of the bad ones.
Those are the strongest magnets I've ever seen.

The one drive and controller I used in a Linux machine I setup for the 
place where I do volunteer work. The machine is an early p-4 with a 
minimal amount of RAM, but with the "ultra-wide" drive performs pretty well.

[toc] | [prev] | [next] | [standalone]

#2588

On Sunday 07 December 2014 17:48, philo  conveyed the following to 
comp.os.linux.hardware...

> On 12/07/2014 10:10 AM, Aragorn wrote:
> 
>> I've got a whole box of anything from Ultra Wide to Ultra 320 lying
>> around, and in capacities from 9 GiB to 36 GiB.  Some of those came
>> out of a machine of my own, and others came out of a now defunct Dell
>> PowerEdge server that we were using back when we were still running
>> an IRC network.  The ones from the Dell were installed in hardware
>> RAID configurations ─ RAID 5 and RAID 10.
>>
>> Some of those disks may be defective but I don't know which ones.  My
>> colleague lumped them all together into a single box.  There's about
>> six or eight of them.  And they're all IBM or Hitachi UltraStars. ;-)
> 
> A lot of people in town know that I'll take just about any computer
> parts they do not want. No CRT's though.
> 
> Sometimes I come home and just find stuff on my porch.
> That's how I got the ultra-wides and scsi controllers.

Well, as it happens to be, we're on opposite sides of the Big Pond, and 
I know for a fact that shipping something to the US is going to be very 
expensive ─ likewise for receiving anything from the US, due to customs 
fees.  I'm afraid my current financial status doesn't allow for it, or 
else I would gladly be shipping the stuff to you.

> I tested all of them and took the magnets out of some of the bad ones.
> Those are the strongest magnets I've ever seen.

Those are still low capacity disks compared to what's mainstream now, 
which means that there was much more of a clearing between the heads and 
the platters in those days, and more space between the sectors and 
cylinders.  So those magnets would indeed have been more powerful than 
the ones in use today.

> The one drive and controller I used in a Linux machine I setup for the
> place where I do volunteer work. The machine is an early p-4 with a
> minimal amount of RAM, but with the "ultra-wide" drive performs pretty
> well.

The thing with SCSI is that it requires a SCSI controller, and then the 
controller takes care of all the head positioning etc.  With IDE 
(PATA/SATA), all of that has to be executed by the CPU on the mainboard.  
This made SCSI drives much faster and snappier in the experience, even 
long after IDE had already surpassed the throughput capabilities of SCSI 
and became UDMA-capable.  So I'm not surprised that it performs very 
well on your old machine; there is less work for the main processor to 
waste CPU cycles on, compared to IDE (PATA/SATA).

The main reason why there is a tendency now towards SATA disks even for 
enterprise solutions is that SATA disks are significantly cheaper than 
SCSI or SAS; you get to pay a much lower price per MiB with SATA than 
from SCSI/SAS.  Still, there still are enterprise environments where SAS 
has preference over SATA, for instance in storage area networks ─ 
parallel SCSI is considered way outdated now.

-- 
= Aragorn =

         http://www.linuxcounter.net - registrant #223157

[toc] | [prev] | [next] | [standalone]

#2589

On 12/07/2014 11:37 AM, Aragorn wrote:
>

<snip>
>
> Well, as it happens to be, we're on opposite sides of the Big Pond, and
> I know for a fact that shipping something to the US is going to be very
> expensive ─ likewise for receiving anything from the US, due to customs
> fees.  I'm afraid my current financial status doesn't allow for it, or
> else I would gladly be shipping the stuff to you.
>

Thanks. As long as you are offering them /free/ I'll be right over to 
pick them up!


I do sometimes order stuff from overseas if it does not weigh much.

A few months back I ran out of space CMOS batteries and ran over to the 
local drug store where I figured they be 2 or 3 USD. Heck , they were 7 
USD each....absurd.

I ordered 20 of them from China at a total of 3.58 USD and that 
/included/ shipping!

>> I tested all of them and took the magnets out of some of the bad ones.
>> Those are the strongest magnets I've ever seen.
>
> Those are still low capacity disks compared to what's mainstream now,
> which means that there was much more of a clearing between the heads and
> the platters in those days, and more space between the sectors and
> cylinders.  So those magnets would indeed have been more powerful than
> the ones in use today.

If I put two of them together I needed to put one in a vice to get the 
other one off. Had to twist it off with a pliers

>
>> The one drive and controller I used in a Linux machine I setup for the
>> place where I do volunteer work. The machine is an early p-4 with a
>> minimal amount of RAM, but with the "ultra-wide" drive performs pretty
>> well.
>
> The thing with SCSI is that it requires a SCSI controller, and then the
> controller takes care of all the head positioning etc.  With IDE
> (PATA/SATA), all of that has to be executed by the CPU on the mainboard.
> This made SCSI drives much faster and snappier in the experience, even
> long after IDE had already surpassed the throughput capabilities of SCSI
> and became UDMA-capable.  So I'm not surprised that it performs very
> well on your old machine; there is less work for the main processor to
> waste CPU cycles on, compared to IDE (PATA/SATA).
>
> The main reason why there is a tendency now towards SATA disks even for
> enterprise solutions is that SATA disks are significantly cheaper than
> SCSI or SAS; you get to pay a much lower price per MiB with SATA than
> from SCSI/SAS.  Still, there still are enterprise environments where SAS
> has preference over SATA, for instance in storage area networks ─
> parallel SCSI is considered way outdated now.


Yes. even though I always build my machines from "old junk" at least I 
am now using SATA drives
>

[toc] | [prev] | [next] | [standalone]

#2590

Aragorn a écrit :
> 
> The thing with SCSI is that it requires a SCSI controller, and then the 
> controller takes care of all the head positioning etc.  With IDE 
> (PATA/SATA), all of that has to be executed by the CPU on the mainboard.

Neither the SCSI host controller nor the host CPU care about head
positioning. They don't even know how many physical heads the disk has.
That's all the task of the disk embedded controller.

[toc] | [prev] | [next] | [standalone]

#2591

On Sunday 07 December 2014 21:46, Pascal Hambourg conveyed the following 
to comp.os.linux.hardware...

> Aragorn a écrit :
>> 
>> The thing with SCSI is that it requires a SCSI controller, and then
>> the controller takes care of all the head positioning etc.  With IDE
>> (PATA/SATA), all of that has to be executed by the CPU on the
>> mainboard.
> 
> Neither the SCSI host controller nor the host CPU care about head
> positioning. They don't even know how many physical heads the disk
> has. That's all the task of the disk embedded controller.

You are right ─ I was miswording what I actually meant  In the event of 
a machine with an ATA hard disk, almost everything on account of disk 
I/O is handled by the CPU on the mainboard, while in the event of a SCSI 
controller, the operating system offloads a great deal of those low-
level tasks onto the controller itself, which (of course) has its own 
dedicated (and typically specialized) processor(s).  (Some Adaptec 
SAS/SCSI/SATA RAID adapters for instance run their own local embedded 
Linux-based system via a SoC processor.)

Of course, things like filesystem layout and block/extent addressing are 
still handled by the filesystem driver in the operating system kernel, 
and thus take up a certain percentage of normal mainboard CPU time.

-- 
= Aragorn =

         http://www.linuxcounter.net - registrant #223157

[toc] | [prev] | [next] | [standalone]

#2592

Aragorn a écrit :
> 
> In the event of 
> a machine with an ATA hard disk, almost everything on account of disk 
> I/O is handled by the CPU on the mainboard, while in the event of a SCSI 
> controller, the operating system offloads a great deal of those low-
> level tasks onto the controller itself, which (of course) has its own 
> dedicated (and typically specialized) processor(s).

Is this still true with modern AHCI SATA host controllers ?
What are these low-level tasks that are performed by a SCSI host
controller but not by a SATA host controller ?

[toc] | [prev] | [next] | [standalone]

#2593

Pascal Hambourg <boite-a-spam@plouf.fr.eu.org> writes:
> Aragorn a écrit :

>> In the event of a machine with an ATA hard disk, almost everything on
>> account of disk I/O is handled by the CPU on the mainboard, while in
>> the event of a SCSI controller, the operating system offloads a great
>> deal of those low- level tasks onto the controller itself, which (of
>> course) has its own dedicated (and typically specialized)
>> processor(s).
>
> Is this still true with modern AHCI SATA host controllers ?  What are
> these low-level tasks that are performed by a SCSI host controller but
> not by a SATA host controller ?

AFAIK it hasn’t been true since the 1980s.  The “integrated” in
Integrated Drive Electronics (i.e. the original name for PATA) refers to
the integration of the controller into the drive.

-- 
http://www.greenend.org.uk/rjk/

[toc] | [prev] | [next] | [standalone]

#2594

On Monday 08 December 2014 10:26, Richard Kettlewell conveyed the 
following to comp.os.linux.hardware...

> Pascal Hambourg <boite-a-spam@plouf.fr.eu.org> writes:
>> Aragorn a écrit :
> 
>>> In the event of a machine with an ATA hard disk, almost everything
>>> on account of disk I/O is handled by the CPU on the mainboard, while
>>> in the event of a SCSI controller, the operating system offloads a
>>> great deal of those low- level tasks onto the controller itself,
>>> which (of course) has its own dedicated (and typically specialized)
>>> processor(s).
>>
>> Is this still true with modern AHCI SATA host controllers ?  What are
>> these low-level tasks that are performed by a SCSI host controller
>> but not by a SATA host controller ?
> 
> AFAIK it hasn’t been true since the 1980s.  The “integrated” in
> Integrated Drive Electronics (i.e. the original name for PATA) refers
> to the integration of the controller into the drive.

No, not the controller.  That which is integrated into the drive are the 
most elementary drive electronics ─ voltage regulators and such ─ which 
is not the same thing as an actual disk controller.  For that matter, 
SCSI disks are IDE disks too, and yet they require a separate controller 
─ whether installed on the motherboard or on a plug-in adapter card.  

Command queuing for instance is a controller task, and when it comes to 
IDE disks, the kernel will take care of that.  When it comes to SCSI or 
SAS, the controller card does all of that, leaving the mainboard CPU 
free to do other things.

A major breakthrough in performance for IDE disks happened when they 
became UDMA-capable, though, and the bandwidth of modern SATA disks is 
about the same as for SAS.  However, you only seriously notice the 
difference between SCSI/SAS on the one hand and SATA/PATA on the other 
hand when there is a lot of very intense disk I/O going on.  With a SATA 
or PATA disk, the machine will become less responsive.  With SCSI or 
SAS, it won't ─ again, because a lot of the activity is offloaded onto 
the SCSI/SAS controller.

-- 
= Aragorn =

         http://www.linuxcounter.net - registrant #223157

[toc] | [prev] | [next] | [standalone]

#2595

Aragorn <thorongil@telenet.be.invalid> writes:
> Richard Kettlewell wrote:
>> AFAIK it hasn’t been true since the 1980s.  The “integrated” in
>> Integrated Drive Electronics (i.e. the original name for PATA) refers
>> to the integration of the controller into the drive.
>
> No, not the controller.  That which is integrated into the drive are the 
> most elementary drive electronics ─ voltage regulators and such ─ which 
> is not the same thing as an actual disk controller.

Obviously you’re guessing.  Have a look at the ATA-1 spec, the commands
are things like ‘read sector’ and ‘write sector’, not ‘move disk head’.

-- 
http://www.greenend.org.uk/rjk/

[toc] | [prev] | [next] | [standalone]

#2599

On Monday 08 December 2014 11:53, Richard Kettlewell conveyed the 
following to comp.os.linux.hardware...

> Aragorn <thorongil@telenet.be.invalid> writes:
>
>> Richard Kettlewell wrote:
>>
>>> AFAIK it hasn’t been true since the 1980s.  The “integrated” in
>>> Integrated Drive Electronics (i.e. the original name for PATA)
>>> refers to the integration of the controller into the drive.
>>
>> No, not the controller.  That which is integrated into the drive are
>> the most elementary drive electronics ─ voltage regulators and such ─
>> which is not the same thing as an actual disk controller.
> 
> Obviously you’re guessing.  Have a look at the ATA-1 spec, the
> commands are things like ‘read sector’ and ‘write sector’, not ‘move
> disk head’.

I did say that I misworded that part.  And yes, I don't know *all* the 
details, but I wasn't aware that this was a requirement for me to be 
allowed to speak about the differences between SCSI and ATA on Usenet.

-- 
= Aragorn =

         http://www.linuxcounter.net - registrant #223157

[toc] | [prev] | [next] | [standalone]

Page 1 of 2  [1] 2  Next page →

Back to top | Article view | comp.os.linux.hardware

csiph-web