UDMA33 hard drive runs at 6 megs/second - WTF?
Gene Heskett
Having just built this system, using a TYAN S1590S board, bios 1.14, and
buying a UDMA33 capable drive, also according to the docs, PIO modes 3
and 4 are supported.
Under RH linux 5.2, original kernel 2.0.36 it default powerups at 4.75
megs/second. Regardless of what I do in the bios, or with hdparm (many
of the operations are 'not supported'), the best I can get it to do is
about 6.75 megs/second.
The drive is a WDC AC14300R, FwRev 15.01J15
Is there any hope of ever getting this thing to move data with a bigger
spoon?
Cheers, Gene
--
Gene Heskett, CET, UHK |Amiga A2k Zeus040 50 megs fast/2 megs chip
Ch. Eng. @ WDTV-5 |A2091,GuruRom,1g Seagate,CDROM,Multiface III
<gene_h...@iolinc.net> or |Buddha + 4 gig WDC drive, 525 meg tape
<gene_h...@westvirginia.net>|Stylus Pro, EnPrint, Picasso-II, 17" vga
RC5-Moo! 22kkeys/sec isn't much, but it all helps
--
The Everette's
for?
Even the Seagate Cheetah U2 is only capable of 18 megs\sec sustained. the 33
I think you expected is the controller's max transfer rate, not the Hard
Drives.
Gene Heskett <gene_h...@iolinc.net> wrote in message
news:943.803T22...@iolinc.net...
Philip Morris
wrote:
>Reply to: <gene_h...@iolinc.net>
>
>Having just built this system, using a TYAN S1590S board, bios 1.14, and
>buying a UDMA33 capable drive, also according to the docs, PIO modes 3
>and 4 are supported.
>
>Under RH linux 5.2, original kernel 2.0.36 it default powerups at 4.75
>megs/second. Regardless of what I do in the bios, or with hdparm (many
>of the operations are 'not supported'), the best I can get it to do is
>about 6.75 megs/second.
>
>The drive is a WDC AC14300R, FwRev 15.01J15
>
>Is there any hope of ever getting this thing to move data with a bigger
>spoon?
>
>Cheers, Gene
>--
> Gene Heskett, CET, UHK |Amiga A2k Zeus040 50 megs fast/2 megs chip
> Ch. Eng. @ WDTV-5 |A2091,GuruRom,1g Seagate,CDROM,Multiface III
><gene_h...@iolinc.net> or |Buddha + 4 gig WDC drive, 525 meg tape
><gene_h...@westvirginia.net>|Stylus Pro, EnPrint, Picasso-II, 17" vga
> RC5-Moo! 22kkeys/sec isn't much, but it all helps
Gene,
Make sure you apply the utility found on WD's site to back the drive
back down to UDMA33. The drive you have is a UDMA66 drive and will
only run in PIO mode on your motherboard unless you run the utility.
If you ever get a board that supports UDMA66, you can run the utility
again to switch it back.
Gene Heskett
Gene Heskett sends Greetings to Philip Morris;
PM> On 14 May 99 08:27:45 -0500, "Gene Heskett" <gene_h...@iolinc.net>
PM> wrote:
>>Reply to: <gene_h...@iolinc.net>
>>
>>Having just built this system, using a TYAN S1590S board, bios 1.14, and
>>buying a UDMA33 capable drive, also according to the docs, PIO modes 3 and 4
>>are supported.
>>
>>Under RH linux 5.2, original kernel 2.0.36 it default powerups at 4.75
>>megs/second. Regardless of what I do in the bios, or with hdparm (many
>>of the operations are 'not supported'), the best I can get it to do is
>>about 6.75 megs/second.
>>
>>The drive is a WDC AC14300R, FwRev 15.01J15
>>
>>Is there any hope of ever getting this thing to move data with a bigger
>>spoon?
[...]
PM> Gene,
PM> Make sure you apply the utility found on WD's site to back the drive
PM> back down to UDMA33. The drive you have is a UDMA66 drive and will
PM> only run in PIO mode on your motherboard unless you run the utility.
PM> If you ever get a board that supports UDMA66, you can run the utility
PM> again to switch it back.
Unforch, the disk I got with the drive on the second shipment (the first
was diskless, and DOA) seems to be incapable of being run from a dos
boot, appearing to be a windoze only diskette. And the *only* windows
is X here.
I just rebooted it to a dos6.21 state, and the only thing that seems to
want to run is EZ.EXE. And the first thing it wants to do is format my
drive so it can install itself, and there is apparently no way to bypass
the EZ Install and actually get to the disk tuning utilities it *might*
be able to do. Format my drive, just so it can install itself? Not
bloody likely.
Probably stupid question, but it doesn't appear that TYAN's bios even
offers the 66mhz option, and that was at the time, one of the reasons I
bought the S1590S board.
Is there any other way around this pile of brush? I'm trying to buy a
scsi card, which should have been here Wednesday, but on checking as to
why it wasn't here yet on Friday, somebody missed the 3 day air orders
on the shipping, so its in the UPS warehouse in CACH, IL as of noon
today.
Once that arrives, then I can make backups. Then I can play...
Cheers, Gene
--
Gene Heskett, CET, UHK |Amiga A2k Zeus040 50 megs fast/2 megs chip
Ch. Eng. @ WDTV-5 |A2091,GuruRom,1g Seagate,CDROM,Multiface III
<gene_h...@iolinc.net> or |Buddha + 4 gig WDC drive, 525 meg tape
<gene_h...@westvirginia.net>|Stylus Pro, EnPrint, Picasso-II, 17" vga
RC5-Moo! 22kkeys/sec isn't much, but it all helps
--
Philip Morris
wrote:
>Reply to: <engin...@wdtv.com>
Gene,
Go to WD's web site and look for the utility "WD ATA66". It does not
attempt to format your hard drive, it simply switches it to an ATA33
compliant drive.
The 1590 does not support ATA66 capable drives. VIA has a new chipset
out that does support ATA66, but it is not the chipset that is on the
1590.
Mark Hahn
> Tht's about right. What type of sustained transfer rate were you looking
> for?
he's right: it's too slow. the disk should be delivering more like 8-12MB/s.
>> Having just built this system, using a TYAN S1590S board, bios 1.14, and
>> buying a UDMA33 capable drive, also according to the docs, PIO modes 3
>> and 4 are supported.
you certainly don't want pio.
>> Under RH linux 5.2, original kernel 2.0.36 it default powerups at 4.75
why are you running a kernel that's basically 3 years old?
>> megs/second. Regardless of what I do in the bios, or with hdparm (many
>> of the operations are 'not supported'), the best I can get it to do is
>> about 6.75 megs/second.
get 2.2.9, and configure it to support your controller (VIA82C586).
>> The drive is a WDC AC14300R, FwRev 15.01J15
it's not the fastest drive, but you should be seeing >10 MB/s
on outer tracks.
Gene Heskett
Gene Heskett sends Greetings to Philip Morris;
[...]
PM> Gene,
PM> Go to WD's web site and look for the utility "WD ATA66". It does not
PM> attempt to format your hard drive, it simply switches it to an ATA33
PM> compliant drive.
PM> The 1590 does not support ATA66 capable drives. VIA has a new chipset
PM> out that does support ATA66, but it is not the chipset that is on the
PM> 1590.
I did that last nite, and ran it this morning, turning the ata66 mode
off, then put everything in the bios back to auto. Still 5.98
megs/second. The 'hdparm -d1 /dev/hda' option apparently is not
supported with this kernel/chipset, so the use_dma flag cannot be turned
on. But, setting the prefetch on with a -A1 gets it all the way up to
3.18 megs/sec! And an A0 setting doesn't get the speed back.
Don't ask me whats going on, everytime I call hdparm the damned thing
gets slower! I've had ns, v4.51, take as long as 5 minutes, with very
little disk activity, to open its first screen!
Tell ya what, this 28 mhz 040 equipt amiga can run that 400 mhz linux
box to the ground, stand on its neck and administer the coupe de grace,
and do it without even raising a sweat! So far, the only place its
faster is at keycracking, 22kkeys for the miggy vs 670 for the linux box.
Cheers, Gene
--
Gene Heskett, CET, UHK |Amiga A2k Zeus040 50 megs fast/2 megs chip
Ch. Eng. @ WDTV-5 |A2091,GuruRom,1g Seagate,CDROM,Multiface III
<gene_h...@iolinc.net> or |Buddha + 4 gig WDC drive, 525 meg tape
<gene_h...@westvirginia.net>|Stylus Pro, EnPrint, Picasso-II, 17" vga
RC5-Moo! 22kkeys/sec isn't much, but it all helps
--
Gene Heskett
Gene Heskett sends Greetings to Mark Hahn;
MH> In comp.os.linux.hardware The Everette's <ever...@erols.com> wrote:
>> Tht's about right. What type of sustained transfer rate were you looking
>> for?
MH> he's right: it's too slow. the disk should be delivering more like
MH> 8-12MB/s.
>>> Having just built this system, using a TYAN S1590S board, bios 1.14, and
>>> buying a UDMA33 capable drive, also according to the docs, PIO modes 3 and
>>> 4 are supported.
MH> you certainly don't want pio.
>>> Under RH linux 5.2, original kernel 2.0.36 it default powerups at 4.75
MH> why are you running a kernel that's basically 3 years old?
Thats what comes with the RH5.2 (Unleashed) distrib. A distribution I
now consider to be thoroughly broken, since I've had to get XF86 stuffs
from the 6.0 to make my video work, and the libc from that distrib, plus
printtool and rhs-printfilters from the 6.0 database, just to get the 2
main things to work, video, and printing!
>>> megs/second. Regardless of what I do in the bios, or with hdparm (many of
>>> the operations are 'not supported'), the best I can get it to do is about
>>> 6.75 megs/second.
MH> get 2.2.9, and configure it to support your controller (VIA82C586).
I just did get that last night. At ~700cps :(
What else do I need to get to keep pppd working? I understand the
version I have is killed by later kernels.
I had downloaded that to /root, then moved it to my ls120 which brought
the disk up to 50% full, and bzip2 -d can't unpack it, 60 megs is not
enough space. So it just got moved back to the hard drive. I might see
if a fresh disk has enough room.
>>> The drive is a WDC AC14300R, FwRev 15.01J15
Cheers, Gene
Gene Heskett
Yeah, I know, talking to onesself is a sign of alzheimers :)
>>>> megs/second. Regardless of what I do in the bios, or with hdparm (many
>>>> of the operations are 'not supported'), the best I can get it to do is
>>>> about
>>>> 6.75 megs/second.
MH>> get 2.2.9, and configure it to support your controller (VIA82C586).
GH> I just did get that last night. At ~700cps :(
GH> What else do I need to get to keep pppd working? I understand the
GH> version I have is killed by later kernels.
GH> I had downloaded that to /root, then moved it to my ls120 which brought
GH> the disk up to 50% full, and bzip2 -d can't unpack it, 60 megs is not
GH> enough space. So it just got moved back to the hard drive. I might see
GH> if a fresh disk has enough room.
And now I've had a chance to play with it, and the compile is failing
because of what seems to be a totally off the wall mistake.
In ./arch/i386/lib/compress.c, line 104-105 is a definition of a
funtion,
csum_partial_copy_fromuser( 2 ptrs, 2 ints ) ; a 27 char name!
and at line 200, this
csum_partial_copy( same vars as above ) ; but name is only 18 chars...
Make, it appears, is treating them as if the names have been clipped,
therefore making them identical according to the error messages output
to the shell window. Make exits with an error 2 from the
apparent redefinition attempt.
While I've written some C code, I'm not a guru, so what now folks?
Steve Ponsford
Mmm, pull it out of your PC and give it a whirl on you miggy :) if you got
Linux installed
on it and the proper cables to support that drive to your interface. The newer
2.3.x kernels support Amigas native Gayle IDE controller and the
Buddha/Catweasel IDE interface. Of course your PIO settings may not even be
recognized but it may be interesting to see the speed difference between the
two systems.
Also try this with hdparm
hdparm -i /dev/<drive> and look at you MaxMultSect if it's supported by your
drive. Take
the number returned from MaxMultSect=?? and set the -m option to that number.
However,
if you have a WD Cavair series drive set -m to a lower number (try 4 or 8).
Read the man
pages on hdparm for more info on WD drives. I might also help if you have
32-bit transfer
mode on by setting -c option to on. Then do a 'hdparm -t -T /dev/<drive>'
before and
after you changed the settings to see if your performance as increased any.
Hope this helps.
Steve Ponsford | Team //\ | A3000-40P96/A500-30+/GVP8+GuruROM
fleg | \\//miga | K6III-450/MandrakeLinux&UAE
Gene Heskett wrote:
> Reply to: <engin...@wdtv.com>
>
> Gene Heskett sends Greetings to Philip Morris;
>
> [...]
>
> PM> Gene,
>
> PM> Go to WD's web site and look for the utility "WD ATA66". It does not
> PM> attempt to format your hard drive, it simply switches it to an ATA33
> PM> compliant drive.
>
> PM> The 1590 does not support ATA66 capable drives. VIA has a new chipset
> PM> out that does support ATA66, but it is not the chipset that is on the
> PM> 1590.
>
> I did that last nite, and ran it this morning, turning the ata66 mode
> off, then put everything in the bios back to auto. Still 5.98
> megs/second. The 'hdparm -d1 /dev/hda' option apparently is not
> supported with this kernel/chipset, so the use_dma flag cannot be turned
> on. But, setting the prefetch on with a -A1 gets it all the way up to
> 3.18 megs/sec! And an A0 setting doesn't get the speed back.
>
> Don't ask me whats going on, everytime I call hdparm the damned thing
> gets slower! I've had ns, v4.51, take as long as 5 minutes, with very
> little disk activity, to open its first screen!
>
> Tell ya what, this 28 mhz 040 equipt amiga can run that 400 mhz linux
> box to the ground, stand on its neck and administer the coupe de grace,
> and do it without even raising a sweat! So far, the only place its
> faster is at keycracking, 22kkeys for the miggy vs 670 for the linux box.
>
Gene Heskett
Soneone (Mark Hahn) said:
MH>>> get 2.2.9, and configure it to support your controller (VIA82C586).
Plz note that *only* in 'Make xconfig' are the options to use that shown
to the poor user. Make menuconfig doesn't give you the choice that I
could see.
What else do I need to get to keep pppd working? I understand the
version I have from the 5.2 install is killed by later kernels.
[...]
And now I've had a chance to play with it, and the compile is failing
because of what seems to be a totally off the wall mistake.
In ./arch/i386/lib/compress.c, lines 104-105 is a definition of a
funtion:
csum_partial_copy_fromuser( 2 ptrs, 2 ints ) ; a 27 char name!
and at line 200, this:
csum_partial_copy( same vars as above ) ; but name is only 18 chars...
Make, it appears, is treating them as if the names have been clipped,
outputting the error that 'csum_partial_copy' is defined in line 105,
and redefined in line 200, therefore making them identical according to
the error messages output to the shell window. Make exits with an error
2 from the apparent redefinition attempt.
While I've written some C code, I'm not a guru, so what now folks?
Cheers, Gene
Steven Ellis
It sounds like your UDMA chipset isn't supported with your version of
the Kernel. I had the same problem with the VIA chipset on my FIC
motherboard.
I did try including the VIA patches for the 2.0.36 kernel which helped,
but I now run 2.2.x which already includes a lot more chipsets. Also you
can turn on UDMA by default.. Boy does it make a big difference.
Steve
--
------------------------------------------------------------------------------
If you listen very carefully you can sometimes hear the dolphins
sing.
Email: steven...@pmail.net
Steven Ellis
Steven Ellis <steven...@pmail.net> wrote:
>Hi Gene
>
>It sounds like your UDMA chipset isn't supported with your version of
>the Kernel. I had the same problem with the VIA chipset on my FIC
>motherboard.
>
>I did try including the VIA patches for the 2.0.36 kernel which helped,
>but I now run 2.2.x which already includes a lot more chipsets. Also you
>can turn on UDMA by default.. Boy does it make a big difference.
Hm.. well it used to make a big difference, but now I appear to have
problems with my UDMA 33 hard drive. I keep getting dma read errors when I
check it using HDPARM or FSCK.EXT2
Any ideas
Steve