sil3114 sata card

[T. Ment]

I bought a cheap sil3114 sata card to try as an add-in card in a
motherboard that has no SATA ports. It worked OK until I tried flashing
its BIOS to the latest version. The updflash.exe utility started up, but
failed.

Now the sil3114 card hangs the computer when booting. Seems I bricked
its BIOS. I tried powering up and hot-inserting the card after boot, but
that freezes the keyboard and I can't do anything.

The card has a soldered on BIOS chip. I have the tools to get it off
without destroying the card. Then I can put it in my USB programmer and
reload the BIOS. Then I can put a chip socket on the card to take the
fixed BIOS.

But that means I have to drag out the tools and make a big mess. It's a
lot of work to fix a $9.49 card.

I've not found a way to get the motherboard BIOS skip the bad BIOS on
the add-in card, and boot up. I've tried two different motherboards.

Any ideas?

[Paul]

Turn off "Interrupt 19 capture".

I haven't tested this with a corrupted config EEPROM,
so I don't know how early the BIOS consults the
chip and whether it will still screw up.

But turning off capture should stop it from
loading config EEPROMs.

Obviously, if some *other* port was relying
on that kind of config EEPROM loading, there
could be trouble.

Paul

[T. Ment]

On Mon, 01 Jul 2019 17:21:10 -0400, Paul wrote:

> Turn off "Interrupt 19 capture".

I never heard of that, I had to look it up:


> Interrupt 19 Capture - Tech ARP
> https://archive.techarp.com/showFreeBOG311d.html?lang=0&bogno=290
>
> Interrupt 19 is the software interrupt that handles the boot disk
> function. It is typically handled by the motherboard BIOS although
> it can also be handled by the optional boot ROM BIOS in some
> IDE/SCSI host adaptors.

May be worth trying, but I don't have that BIOS option in any computer
of mine. Unless it goes by some other name.

[Paul]

In my non-UEFI computer (P5E Deluxe), the setting is

Boot Settings Configuration

Interrupt 19 Capture [Enabled] <== makes add-on cards boot

The number is 19 decimal or 0x13 hex. So you might have
to search for 0x13 in the manual.

I believe that routine is the Read Routine used for
storage controllers. Each storage controller "registers"
as bootable, and then can participate in boot menus
and the like. Or in the case of RAID cards, you might
even get the RAID BIOS screen add-on to load.

This setting might be needed to get a RAID addon
screen to show up.

Addon ROM display [Force BIOS]

*******

I cannot find this setting in the UEFI machine I got.

The user manual for my UEFI machine makes reference
to "storage OPROM" (Option ROM) but only for the
hardware on the motherboard. No control is offered
for PCIe cards that I can see.

Perhaps you have to use a UEFI shell command
to fix it ? Dunno.

Paul

[T. Ment]

On Mon, 01 Jul 2019 19:47:24 -0400, Paul wrote:
to search for 0x13 in the manual.

>This setting might be needed to get a RAID addon
>screen to show up.
>
> Addon ROM display [Force BIOS]
>
>*******
>
>I cannot find this setting in the UEFI machine I got.
>
>The user manual for my UEFI machine makes reference
>to "storage OPROM" (Option ROM) but only for the
>hardware on the motherboard. No control is offered
>for PCIe cards that I can see.
>
>Perhaps you have to use a UEFI shell command
>to fix it ? Dunno.

My computers are old, pre-UEFI. I can't find any of the options you
have.

The sil3114 flash update program supports SST39SF010, and mine is the
same, but with an A suffix.

I found a web thread that says not all cards can be flashed. I guess if
they don't connect all the lines to the chip, that could be true. That's
probably why it failed.

My USB programmer claims to support the A suffix chip, so if that won't
work, nothing will. I had hoped to avoid all that work, but it seems the
hard way is the only way.

[Paul]

Is your computer a Dell/HP/Acer OEM type ?

Or a home build with an Asus or MSI motherboard ?

The home build boards come with pretty decent
manuals.

If you have the model number for your home build,
I might be able to find a copy.

The OEM computers, only in the rare exception cases
(Trigem boards), might you get an actual manual.

On PCChips motherboards (in a cheap OEM), the board
doesn't even get a model number! There used to be
a website called "PCchips Lottery" to cover those
cases :-) Not everybody is born lucky.

*******

"not all cards can be flashed"

What this means is, the Flasher program is designed for PMC or
SST brand flash chips. If the manufacturer uses an "oddball"
chip, if you have a universal programmer, of course you
can flash it up. But that means getting out your soldering
iron and so forth.

The boards I've seen, the SIL ones, have pretty good
compliance on Flash brand.

The "A" could be a speed grade, which means the pinout
and registers would likely be the same as the non-A one.

I think those flash used to run at 33MHz or so. Or
whatever the PCI bus clock happens to be. That might
be a typical value. It might say on the lid, what
the speed is.

The only SIL I've got here, is a SIL3112 on a motherboard,
where the "flash" is a chunk of code in the BIOS chip.
The Flash in that case, I needed a newer BIOS to take
care of the "won't work with 1TB drives" problem. That
was a bug in the Silicon Image code, which could be fixed.

Paul

[Flasherly]

On Mon, 01 Jul 2019 21:00:40 +0000, T. Ment <t.m...@protocol.invalid>
wrote:

Good ol' SIL. Look familiar?
07/26/2005 10:16 AM 25,468,006 SI3114.exe
SiI3114_RAID_1002.zip
Silicon Image Serial ATA IDE Driver
V1.0.0.2 and Raid Driver V1.0.0.7
for Windows 2000/XP/2003
Ver 1.0.0.2 2005/07/26 update
OS Win2K / WinXP / Win2003
Size 24.69 (MBytes)

Ran it with ASUS, so it was ages ago. I've since switched two times,
maybe three, upgrading with Gigabyte MBs.
asus_K8N-E+ AMD 754 3000

I also ran with a couple other controller boards. It really doesn't
matter what the names are in terms of performance. They're basically
cheap, less than a gamble, as you may have to take what you can get
(without a lot of offerings). So SIL is just fine and peachy.

I had one, I recall, the performance was off or operation
characteristics didn't seem right. I replaced that was an updated
model, which didn't evince problems -- so it was fixed. Same board
maker -- or call it SIL for convenience or an example. (Only a
Gigabyte with only two SATA ports).

And I had one, as you do, which I flashed its BIOS and it ate the big
one. Not sure which card I killed, but it's strange they're made
with, rather for a utility to download that then it kills perhaps
better than your average mousetrap. You might want to kiss it goodbye
before tossing it into the trash.

You need to reset your MB with a paper-clip bend across the two pins
marked for a RESET;- or pull the MB battery for a minute or thirty.

That I never managed to do, to kill a BIOS boot bunged-up settings
with SIL. That I've bricked a BIOS, done in other inventive ways
perhaps goes unsaid. As in permanently bricked, though, that should
be able to happen unless it's a BIOSTAR (that one and another brand,
twice, I've had MBs sold to me that would boot, except they wouldn't
hold user changes to BIOS settings). It's permanently 1980 everyday
for the rest of life.

Maybe you should update like me. Now I have 6 and 8 SATA slots on my
newest MB and I still don't like the MB's HDD controller chipset
performances;- In fact the last performance I did like was ...

Discounting when MBs hadn't totally cheaped out on slots and didn't
have no stinkin' HDD controllers (ISA HDD controllers). That kind of
talk makes my memory go hazy.

These SIL cards I've also owned ... Guess this makes four. And what
have we now learned? ...That's right -- don't ever flash one again.
But you can buy another one now, if you want. And it may as well be a
SIL: for the money you may not find better in another brand. (Under
$10 on Ebay ... original-packaging new and m a y b e).

-

; This file installs the SiI 3112 non-RAID serial ATA driver as part

-

Description: PCI 2-channel Serial-ATA host controller card. With
optional software RAID function. The most popular version of 2-channel
Serial-ATA host controller add-on card, with optional RAID 0, 1
function.

* PCI Specification Revision 2.2 compliant
* Silicon Image SIL 3512 host controller chip
* Support 66 Mhz PCI with 32-bit data
* Compliant with programming interface for Bus Master IDE
Controller, Rev1.0
* Support programmable and EEPROM, FLASH & EPROM loadable PCI
class mode
* Integrated SATA Transport, Link Logic & PHY layer
* 48-bit sector addressing
* Virtual DMA
* Serial ATA Specification Revision 1.0 compliant
* Dual independent DMA channels with 256KB FIFO per Serial-ATA
channel, transfer rate up to 1.5Gb/s
* Internal Serial-ATA port x 2
*
* Supports 3TB HDDs
* Supports SSD.
* Support Boot to CD/DVD

-

use included CD and at driver prompt:
search CD (for drivers)

SD-SATA150R
Description: PCI 2-channel Serial-ATA host controller card. With
optional software RAID function. The most popular version of 2-channel
Serial-ATA host controller add-on card, with optional RAID 0, 1
function.

* PCI Specification Revision 2.2 compliant
* Silicon Image SIL 3512 host controller chip
* Support 66 Mhz PCI with 32-bit data
* Compliant with programming interface for Bus Master IDE
Controller, Rev1.0
* Support programmable and EEPROM, FLASH & EPROM loadable PCI
class mode
* Integrated SATA Transport, Link Logic & PHY layer
* 48-bit sector addressing
* Virtual DMA
* Serial ATA Specification Revision 1.0 compliant
* Dual independent DMA channels with 256KB FIFO per Serial-ATA
channel, transfer rate up to 1.5Gb/s
* Internal Serial-ATA port x 2
*
* Supports 3TB HDDs
* Supports SSD.
* Support Boot to CD/DVD

[T. Ment]

On Tue, 02 Jul 2019 11:45:17 -0400, Flasherly wrote:

>Good ol' SIL. Look familiar?

> And I had one, as you do, which I flashed its BIOS and it ate the big
> one. Not sure which card I killed, but it's strange they're made
> with, rather for a utility to download that then it kills perhaps
> better than your average mousetrap. You might want to kiss it goodbye
> before tossing it into the trash.
>

> That I never managed to do, to kill a BIOS boot bunged-up settings
> with SIL. That I've bricked a BIOS, done in other inventive ways
> perhaps goes unsaid. As in permanently bricked, though, that should
> be able to happen unless it's a BIOSTAR (that one and another brand,
> twice, I've had MBs sold to me that would boot, except they wouldn't
> hold user changes to BIOS settings). It's permanently 1980 everyday
> for the rest of life.

Even a bricked BIOS can be fixed with skill and the right tools. It's
harder if the chip is soldered on instead of socketed, but still doable.
I just finished this one, putting a PLCC32 socket where the chip was.

I reprogrammed the BIOS chip with my USB programmer and put it in the
socket. Now it boots again.

But I get disk errors when writing to the disk. Seems the card was bad
all along. I did a lot of soldering work to figure that out. And I can't
return the card now that I modified it.

Maybe I trashed the card when I hot plugged it trying to get it booting
after the bad flash. I won't do that again. Oh well, $9.49 is not a big
loss.

> Maybe you should update like me.

I have other motherboards with SATA onboard, and the disk works fine
there. But I want to get it running in an old board that has no SATA
onboard.

> But you can buy another one now, if you want. And it may as well be a
> SIL: for the money you may not find better in another brand. (Under
> $10 on Ebay ... original-packaging new and m a y b e).

Maybe I'll try again for $10 more. I'll think about it.

[T. Ment]

On Mon, 01 Jul 2019 22:52:27 -0400, Paul wrote:

> Is your computer a Dell/HP/Acer OEM type ?
> Or a home build with an Asus or MSI motherboard ?

I don't have any name brands. All white box, most from Ebay parts for
pennies on the dollar. I did buy some new power supplies a long time
ago. All still running, after some failed capacitor replacements.

> If you have the model number for your home build,
> I might be able to find a copy.

I have all the manuals, either paper or PDF.

> On PCChips motherboards (in a cheap OEM), the board
> doesn't even get a model number!

Some people trash PCChips and Biostar. But I have some running fine.

> "not all cards can be flashed"
>
> What this means is, the Flasher program is designed for PMC or
> SST brand flash chips. If the manufacturer uses an "oddball"
> chip,

The flasher program claims to support the chip but something went wrong.
Just enough to erase the chip it seems. If you don't connect all the
chip pins to the right wires it's not going to work.

> if you have a universal programmer, of course you
> can flash it up. But that means getting out your soldering
> iron and so forth.

That's what I did. Chip flashed. Computer boots. But the card itself is
bad, I get disk write errors.

A lot of work for a small piece of knowledge.

[Flasherly]

On Tue, 02 Jul 2019 20:03:54 +0000, T. Ment <t.m...@protocol.invalid>
wrote:

>Even a bricked BIOS can be fixed with skill and the right tools. It's
>harder if the chip is soldered on instead of socketed, but still doable.
>I just finished this one, putting a PLCC32 socket where the chip was.
>
>I reprogrammed the BIOS chip with my USB programmer and put it in the
>socket. Now it boots again.
>
>But I get disk errors when writing to the disk. Seems the card was bad
>all along. I did a lot of soldering work to figure that out. And I can't
>return the card now that I modified it.
>
>Maybe I trashed the card when I hot plugged it trying to get it booting
>after the bad flash. I won't do that again. Oh well, $9.49 is not a big
>loss.

I've never run into a MB or had a bricked BIOS, apart the two
mentioned. The Biostar I had to fight them they accepted a dispute
and settled for a RMA/refund. The other one, a power outage, the
clock multiplier would drop default down on the CPU to the lowest
multiple. I got rid of and sold that one.

>> Maybe you should update like me.
>
>I have other motherboards with SATA onboard, and the disk works fine
>there. But I want to get it running in an old board that has no SATA
>onboard.
>
>
>> But you can buy another one now, if you want. And it may as well be a
>> SIL: for the money you may not find better in another brand. (Under
>> $10 on Ebay ... original-packaging new and m a y b e).
>
>Maybe I'll try again for $10 more. I'll think about it.

I've three or four of those SLI boards in working order and great
shape. Also one or two older MBs with two at the most SATA ports, or
I'd probably be happy to give them away. All that older SLI stuff,
including the MB HDD controllers, runs great with plattered HDD
defragmentation routines, nice and slick, whereas these two newer MBs
I've updated to, (AMD3+ Bulldozer stuff), it's like pulling hair
trying to perform a defragmentation. Godawful slow.

Not that I did always have perfectly performing SLI boards. One, it
seems, over time got worse, and may have (helped) wear out a HDD. Hard
to say, but the SLI replacement model, one up from that, there were no
issues/complaints. Before, budget MBs cheaped out on SATA slots,
maybe two;- their MB controllers also worked fine. Now that I've a
$59 MB board with 8 SATA ports, I couldn't use those SLI boards on it
anyway, not without giving the only PCI slot it has, one, where my
soundboard sits.

What I'm thinking about, is I hope those newer controllers aren't
adversely affecting my HDDs with defragmentation nonsense that takes
in an order exponentially longer than the copy routines. 15 or 30
minutes to defrag I file I just copied in 15 sec (writes/copies work
fine). Ridiculous. Anything substantially a large data capacity, is
better defraged through mounting up a blank disk and doing manual
copy, which usually results in low to no fragmentation.

$10 more, I hear you. I didn't like it either. Not quite sure now,
but seems I may have made that BIOS rewrite attempt on more than one
card. That it was a possible second instance with another SLI card
that evinced the brick. Bricked mine and never went near another SLI
with another thought of BIOS rewrite, that's for sure. I don't like
spending money on "normal" stuff that shouldn't be lacking on a MB,
whether the MB was cheap to me or not. And things could be different
than me saying, offhand, a $10 board is a remote chance. I'd be as
easily wrong if in fact a SLI is now a pricier proposal.

(I think some SLI boards may POST before the BIOS, it's been a few
years ago, as well have a key sequence procedure to an internal SLI
response for activating RAID -- in conjunction and in either case to
use with supplied OS drivers.)

[T. Ment]

On Tue, 02 Jul 2019 11:45:17 -0400, Flasherly wrote:

> But you can buy another one now, if you want. And it may as well be a
> SIL: for the money you may not find better in another brand. (Under
> $10 on Ebay

I see plenty of PCIe, but not much standard PCI. The next best offer is
$15 for the same card.

I don't want another from the same seller, maybe they have a bad batch.

[T. Ment]

On Tue, 02 Jul 2019 21:15:27 +0000, T. Ment wrote:

> The flasher program claims to support the chip but something went wrong.
> Just enough to erase the chip it seems. If you don't connect all the
> chip pins to the right wires it's not going to work.

Here's another thread where people had problems flashing:

https://www.vogons.org/viewtopic.php?f=46&t=47717

[T. Ment]

On Tue, 02 Jul 2019 20:03:54 +0000, T. Ment wrote:

> Maybe I trashed the card when I hot plugged it trying to get it booting
> after the bad flash.

The card has a jumper jp4. I didn't what it was for, so I tried it. This
was before desoldering the BIOS chip. It didn't help.

But now I think those two pins are for a HDD LED. Shorting it to ground
was probably a bad idea. Maybe that damaged the card.

[T. Ment]

On Wed, 03 Jul 2019 01:37:48 +0000, T. Ment wrote:

> But now I think those two pins are for a HDD LED. Shorting it to ground
> was probably a bad idea. Maybe that damaged the card.

This board has a boat load of surface mount capacitors.

There's probably just one bad component somewhere. Desoldering and
checking them all would take too much time. More than $9.49 worth.

Hey! I found a broken solder joint on one side of C41. Tomorrow I'll
repair it, and see if that helps.

[Paul]

As long as the two pins on the LED header were
not shoved into any other electrical potential,
odds are good you didn't hurt it.

+5V ------ Rseries -----X \
\___ use a LED, or short the pins,
X / won't hurt it. Top pin is (+).
|
activity ____|/
|\
v
|
GND

If you applied a hard rail (voltage source) to the bottom
pin, from elsewhere on the card, you could damage the
open collector drive transistor. If you short the header
pins together, the Rseries (180 ohms or so), protects
your investment.

Yes, shoving a card "hot" into a PCI slot, is a bad idea.
If the slot had advanced power/ground (sequencing), then
it might be OK. Regular slots don't provide sufficient
control of the entry angle of the card, to ensure bad things
don't happen.

You can test-boot your machine now, with the card removed, and
verify the PCI bus was not damaged.

*******

What surprises me about all these Silicon Image based products:

1) Silicon Image is no longer interested in storage.
That happened at least five years ago.

2) They probably sold off that business to someone else.

3) The downside of the sale, is no driver maintenance or BIOS
file maintenance. You're relying on Syba downloads page
for various materials. Silicon Image had its own site at one
time.

The mind-boggler, is why they're selling the cards with
the crusty old RAID BIOS in it. Nobody really expects a
card like that to be good at RAID. The lack of hardware
XOR means it would suck at RAID5. And I already recounted
a failure mode the SIL3112 had in RAID1 mode, that leaves
me wondering at their sanity.

What you want from these cards, is JBOD mode. And for someone
to volunteer at the company that is pumping this shit out,
to make sure there is a Windows 10 driver.

You can use a PCI to PCIe bridge chip, to allow the
usage of PCIe SATA chips with the PCI bus. People
have done that before. There is no particular advantage
to this, other than to put some distance between you
and the "smell" of SIL BIOS chips. The regular PCI bus
can't go faster than around 110MB/sec with a decent
burst size.

The SIL3124 is a SATA card, that could do PCI-X 64/66 and
would have four times the bandwidth of the SIL3114. But
on a regular motherboard it drops back to 32/33MHz and
the 110MB/sec figure. Modern hard drives can reach
250MB/sec (the record is a certain 15K drive that
does 300MB/sec). The drive is not hurt if it "goes slower".

Unless I see a company put some effort into support on these,
I couldn't in good conscience, touch this with a barge pole.
I mean, FFS, they're still using the same labels they
were using more than 10 years ago! That label on the ROM
gives me the chills... (It hides the chip part number
so we don't know what is under there.)

https://www.amazon.com/IOCrest-SATA-RAID-Controller-SY-PCI40010/dp/B002R0DZZ8/ref=sr_1_36

It was bad enough working with this stuff, when it was supported.

Paul

[T. Ment]

On Wed, 03 Jul 2019 00:31:42 -0400, Paul wrote:

> What you want from these cards, is JBOD mode. And for someone
> to volunteer at the company that is pumping this shit out,
> to make sure there is a Windows 10 driver.

I need it for a DOS computer, to scrap a slowly failing PATA drive. I
need a card with BIOS, to boot from SATA and give performance comparable
to PATA. I have the drive, a 320 GB WD, my first drive ever, over 160GB.
It works, and I need a SATA card for standard PCI Not many choices for
standard PCI that I can find.

> The SIL3124 is a SATA card, that could do PCI-X 64/66 and
> would have four times the bandwidth of the SIL3114. But
> on a regular motherboard it drops back to 32/33MHz and
> the 110MB/sec figure. Modern hard drives can reach
> 250MB/sec (the record is a certain 15K drive that
> does 300MB/sec). The drive is not hurt if it "goes slower".

A 3124 sounds good. They price at $35. I have to think about that.

> Unless I see a company put some effort into support on these,
> I couldn't in good conscience, touch this with a barge pole.

The only difference between old junk and new junk is the price. I like
old junk. I spend less money than people who demand new. Today's factory
new is tomorrow's old junk. In the end, it's all junk.

I repaired the solder joint, but still get disk write errors. There must
be some other defect not visible.

[Flasherly]

On Wed, 03 Jul 2019 14:49:32 +0000, T. Ment <t.m...@protocol.invalid>
wrote:

>I have the drive, a 320 GB WD, my first drive ever, over 160GB.
>It works, and I need a SATA card for standard PCI Not many choices for
>standard PCI that I can find.

I was given free USB HDD docking stations when buying something else.
But they're for a SATA drive;- some firmware limited to under 1 or
1.5T (FAT32), others 2T, and that's it -- never got around to docking
at over a 2T HDD. At the time I got them, they very cheap, among
various brands around $10 or $15. I like them, leave them on
occasions along with a tiny 3" 120V fan directed on the exposed half
of the mounted HDD, hooked up for unattended large transfers. Provided
they do not, on rarer occasions, sych-out with an error.

(Also aggravating, the controllers my docking stations fly on a
defragmentation routine, whereas my MB's SATA controllers are sludge
for fragmentation purposes.)

PATA connections, I wouldn't recommend the SATA<>PATA converter I
tried. ...DOS boots, & etc., (HIRENS supports USB in a DOS config),
or in as much and depending on whether a docking station is a viable
option to you.

Easy to say with only one Seagate 200G PATA drive left, never used
anymore, although full as a triplicate measure just to fill its space
among the rest of SATA, or a duplicate data backup drive set.

[T. Ment]

On Wed, 03 Jul 2019 14:49:32 +0000, T. Ment wrote:

> I need it for a DOS computer

The Ebay seller told me it does not support DOS.

So I booted linux, and it works. Seems it needs a driver to poke its
chipset registers the right way.

Looks like my plan for DOS won't fly. I spent a lot of time and solder
learning that. Glad I only spent $9.49 on it. That's one reason to buy
old cheap junk.

[Paul]

https://www.vogons.org/viewtopic.php?f=46&t=49471

"If you want the SATA card to be usable by DOS
and Windows 9x, you need to flash it with the "base" BIOS.

The file you need is called b5403.bin

Reboot. After reboot, you should no longer see an option
to enter the RAID utility for the card."

"It didn't look like there were any IRQ conflicts
when the system booted up, but nonetheless, the card flashed
totally fine after moving to a different PCI slot."

That means you need an older motherboard, one with a motherboard
manual that details IRQ sharing. Do the flash there, then move
the card to a newer machine.

The 0x13 routine is what DOS uses, isn't it ? The BIOS routine
continues to be uses for ATA. For ATAPI, it seems to use
one of several ATAPI drivers (that you put on your MSDOS floppy).

Paul

[T. Ment]

On Wed, 03 Jul 2019 23:49:14 -0400, Paul wrote:

> https://www.vogons.org/viewtopic.php?f=46&t=49471
>
> "If you want the SATA card to be usable by DOS
> and Windows 9x, you need to flash it with the "base" BIOS.
>
> The file you need is called b5403.bin

That's what I tried when the flashing problem started. My USB programmer
got the b5403 BIOS flashed, but it still does not work with DOS 6.22.

Your thread link mentions DOS 7. And the flasher program says to use
FreeDOS. Maybe they make a difference. All I know is, it's not working
with DOS 6.22.

> Reboot. After reboot, you should no longer see an option
> to enter the RAID utility for the card."
>
> "It didn't look like there were any IRQ conflicts
> when the system booted up, but nonetheless, the card flashed
> totally fine after moving to a different PCI slot."
>
>That means you need an older motherboard, one with a motherboard
>manual that details IRQ sharing. Do the flash there, then move
>the card to a newer machine.

Like he said, I tried different PCI slots. Didn't help. His card must be
flashable. I think mine is not.

>The 0x13 routine is what DOS uses, isn't it ? The BIOS routine
>continues to be uses for ATA. For ATAPI, it seems to use
>one of several ATAPI drivers (that you put on your MSDOS floppy).

Yes 0x13 is the hard disk BIOS INT.

I have a multiboot setup with DOS, linux, and Windows. Before testing
the sil3114 card, I cloned the multiboot drive onto the SATA drive with
linux dd.

For testing, I have both drives in the computer. I boot DOS from PATA
(C:), and DOS sees the SATA as D: I can change to it and display a
directory, but copying a large file fails. Same test works fine when I
boot linux.

I wonder why the flasher program recommends FreeDOS. Maybe I can try
that later. Or maybe I'll try a 2 port sil3512 card. I seem to recall
reading it supports DOS.

[Paul]

FreeDOS would be recommended due to availability.

Configuring regular MSDOS on newer systems is difficult.
It's getting harder and harder to figure out what to
do with the address space. I had less trouble with some
older systems.

Paul

[T. Ment]

On Wed, 03 Jul 2019 23:06:34 +0000, T. Ment wrote:

> it needs a driver to poke its chipset registers the right way

I found its data sheet:

www.pix.net/techpubs/silicon_image/SiI-DS-0103-D.pdf

The PDF is 119 pages, and has programming hints. Guess I could write my
own DOS device driver, given enough time.


Page ii:

11/29/06 This datasheet is no longer under NDA. Removed
confidential markings

Page 74:

The programming sequence for the SiI3114 is about the same
as for the SiI3112 or SiI3512. However, SiI3114 supports up
to four SATA devices (instead of two for the others).

[T. Ment]

Now that I've fixed its broken solder joint, and I know the card works
in linux, I can verify the DOS flasher DOES NOT work. It gets far enough
to trash the chip's BIOS and then fails, replacing the card with a brick
that hangs the computer at boot.

I doubt it works with the Windows flasher either. Other sil3114 cards
may flash, but not this one. It has a r5403 BIOS soldered on. Take it or
leave it.

I found the Lattice archive for legacy SIL cards:

https://www.latticesemi.com/Support/ASSPSoftwareArchive

There you can find a 3512 BIOS, version 4.3.70. It seems unique, they
call it an "IDE" BIOS. I flashed it onto my 3114 card, and it boots.

It also works in DOS. I tried two later versions of the 3512 BIOS. Both
boot, but like the 3114 BIOS, fail on a file copy. So there's something
special about the 4.3.70 BIOS that lets it work in DOS.

They call it an "IDE" BIOS. Maybe that means IDE vs AHCI. Don't know.

Though it worked with a DOS file copy, I could not get it to boot from
the SATA drive, in the computer I had in mind for it. Maybe the BIOS is
too old and the hard drive is too big. Or maybe it's incompatible with
the motherboard BIOS. Don't know.

Looks like they gave up supporting DOS after that version. The archive
has no such BIOS for a 3114 card.

[Paul]

OK, and was there a "max disk size for MSDOS" ?

It can't go over 28 bit LBA or 137GB.
A 120GB SATA would work, a 160GB might be too much
(and especially if a partition "spanned" the boundary,
that's supposed to be bad).

On the old IDE drives, you could use the CLIP jumper
to clip to 33.7GB or 2GB or so. The geometry was the
same in each case, and it was the interpretation of
the OS at the time, that decided what that was. The CLIP
causes the drive to declare a magical CHS value.

I don't think SATA drives have the clip jumper, because
of course, the SATA era is 48 bit LBA (double pumped registers).
It's doubtful anybody gave a rats ass about DOS.

You can also clip a drive down with a Host Protected Area (HPA).
My other machine, the BIOS simply doesn't allow those operations
(they're locked out). On this machine, only the Jmicron port
allows setting an HPA.

*******

This is the proposal to move from 28 bit to 48 bit.
While some parts of the ecosystem might handle this,
I don't know if the whole thing (DOS part) does.
DOS was likely too late to the party for this.
DOS could be generating 28 bit addresses only
(good for the beginning of the disk... maybe).

https://web.archive.org/web/20041024150852/http://www.t10.org/t13/technical/e00101r6.pdf

There used to be web pages that would explain every limit,
all the way up to 2.2TB. But I doubt I could find those
pages today.

Paul

[T. Ment]

On Thu, 04 Jul 2019 16:52:12 -0400, Paul wrote:

> OK, and was there a "max disk size for MSDOS" ?

> It can't go over 28 bit LBA or 137GB.
> A 120GB SATA would work, a 160GB might be too much
> (and especially if a partition "spanned" the boundary,
> that's supposed to be bad).

I've heard something like that, but I have DOS multibooting on a
computer with a 160 GB PATA drive. Linux fdisk says:

Disk /dev/sda: 160.0 GB, 160041885696 bytes
255 heads, 63 sectors/track, 19457 cylinders
Units = cylinders of 16065 * 512 = 8225280 bytes
Sector size (logical/physical): 512 bytes / 512 bytes
I/O size (minimum/optimal): 512 bytes / 512 bytes
Disk identifier: 0x00000000

Device Boot Start End Blocks Id System
/dev/sda1 * 1 261 2096451 6 FAT16
/dev/sda2 262 275 112455 7 HPFS/NTFS/exFAT
/dev/sda3 276 295 160650 83 Linux
/dev/sda4 296 19457 153918765 f W95 Ext'd (LBA)
/dev/sda5 296 4030 30001387 b W95 FAT32
/dev/sda6 4031 4871 6755332 83 Linux
/dev/sda7 4872 4998 1020127 82 Linux swap
/dev/sda8 4999 8485 28009327 7 HPFS/NTFS/exFAT
/dev/sda9 8486 14593 49062509+ 7 HPFS/NTFS/exFAT
/dev/sda10 14594 19457 39070079+ 7 HPFS/NTFS/exFAT

DOS is the 1st partition. It boots. I use it.

The active flag changes at boot time, extipl does that. It's an obscure
boot manager. It's no longer supported or developed, but it does what I
need.

[Flasherly]

On Thu, 04 Jul 2019 19:51:33 +0000, T. Ment <t.m...@protocol.invalid>
wrote:

Another card I found (SD-SATA150R) I also have -- maybe I have four or
five SIL cards floating around here. As well a "SI_Buslink card" or
so-named and tossed in that directory, which seems as much a grab bag
driver situation for possibly all sorts of cards;- a rotten JPG I
made of my card looks to be a 2 header PATA setup, so there maybe two
PATA cards back in my stuff-it boxes.

Last is SI_RAID680 PATA card which has
SiI0680 BIOS Version 3.2.20. I really don't recall which board I
killed, except I likely wouldn't have kept it around after bricked,
only I do recall buying another, making both certainly then SATA cntr
boards.

(Goes without saying beyond 64/128G SSD only an updated controller
will see the likes of a quasi-256, 512G SSD 1T;- and they'll be seen
in DOS as well if at FAT32.)

Contents (SiI0680 BIOS)
I. Overview
II. Applicable Hardware/Software
III. Corrections

I. Overview - Contents
b3220.bin - add-in card BASE BIOS
r3220.bin - add-in card RAID BIOS
3220.bin - motherboard BIOS for OEM use in development. This
BIOS is
not intended for general end-users. End-users with a SiI0680A onboard
a motherboard, please contact the motherboard manufacturer for a BIOS
upgrade.
Note: The firmware download should take less than five minutes to
complete.
I. Corrections
1. Includes RAID volume support beyond 1TB.
2. Resolved a missing "Press F3 to enter RAID utility" message state.


Then there's this one, SI_RAID680 PATA card, nicely JPG'd and
decidedly a 2-port PATA header.

; This INF file installs the Silicon Image Serial ATA non-Raid driver
; for the SiI 3112 controller on systems running Windows 98/ME,
; Windows NT 4.0, Windows 2000, Windows XP, and Windows Server 2003.

(Can't offhand recall, but could be wrong, buying added USB ports in a
SIL flavor)

VIA Products Specification
VT6202 l Four Port USB 2.0 Host Controller
l Discrete PCI Adapter
VT6212 l Four Port USB 2.0 Host Controller
l Supports PCI / Cardbus Adapter
VT8235 l Six Port USB 2.0 Host Controller
l Integrated South Bridge

SD-SATA150R
Description: PCI 2-channel Serial-ATA host controller card. With
optional software RAID function. The most popular version of 2-channel
Serial-ATA host controller add-on card, with optional RAID 0, 1
function.

* PCI Specification Revision 2.2 compliant
* Silicon Image SIL 3512 host controller chip
* Support 66 Mhz PCI with 32-bit data
* Compliant with programming interface for Bus Master IDE
Controller, Rev1.0
* Support programmable and EEPROM, FLASH & EPROM loadable PCI
class mode
* Integrated SATA Transport, Link Logic & PHY layer
* 48-bit sector addressing
* Virtual DMA
* Serial ATA Specification Revision 1.0 compliant
* Dual independent DMA channels with 256KB FIFO per Serial-ATA
channel, transfer rate up to 1.5Gb/s
* Internal Serial-ATA port x 2
*
* Supports 3TB HDDs
* Supports SSD.
* Support Boot to CD/DVD

1) SIL680 - Driver for SIL680 chipset (Without RAID Function)

DOS - DOS Driver

Win_Drv - For Windows 98, 98SE, ME, 2000, XP - Ver 1.0.0.12

------

(2) SIL680RAID - Driver for SIL680 chipset (With RAID Function)

DOS - DOS Driver

Win_Drv - For Windows 98, 98SE, ME, 2000, XP - Ver 1.0.1.7

Medley - RAID Utility for Windows 98, 98SE, ME, 2000, XP - Ver
1.2.0.5



(3) SIL3112 - Driver for SIL3112 chipset (Without RAID Function)

Linux - Tweaking Red Hat 9.0 Driver

Novell - Novell Driver - Ver 1.07J

Win_Drv - For Windows 98, 98SE, ME, 2000, XP - Ver 1.0.0.47



(4) SIL3112RAID - Driver for SIL3112 chipset (With RAID Function)

Linux - Tweaking Red Hat 9.0 Driver

Novell - Novell Driver - Ver 1.07J

[Flasherly]

On Thu, 04 Jul 2019 17:50:09 -0400, Flasherly <Flas...@live.com>
wrote:

And also using similar for a ported (rather no dependency on *NIX)
boot manager/arbitrator. As I presume to understand it's originally
developed out of a *NIX environ, hence and but installs to a dev/disk
from within a DOS command interpreter. With the right controller, it's
feasible to attach SBM to every HDD/SSD's MBR, which can then be
defined, on an individual basis and relationship to other partitions,
from the first drive the BIOS points then to selectively boot, as well
as for relationship defined in each instance of a drive/device install
of SBM, to other drives. DOS withstanding, as well as the so-called
hidden FDISK /MBR command for erasing SBM (instead of using a
provision it comes with for un-installation.)

SMB is relatively dumbed down, to say from a standpoint of Microsoft's
distribution, or modifications, of GRUB and Windows 10 (presumptively:
I haven't set that one up yet, in De Loope De Loop.)

[T. Ment]

On Thu, 04 Jul 2019 19:51:33 +0000, T. Ment wrote:

> I can verify the DOS flasher DOES NOT work

I tried again, using the PCI slot near the AGP. I never tried that one
before. Now it works. Maybe I have some bad slots. I'll try them again
later to see what happens.

But I doubt they are bad, since other people reported the same problem.
Must be something else going on. I don't see how IRQ sharing could hurt
in DOS, when there's no activity from the sharing devices.

> Looks like they gave up supporting DOS after that version. The archive
> has no such BIOS for a 3114 card.

Wrong again. There's an old version 5073 for the 3114 card. It only has
a raid version, no basic. That's why I did not try it before. But I had
to try it just for completeness, before giving up. And what do you know,
it works with DOS.

[T. Ment]

On Fri, 05 Jul 2019 00:58:09 +0000, T. Ment wrote:

>> I can verify the DOS flasher DOES NOT work

> I tried again, using the PCI slot near the AGP. I never tried that one
> before. Now it works. Maybe I have some bad slots. I'll try them again
> later to see what happens.

The motherboard is an Asus P3V4X. It has 1 AGP, and 6 PCI slots. Slot 2
has a network card. Slot 1, and slots 3 thru 6 are open.

Slots 3, 4, 5, and 6 all fail. Slots 3 and 6 share INT-C, slots 4 and 5
share INT-D. But with no card in their partner slot, I don't see why it
fails. Four slots can't all be bad.

The sil3114 card only works in slot 1. Slot 1 shares INT-A with AGP, but
the AGP card has a jumper that disables its IRQ.

[T. Ment]

On Fri, 05 Jul 2019 05:52:22 +0000, T. Ment wrote:

> Four slots can't all be bad

I moved the network card to one of the trouble slots, and now the
network card fails too. Looks like this P3V4X motherboard is shot. I
hate to lose an old board. P3V4X is expensive now on Ebay.

Maybe I'll pull all the capacitors and check them before trashing it.
But that's tedious and not much fun. We'll see.

[Flasherly]

On Fri, 05 Jul 2019 15:10:43 +0000, T. Ment <t.m...@protocol.invalid>
wrote:

>I moved the network card to one of the trouble slots, and now the
>network card fails too. Looks like this P3V4X motherboard is shot. I
>hate to lose an old board. P3V4X is expensive now on Ebay.
>
>Maybe I'll pull all the capacitors and check them before trashing it.
>But that's tedious and not much fun. We'll see.

Picked up from Newegg, long ages ago, an ASUS K8N-E+ -- specifications
being: 5 PCI, an AGP, 4 SATA and 2 PATA;- At an half-off NewEgg
sideline offer, either someone returned or wasn't up to how to build.
Things more apt to happen, back before NewEgg became cutthroat about
RMA-ing on customer money, no less indiscriminately over putting
obvious crap merchandise, in order to dump it. So I ran that Asus
forever, which is long enough to became worse for the wear, distinctly
plagued with issues, and then some for an added measure. I wore that
MB into the ground, then to begin going through a regular routine and
series of power supply replacements, which helped for a spell longer,
which broke when I plugged into it the only one server-grade power
supply (Fortron/Sparkle), I've owned, whereupon the ASUS it ate it in
a poof of particularly evil-smelling smoke.

Not to mention expensive: A value that, to me, exceeded mere special
PS units. That PS was unlike anything before or since. It was
literally server grade all the way, smaller and very much like a
brick, (for low profile cases), and not an ounce less than as heavy.

I don't trust going back, as a rule, risking dated OEM stock and
rejects. Not that I haven't had good luck with limited purchases from
white-boxed MBs, DELL or HP sells unbranded for nickel-dime, from
special production runs lacking any characteristic nomenclature off
the Pacific Rim.

It's still at risk over what averages are, on a market, with the most
value, for DollarCostAverging all sales, has to offer. That's
industry bread and butter: The mean on recent production fabs and at a
volume significant enough for decent representative Quality Control.

Which is secondary to why I then switched to Gigabyte. I'd avoided
them since being told by those I respected of their value and quality.
And I'd even run into an odd instance of an Asus employee, who was
fired, who advised me of Asus deteriorating core values. Some day,
perhaps, I'm due up and I'll now have to try an ARock.

Looking at my Gigabyte in a UK link, the first thing I saw was a
negative comment on some outlandish support-chipset temperatures I
hadn't read to notice prior. I mean I've since run into it: recorded
my own at 140F extremities (one of note out of 4 Gigabyte MB
thermistors) on the MB chipsets. But it was satisfactorily cheap
enough of a MB for me -- for 8 SATA ports and more AMD overlap
support, among generations of CPUs, than is reasonably sane to
expect;- Probably, something that's addressable for enduser
modifications in cooling.

All in all, at one PCI slot, which is more along contemporary
standards (inasmuch a sixfold hard focus as adapted to solidstate
transfers on memory storage considerations, than CPU efficacy refined
for thread/core count on non-gaming measures), which, of course,
places your six PCI slots as outlandish, even if only one were out of
the loop for nothing much else spectacularly off kilter.

Indeed, the augmented computer of ISA and PCI MBs is increasingly one
falling behind, the shrinking micro-platform of advancing form, where
MBs increasingly are defined for self-containment within admissible
limits of their own specifications. A skew, to be sure: 3 PCI slots,
no less an rich estate, and just as likely to be a $300+ concession,
in today's pricing and, not to discount ASUS, more and above any
$50US, give or take a little, on value-oriented MB ranges.

[Paul]

The PCI bus is a shared bus. Each slot has a chip select
and a copy of a clock signal (which would be unique).

If the bus was shot, then chances are, moving a card to
any slot would not work.

If the motherboard is not positioned properly in the
tray, a card can be inserted crooked and cause a problem.
You have to ensure when placing a motherboard, that the slots
at the extreme ends are equally smooth for (unpowered)
test card insertion.

Paul

[T. Ment]

On Fri, 05 Jul 2019 13:28:17 -0400, Flasherly wrote:

> Which is secondary to why I then switched to Gigabyte. I'd avoided
> them since being told by those I respected of their value and quality.
> And I'd even run into an odd instance of an Asus employee, who was
> fired, who advised me of Asus deteriorating core values. Some day,
> perhaps, I'm due up and I'll now have to try an ARock.

I don't trust name brand reputation. I buy cheap junk and try to fix it
up. My K7S5A was an Ebay "parts or not working" special for $6, shipping
included. I replaced a couple of bad caps. It works.

[T. Ment]

On Fri, 05 Jul 2019 15:50:29 -0400, Paul wrote:

>The PCI bus is a shared bus. Each slot has a chip select
>and a copy of a clock signal (which would be unique).

>If the bus was shot, then chances are, moving a card to
>any slot would not work.

I don't know much about PCI design.

>If the motherboard is not positioned properly in the
>tray, a card can be inserted crooked and cause a problem.
>You have to ensure when placing a motherboard, that the slots
>at the extreme ends are equally smooth for (unpowered)
>test card insertion.

But I make sure the motherboard and cards are fully seated and making
good contact.

There are some capacitors near and between the PCI slots. I'll desolder
and check them, to see if I can find any bad ones. Something is messing
up INT-C and INT-D. INT-A and INT-B work.

[Paul]

T. Ment wrote:

> There are some capacitors near and between the PCI slots. I'll desolder
> and check them, to see if I can find any bad ones. Something is messing
> up INT-C and INT-D. INT-A and INT-B work.

Interrupts are swizzled on the motherboard.

https://www.coreboot.org/images/8/84/Routing_full.png

Most cards think they are accessing INTA. (A card doesn't
necessary need to use more than one signal, so INTA on the
card is the one that is used. The table in the motherboard
manual is likely how "INTA" on all slots is wired. The swizzle
diagram in the previous picture, is more important on expensive
PCI cards with say, multiple chips or functions.)

But the IRQ from the motherboard could be one
of many such signals. And the old manuals
had a table with the details.

Chipsets began to acquire more signals, making
a difference to how onboard chips were treated,
versus the "empty slot" wiring. The more modern
systems had "a few good choices" for tricky cards.

https://i.postimg.cc/2jWmqhzB/INTA-slot-wiring.gif

PCI Express don't need this, because they can
use "Interrupt packets", and then the wiring
doesn't matter. Everything is carried over
a physical star connection (hub in center inside
chipset, each slot has a "private" bus and does not
share resources). This makes PCIe virtually
bulletproof to the issues that existed on PCI.
Unless someone designs a bad motherboard with
bad controlled impedance values, the signal
quality on PCIe approaches "perfection". Nice
wide eye opening. No wiggly-jiggly shit like
on PCI.

Intel did 1000 hours of analog simulation (something
not common at the time of introduction), to ensure
that sufficient PCI slot combinations were tested so
that there would be no slot dependencies. I've done
some of those simulations, for projects at work,
the difference being, the stuff I worked on,
the chips were "permanently affixed" to their
slots, so I only needed to run one sim topology
to prove it worked. The bus could seemingly be
quite long, before it ran into trouble (longer
than the wiring we needed at the time).

The P2B-S would not allow a SATA card to operate.
The theory was, that the BIOS PNP code refuses to
map cards where the "type" field is not recognized.
The P2B was an IDE ribbon cable system, and nobody
knew of SATA at the time. Selecting PNP_OS=yes might
change the behavior, as the OS is then tasked with
recognizing cards. Normally, PNP_OS=no is the
recommended value. The P2B-S cannot boot from a DVD
drive either. Once it determines the drive isn't
exhibiting CD behavior, it won't even send out
any probes. The light won't blink.

It's hard to say what kind of table manners a P3V4X would have.

Paul

[Flasherly]

On Fri, 05 Jul 2019 19:53:53 +0000, T. Ment <t.m...@protocol.invalid>
wrote:

>I don't trust name brand reputation. I buy cheap junk and try to fix it
>up. My K7S5A was an Ebay "parts or not working" special for $6, shipping
>included. I replaced a couple of bad caps. It works.

I saw the same on my MB, about dozen instances of the Gigagbyte on
Ebay, two or three similarly listed for parts. $20, or $6 and a raise
the postal services gave themselves since then. But it's also on the
"brand reputation" circuit, or should be. I only got as far as UK's
Amazon listing while conveniently echoing off a virtual TOR network.
New and in a box for $60 and still boots DOS or a UNIX variant.

I trust Gigabyte. I don't know why the HDD controller balks, on my
particular model, over fragmentation routines and is too slow, nor how
a support chip can reach 140F. And I don't want, not especially, to
solder on it to find out more to fix it.

(OK, I will go so far to cut up old heatsink material, custom mount
and fit a small fan. And if this "new socket AM3+ MB" actually did
had more than one PCI slot, I might be tempted to put back in a SLI
controller, have 10 SATA ports, and kludged it up further to get a
smooth defrag routine.)

Besides Gigabyte was using solid-state capacitors ahead of the
industry curve, notably for a MB advertised as adverse to operating
under rugged conditions. Solid-state capacitors is "hot stuff" as I
understand.

There had been an issue, sometime prior to SS capacitors, with some
motherboard manufacturers buying "bad-batched" and substandard
electrolytic capacitors - ASUS being named pre-eminent in the
deceptive practice. I'm not sure to what extent subsequent
distribution may have been affected or how long it went on. As
mentioned the ASUS K8N-E+ is old enough as it, a 3rd generation prior
AMD socket, yet I was already raising my eyebrows at ASUS pricing
schemes, or I may as well instead have bought it new rather than at
half off from Newegg.

[T. Ment]

On Fri, 05 Jul 2019 15:50:29 -0400, Paul wrote:

> If the motherboard is not positioned properly in the
> tray, a card can be inserted crooked and cause a problem.

I desoldered one capacitor. My cheapo capacitor checker said it's good.
I soldered it back on the board. I didn't check any more. They all look
good, no bulges or leaks.

I put the motherboard back in the case, and the network card in one of
the trouble slots. Now it's working.

I temporarily used three screws to hold the motherboard in the case.
Maybe it will fail again, when I put the other three screws in. Maybe
that warps it just enough, to expose a defect not visible to the eye.

[T. Ment]

On Fri, 05 Jul 2019 18:55:37 -0400, Flasherly wrote:

> I trust Gigabyte. I don't know why the HDD controller balks, on my
> particular model, over fragmentation routines and is too slow, nor how
> a support chip can reach 140F. And I don't want, not especially, to
> solder on it to find out more to fix it.

I killed more than one motherboard learning how to solder.

It's not hard to get a capacitor off the board. You just add a little
blob of solder to the pins on the reverse side of the board, heat one
pin and wiggle with your hand on the opposite side of the board, heat
the other pin and wiggle, and repeat that cycle until you have it out.
That's the easy part.

But then you're left with two holes still plugged with solder. This is
what's hard. You have to clear the holes without damaging them, and they
are delicate.

The factory uses lead free high temp solder. With a common soldering
iron, it's not easy getting enough heat on the holes, to wick out the
factory solder without damaging the holes. It's doable, but tedious.
Once cleared, you put the capacitor back in, and solder it with leaded
solder, which is easier to work with.

Motherboards are not designed for easy repair.

[Flasherly]

On Sat, 06 Jul 2019 00:38:25 +0000, T. Ment <t.m...@protocol.invalid>
wrote:

>I killed more than one motherboard learning how to solder.
>
>It's not hard to get a capacitor off the board. You just add a little
>blob of solder to the pins on the reverse side of the board, heat one
>pin and wiggle with your hand on the opposite side of the board, heat
>the other pin and wiggle, and repeat that cycle until you have it out.
>That's the easy part.
>
>But then you're left with two holes still plugged with solder. This is
>what's hard. You have to clear the holes without damaging them, and they
>are delicate.
>
>The factory uses lead free high temp solder. With a common soldering
>iron, it's not easy getting enough heat on the holes, to wick out the
>factory solder without damaging the holes. It's doable, but tedious.
>Once cleared, you put the capacitor back in, and solder it with leaded
>solder, which is easier to work with.
>
>Motherboards are not designed for easy repair.

My impression of my first Gigabyte was, after I've went through so
many others, why can't I kill this one? (Seems it may have seen 8,
going on 10 years use;- I run them 24/7.)

Someone was listening and heard me, because a lightning storm emerged,
struck the transformer pig on a pole in my back yard, leaving me
looking at one fried modem and MB.

Maybe there's some truth to 3-year capacitor (liquid) average
lifespan. Five years is what I'd expect of a MB prior to Gigabyte. SS
caps are also now to be expected, more widely adapted for an industry
standard. MSI, a year or two ago, was trying on a comeback for
reputability to foist SS caps for MBs built for high-quality standards
at Military Specs. I left MSI to ASUS for reliability after a couple
bad MBs, long before SS caps.

https://www.gigabyte.com/webpage/8/article_02_all_solid.htm

[T. Ment]

On Wed, 03 Jul 2019 23:06:34 +0000, T. Ment wrote:

>> I need it for a DOS computer

> The Ebay seller told me it does not support DOS.

> Looks like my plan for DOS won't fly.

So back to Ebay for a Promise SATA/150 TX4. It can see the SATA drive,
but wont' use it, until you enter its BIOS and create an array -- which
kills any data on the drive.

I had already cloned the drive using linux DD on another computer which
has onboard SATA, and I didn't want to kill the data and start over. But
at this point I had no other option.

With nothing more to lose, I put the sil3114 back in the P3V4X, booted
from IDE, and ran updflash with BIOS version 5.5.00, the latest one. But
as before, updflash failed, and when rebooting, the computer hung again.
That problem was gone for a while, but it came back. Not sure why.

So back to the USB programmer to reflash. Then the computer booted OK.

Next I booted into the sil3114 BIOS and ran the low level quick format,
hoping to make the sil3114 happy.

I disabled the onboard IDE, then installed DOS 6.22 from floppy to the
SATA drive on the sil3114. Then I enabled IDE, and copied a 40 meg file
from IDE to SATA, all in DOS. That worked too. No errors as before.

Disabled IDE again, booted DOS from SATA. Copied the big file again,
from one directory to another. Very fast. about 1.5 seconds, compared to
10 seconds with the IDE drive and P3V4X bios.

So my conclusion is, you can't clone the drive on some other computer
and expect it to work in DOS with the sil3114. You must make the sil3114
happy by using its low level format first. I don't know what INT13 magic
is going on there, but it seems to work.

I will do more testing after reinstalling the other partitions, but DOS
was the only problem before, so hopefully it will all work.

[Paul]

Even for a JBOD disk (not part of an array),
some RAID cards insist on adding metadata that
says the hard drive is not part of any array.

Some other brands, the firmware looks for metadata. If
is is missing, the drive is assumed to be JBOD.

Not that this has anything to do with a file copy
malfunctioning part way through. That doesn't make
any sense.

Metadata and configurations are things that block
mounting or seeing stuff. For example, if I plugged
an Intel disk into a Promise card, the first
partition would disappear. Plug it back into an
Intel port, the partition comes back. That's an example
of metadata at work, and something which isn't compatible
(or the second controller ate), preventing some info
from being used.

A lot of cards put the metadata at the end of the disk,
but a few put it at the beginning. I suspect my disappearing
partition case, plays into that. (

Paul

[T. Ment]

On Tue, 16 Jul 2019 16:32:09 -0400, Paul wrote:

>Even for a JBOD disk (not part of an array),
>some RAID cards insist on adding metadata that
>says the hard drive is not part of any array.
>
>Some other brands, the firmware looks for metadata. If
>is is missing, the drive is assumed to be JBOD.
>
>Not that this has anything to do with a file copy
>malfunctioning part way through.

A RAID card BIOS has its own INT13 code. It can't rely on motherboard
INT13 code, which may not support large drives. But that's just for DOS.
Linux or Windows don't call BIOS, their device drivers do all the work
internally.

>That doesn't make any sense.

With DOS, it does. The sil3114 INT13 code does not work right in DOS
without something on the disk put there by its low level format. I don't
need to know what, I just want it to work.

[T. Ment]

On Tue, 16 Jul 2019 21:22:59 +0000, T. Ment wrote:

> The sil3114 INT13 code does not work right in DOS without something
> on the disk put there by its low level format.

Oops, wrong again.

The quick low level format just wipes the beginning of the disk. I don't
know how many sectors it writes, but it's enough to wipe the partition
table.

I now see that when DOS creates its maximum size partition on a wiped
disk, it chooses a different CHS geometry than linux fdisk. Some kind of
interaction between the disk and the sil3114 BIOS tells it to, I think.

DOS sets heads=16, but linux fisk sets heads=255. That's what causes the
DOS copy failure. 16 heads is the limit if you want to use DOS with this
card.

With DOS, cylinders are limited to 1024, no matter what the heads and
sectors per track are. So with heads=16, sectors=63, and cylinders=1022,
your maximum DOS partition is about 500 MB. My problem is, I have a DOS
partition of 2GB, the maximum size with heads=255.

If you don't need a huge DOS partition like mine, the card works.

[T. Ment]

On Wed, 17 Jul 2019 01:15:49 +0000, T. Ment wrote:

>With DOS, cylinders are limited to 1024, no matter what the heads and
>sectors per track are. So with heads=16, sectors=63, and cylinders=1022,
>your maximum DOS partition is about 500 MB. My problem is, I have a DOS
>partition of 2GB, the maximum size with heads=255.

>If you don't need a huge DOS partition like mine, the card works.

But only up to some disk size limit, unknown to me.

I'm not trying to recommend this card for new builds. I just wanted to
know if it works on older PCI bus computers that don't have any SATA. It
works on one of mine. YMMV.

[T. Ment]

On Wed, 17 Jul 2019 01:15:49 +0000, T. Ment wrote:

> DOS sets heads=16, but linux fisk sets heads=255. That's what causes the
> DOS copy failure. 16 heads is the limit if you want to use DOS with this
> card.

Some final observations:

With the DOS partition working right (heads=16), I flashed the BIOS back
to 5.0.73. The flash worked. That's the trick to prevent updflash from
failing. You must use a DOS partition sized within the limit the sil3114
BIOS supports. To ensure that, use DOS fdisk on a wiped drive and let it
choose its maximum partition size.

Even if you don't use DOS, create a DOS partition anyway, so linux fdisk
uses the same geometry when adding other partitions. Otherwise, you will
get weird linux errors when booting.

Using the older 5.0.73 BIOS, I wiped the disk and let DOS fisk create
its maximum partition again. This time, it set heads=255 to give a 2GB
partition like linux fdisk.

The newer 5.4.0.3 and 5.5.0.0 probably support larger hard drives, and
heads=16 was a trade off required for that. I don't know what the hard
drive size limits are for any of these BIOS, but they all work with my
320GB.

If you get a card with a 5.4.0.3 BIOS and don't care about DOS partition
size, don't mess with it. Flashing with updflash is risky. It won't work
if your DOS partition head count is not compatible with the current BIOS
on the sil3114 card.

[T. Ment]

That's all true, but I later learned more:

https://www.vogons.org/viewtopic.php?f=46&t=49471

After flashing the right BIOS, and following the instructions to create
a JBOD array, given in the thread linked above, NOW IT DOES WORK IN DOS
with a heads=255 partition.

The older 5.0.73 BIOS does not have a JBOD option, but the 5.4.0.3 and
5.5.0.0 do. They show it as "concatenation" instead of JBOD, same thing.

What a saga to make this card work in DOS. Maybe this is the final word.