K7S5A + SCSI

[BipolarBill]

Hey everyone - I need ideas.

Broke out a brand new K7S5A and plugged in this brand new LSI u160 SCSI card. I set up RedHat, reboot and :( . I then notice that I've never seen the SCSI BIOS load during POST (like all add-in storage cards do). So, we have a card that is seen by BIOS and assigned an IRQ. I can get into the basic config with a floppy-based program, but I cannot access the SCSI BIOS and setup because I never see it flash by. Because there is no resident SCSI BIOS, I cannot boot from the drive that's on this card.

Got the latest BIOS and everything else seems fine.

[Peter M]

Feel free to paste our private conversation here, to let the others where we are.

[BipolarBill]

I'm gonna try to "kick-start" this with a Promise card. If that doesn't work, I gotta assume defect.

Peter's first response:

OK ... since I have an older LSI SCSI card in my own K7S5A (an 21002 dual channel card), here's what to do:

First of all you need to NOT have a February or June BIOS in your K7S5A. The current 10/29 one is fine.

That settled, you should have the LSI SCSI BIOS do its "Press Ctrl-C ..." routine, scanning for devices, and find some. You should press Ctrl-C once and browse through the menus (don't forget the global menu reached by F2), to make sure everything is the way you want it. There's a Device Selection menu too, where you can see which ones have been detected; you also select which IDs to scan at boot, which reduces boot time.

With that done, the "BBS-x" slots in system BIOS are replaced with the name strings that the LSI BIOS registers for the drives it found, something like "ID 68-SEAGA". Select that, and you're done.

No tricks, other than having a system BIOS release in which ECS did not screw that up

Good luck.

Me:

Thanks Peter, but I can't access the LSI BIOS/setup at all. The ECS assigns IRQs, but won't let me at the LSI controls - it just bypasses the SCSI card altogether.

I'll look into a replacement and leave this one as IDE.

Peter:

Do you have any other cards in there that consume ROM space (PCI IDE or something?) Or did you fiddle with the shadow/cache settings? If you don't even get the Ctrl-C prompt, dowloading the SCSI BIOS image and flash utility from www.lsilogic.com could also be worth a try. See what that utility says - and if you have, use Microsoft's good old MSD.EXE DOS diagnostics tool to show you the UMB memory map.

Given that the U160 card still uses the exact same BIOS as all the older and slower LSI cards, I'm absolutely certain this does work if done right. Don't you dare give up!

Me:

quote:
--------------------------------------------------------------------------------
Do you have any other cards in there that consume ROM space (PCI IDE or something?)
--------------------------------------------------------------------------------

No Sir. Just a USR Winmodem.

quote:
-------------------------------------------------------------------------------
Or did you fiddle with the shadow/cache settings?
--------------------------------------------------------------------------------
No - strictly default there.

quote:
--------------------------------------------------------------------------------
If you don't even get the Ctrl-C prompt, dowloading the SCSI BIOS image and flash utility from www.lsilogic.com could also be worth a try.
--------------------------------------------------------------------------------
I would, but the card works first time, every time in every other PC I have.

quote:
--------------------------------------------------------------------------------
See what that utility says - and if you have, use Microsoft's good old MSD.EXE DOS diagnostics tool to show you the UMB memory map.
--------------------------------------------------------------------------------

Config.exe showed the drive and the SCSI bus in it's entirety. I just can't evoke the LSI BIOS. I'll try to use MSD.EXE, but if the ECS board isn't going to cooperate and let the LSI do it's thing, I don't know what I can do with that info.

quote:
--------------------------------------------------------------------------------
Given that the U160 card still uses the exact same BIOS as all the older and slower LSI cards, I'm absolutely certain this does work if done right. Don't you dare give up!
--------------------------------------------------------------------------------

Who's giving up? I'm gonna go around the dammed obstacle!

Peter

What we need to find out (from MSD) is how much UMB space is being consumed e.g. by the graphics card's ROM. The LSI SCSI ROM needs 64KB free during initialization, and if you don't have that, then system BIOS won't be able to map and start it.

My K7S5A has a Radeon 7500LE in the AGP slot, an LSI 21002 in the PCI slot right below it, plus a Realtek NIC and a TV card in the lower two. Onboard LAN, USB and sound are in use too.
Can't say more than: From all I know and from personal experience with this combination, it SHOULD work. I'd try doing a "Load Optimal Defaults" in system BIOS too ...[/i]

[BipolarBill]

OK - I fired up MSD and here are the machine results:

ROM BIOS - 65536

Option ROM - 32768

Video ROM BIOS - 64512

As you can see, there's no SCSI ROM loaded. :(

[BipolarBill]

You might be interested in the ECS response:

Sent: Tuesday, November 26, 2002 11:22 AM
To: Support
Subject: K7S5A + LSI U160 = grief

Hello and thanks for reading this.

I'm have a hard time getting the latest BIOS for the K7S5A to boot the BIOS on an LSI u160 SCSI card. It just skips right over the card as if it weren't there. Is there some switch that I need to use to address this or is it a BIOS limitation?

Thanks.

From the desk of
Bill DeVercelly
Win2K MCSE and Pseudo-Geek

---------------------------------------------
Dear Sir/Madam:

I don't know what you mean about just skips right over the card mean ?

We been test adaptec U160 without any problem. Please try change differnet one then try again.

"""""""" Please include all the previous correspondence or case number when replying """"""""

--------------------------------------------

It is remarkable that they answered the same day, but it's the typical rude-**** Taiwanese answer - truly useless.

[BipolarBill]

Got it! The boot ROM for the SiS onboard LAN was blocking the LSI BIOS.

Now - how can I use the darned onboard NIC without the boot option? The only way to get around it is to disable it altogether.... :mad:

[$1500-P4 gamer]

Dont know. Maybe try turning off plug n play and let bios assign the irq's etc. from start. Worth a shot, prob. not gonna help. Do you have a reg. pci nic around. Just to see if its cause its internal or cause its trying to boot off a netwrok device. Also hav eyou tried making sure that the option to boot off other devices is to no and etc.???

[Peter M]

So why does this work on mine then? I do have the onboard NIC enabled. Strange. Again, how much UMB space does your VGA card eat? If no MSD.EXE, you can check in Windows Device Manager, look for memory resource usage from 000C0000.

Next time you want a useful answer from ECS, you better hop over to the PC-Chips Lottery site BBS, and post your question there - ECS UK have an eye on us there :).

regards, Peter

[BipolarBill]

Originally posted by BipolarBill
OK - I fired up MSD and here are the machine results:

ROM BIOS - 65536

Option ROM - 32768

Video ROM BIOS - 64512

[Peter M]

Eeek, 63 KBytes of VGA BIOS *shudder*. What's that? The 32K thing is the netboot ROM I suppose (use Utilities-Memory Browser, and press Enter on a listed item to see some strings from inside it).

Doing the math ... 63+32+64 > 128, you're busted ... If putting the SCSI card into another slot doesn't get it enumerated before the onboard LAN, I guess I'll have to talk to my ECS contact again. (The SCSI ROM image shrinks to 16 KBytes after init, which then would leave enough room for the netboot ROM to come after it.)

We need a setup item that lets us disable the LAN boot ROM without disabling the LAN function.

regards, Peter

[BipolarBill]

That's a GeForce4 MX440. I could "downgrade" it to a GF2 GTS.

Those memory numbers are with LAN enabled - the "Option" ROM is the ECS NIC! With it disabled, you're spot-on - the LSI uses 16K. ;)

Yep - we need to disable the NIC ROM apart from the adapter itself. That's a tall order from such a rudimentary BIOS. It looks like MSN designed it. :p

I emailed those bozos in California. I don't expect much from them.


*sidenote*

I managed to install Redhat before I got the adapter to boot. Now the OS is pooched - "ID "x" respawning too fast: Disabled for 5 minutes"

Looks like a reinstall. :(

[BipolarBill]

Well, i replaced the MX card with a GF2 Ti and everything works now. The new vid card only uses 49K and allows everything else to drop. The SiS NIC only uses 16K now. I also now have a boot selection called "#48 LUN 6".

Remarkable. Thanks for the ideas, Peter. I guess we can call this and incompatible video card.

[BipolarBill]

Latest response from the California nuts and fruits:

Dear Sir/Madam:

You should ask LSI special Real DOS mode utilty allow you change IRQ and address. sound like they are using same IRQ or same address.

When you disable onbonard LAN , the boot from LAN function will disable too.

IRQs again. :rolleyes:

[Peter M]

Ah, lovely blame-IRQ-sharing support tactics.

All together now: B-U-L-L-S-H-I-T !!!

Well, the VGA card isn't exactly incompatible, it just eats an unreasonable amount of precious Upper Memory. We had an exact copy of the issue at work recently (also with an NVidia VGA) - but there the LSI SCSI chip was old enough to allow us to revert to a much older revision of the SCSI boot ROM that is only 32K. With the 53C1010, you can't do that.

The other trick we could pull is AMIBCP the system BIOS, removing the embedded SiS LAN boot ROM so the system BIOS simply can't use it anymore.

regards, Peter

[BipolarBill]

Originally posted by Peter Missel
The other trick we could pull is AMIBCP the system BIOS, removing the embedded SiS LAN boot ROM so the system BIOS simply can't use it anymore.
Oh - you mean strip it with something like CBROM? I'm game - I have no need for boot to LAN and I doubt that anyone else I know will.

[Peter M]

CBROM is for Award BIOSes, but yes, AMI BIOSes have a modular structure too. Things like the LAN boot ROM for the SiS 900 are just binary modules with no code dependancy to system BIOS - the association just comes from the PCI device ID they're meant for.

So to get rid of an autoloading, embedded PCI ROM, one can associate it with a non-existing PCI device ID, or take it out altogether.

regards, Peter

[BipolarBill]

Latest BS from CA:


Dear Sir/Madam:

We only support US customer.

We has been test this board with Geforce Ti4200 and TI4600 with Adaptec 160 SCSI card without any problem at all.

Thanks your information.

"""""""" Please include all the previous correspondence or case number when replying """"""""

Regards,


ECSUSA Support Team

What sh!theels these people are! My response:

Are you people stupid or mean? I am a US Customer! Look at my e-mail address. The last time I checked, Houston was in Texas and Texas was in the US.

I don't give a **** what works in your samples. It doesn't work in MY PC. I am the customer. I give your company money so that they can pay you.

Remember what happened to PCChips? Yes you do. Poor customer service will kill ECS too. Keep up the rotten service.

Screw 'em. :mad:

[Peter M]

Ummm ... it's rather like the PC-Chips group assimilated ECS, not the other way round :)

Besides, in their defence, it might very well be that other GeForce cards use smaller footprint VGA BIOSes. Usually the VGA chip makers offer a selection, depending on what interfaces and features the card making OEM wants to support.

regards, Peter

[Peter M]

... and it gets weirder ...

I just looked into mine. I have:

C0000 VGA, 52KB
CD000 SCSI, 16 KB
D1000 LAN, 32 KB

Which means that the SCSI card (in slot 1 btw) does get enumerated before the onboard LAN - else the SCSI ROM wouldn't be on a lower address.

Which in turn means that your system should have it the same way round?!?

regards, Peter

[BipolarBill]

I re-enabled the onboard LAN, but I don't think that it's going to work. I can boot, but RedHat won't give me a GUI.

Here's the current spread:

F000 ECS BIOS, 65536
D000 Option ROM, 32768 (LAN)
CC00 Option ROM, 16384 (LSI)
C000 Video ROM, 49152

I just ordered an MSI Ultra2 - 166MHz Athlon support for $81. Good board. I need a good board - this is gonna be a server. I can sell the ECS as a desktop.

[Peter M]

Well given that, you are on the same grounds as I am, the LSI SCSI gets enumerated before the SiS LAN. Meaning that with the larger VGA ROM back in, the SiS LAN ROM should fall off the edge, not the LSI SCSI.

Quality wise, you can't get a more mature board than the K7S5A. Besides, with the same graphics card, you'll see the same problem on any of today's mainboards - there simply aren't any that provide more than 128 KBytes of UMB.

regards, Peter

[BipolarBill]

I'll kee working on it. ;)

[$1500-P4 gamer]

I see that once again Nvidia is pushing the pc spec above an beyond its limits. AGP now UMB this is ridiculous. How could they not know that! They ahve to have a guy like you (peter) around. Or are they just taking guesses on the size of mem the bios takes up for these cards? If its gonna be like that on all mobo's Id say its a vid card manufac.'s problem noone elses but ours.:(

[Peter M]

Well yes, the size of the VGA BIOS is entirely up to the company that makes the VGA chip. As an OEM putting the VGA chip onto a card or mainboard, you use the binary ROM image as provided, configure a few data areas for choice of RAM size, type and frequencies and other misc stuff like supported output channels, and that's it. Sometimes you get to choose smaller ROMs with fewer features (like VESA 3.0 support, DVI or TV-out et al), but normally there's one ROM flavor and that's it.

That said, a PCI or AGP ROM image is allowed to be up to 64 KBytes ... but after firing its chip up, it is supposed to ditch the initialization code and shrink to whatever is needed at runtime - as demonstrated by LSI's SCSI BIOS, which is 64 KBytes initially but shrinks to 2-16 KBytes for runtime, depending on what kind of devices it found.

NVidia's VGA BIOS ROM currently seems to be at 62-63 KBytes _runtime_ size, which isn't beyond any specification - it's just extremely impolite toward the non-VGA cards that come after it. Given the popularity of LAN boot, PCI IDE and SCSI controllers and soon bootable FireWire cards, this is asking for trouble. But we know THAT already, don't we :mad:

But at the end of the day, what's ACTUALLY biting us here is the legacy of the PC. Card initialization and runtime support, as well as the system BIOS runtime image and VGA compatibility frame buffer still all have to take place in "real mode" address space - limiting even nice and shiny 2-GByte-RAM boxes to the one MByte that made the original 8088 processor's world. Add-on card space in there originally was 192 KBytes (from C0000 to EFFFF), but stuff like ATAPI, USB, PnP, PCI and whatnot made the system BIOS grow down from its F0000..FFFFF space all the way down to E0000. Now all we have left for card ROMs is 128 KBytes. Ridiculous, yes. It's not the only leftover from 1982 or earlier in our oh-so-new boxes. Don't ask.

regards, Peter

[BipolarBill]

I'm curious, Peter - do SGI and Sun machines have the same limitations?

[$1500-P4 gamer]

So basically nvidia isnt breaking the rules just kinda being rude!:mad: Cant we just expand the 1 Mbyte to 2? What of current dual bios's. Same limitations? If one disabled USB would you get back more UMB??? Trust me I believe you about our new pc's legacy problems. I was all up in joy when my P4 mobo came with NO ISA slots. About time they killed that. I mean it was truely hardware run for like modems (no software win modem junk yet) etc. But none the less slowwwww and supporting that legacy ISA, slowed the whole system down from what I've been told. This true?:confused:

[Peter M]

Bill, no, those machines, and the Apple and Alpha hardware too, have often been re-architectured, with only the operating system abstraction remaining compatible.

Our x86 platform still is HARDWARE compatible to the original IBM AT and, in parts, even to the IBM PC.

gamer, well, the original 8086 processor couldn't reach more than one megabyte of memory address space. It was layed out like this:

00000..7FFFF RAM
80000..9FFFF RAM expansion
A0000..BFFFF video card frame buffer
C0000..EFFFF expansion cards
F0000..FFFFF system BIOS

Still maintaining compatibility, even your shiny new Athlon or P4 fires up in 8086-compatible world model, so the above is all we have. The rest is explained further above.

And yes, although the ISA bus physically is gone, there's still plenty of ISA like legacy bits and pieces in there. From the keyboard/mouse controller (a 1977 general purpose IO controller, 8-bit ISA) including the dreaded GATEA20 function (look it up), floppy disk controller, serial and parallel ports, real time clock and CMOS storage down to vital stuff like the interrupt controllers and system timers, it's all still the same **** as in the original IBM PC and AT - which in itself was hastily slapped together from proven (read: old) parts back then in 1981/83 respectively. Innovation? Sorry folks, not in this one.

Yes, chip design has advanced, so you won't physically see all that stuff on the mainboard ... but it's there, integrated into the south bridge and/or in a super-I/O chip. The latter isn't connected through real ISA anymore (it's a fast-and-narrow "LPC" bus - just easier to route and using less pins, but no faster than ISA), but for the sake of compatibility *sigh*, the south bridges still make that look like ISA.

regards, Peter

[$1500-P4 gamer]

Hmm, I see. Thanks peter that was good info there. So basically while we are cranking the speed up on video mem and cpu the actually roots of the pc are still the old 8086 junk. God thats sad. We have some major bottleknecks and etc. that are just being accepted as part of the pc turf. LOL these guys need to get the ball rolling. 90% of that must be the problem with hardware compatability. No wonder with apple its plug it in and let her rip.:eek: Some rich type genious needs to sit down and redo this junk we have. Then he'd be even richer.

[Peter M]

What we need to have is operating systems that don't found on DOS anymore. After all this is what the hardware compatibility thing is all about - DOS didn't provide a full hardware abstraction, so you could never pull the rugs out under the operating system without breaking the application software.

Now that we finally have NT-based Windowses all around, and since Linux is multi-platform anyway, the PC world would be ready for a new hardware architecture underneath.

That's what you thought.

In reality, just look at how many people whine about their old games no longer working because someone dared make sound chips that aren't "SoundBlaster compatible" anymore ... just to get an idea of what would happen if you changed the platform altogether.

The step to a new platform should have been made right with the move to 286 processors IBM AT, at least with the 386, but noone dared lose their base of customers and rewrite their applications back then.

Now, it's way too late. Everything that'll happen in the PC area will be backward compatible to computing stoneage. Itanium has GATEA20. Need I say more?

[$1500-P4 gamer]

O.K. I see the problem now. But alas wouldnt it be a safe call now to update to say a 386 compat. instead of 8086. Cuase noone uses that old of software anymore. Heck even a 486 compatable system would drive most everyones needs. I see what you are saying about DOS aswell. It doesnt run the hardware by itself but relys on the bios to. So NT like Xp and etc. runs it through the OS with the death of Dos. But there again your right, lots of older progs are a no go on nt Os's. What a jumbled mess they left this become. Is ther any hope??? Sooner or later BP's hoggy Nvidia card issue is gonna become self evident on more pc's. I predict utter chaos.:(

Thanks again. As always you are a fountain of bios info. I always learn somthing new chatting with ya.;)

[Peter M]

Ummm ... you didn't quite get it. The problem with DOS is that it exactly does NOT provide a complete, abstract software interface to the hardware below it. So application software has to go directly onto the hardware to get the interesting things done.

If an operating system provides a complete abstraction (like NT, Linux or MacOS do), it's quite easy to get things running on completely different hardware. Port the operating system's lower levels to use the new hardware, reCOMPILE the higher levels of the OS and the user applications (not needed in case the processor architecture is the same), and you're set. With incomplete abstraction, you have to reWRITE the application software to get them to run on new hardware.

That's what kept the x86 world from building really new machines. Starting with the 286 AT and not stopping with PCI, AGP, and whatnot, it's all patchwork piled on top of the original PC design IBM slapped together more than 20 years ago - which was pretty awful even by the standards of back then.

[BipolarBill]

Wow. I feel like I've been dipped in dog doo. :( I'm sorta thankful now that this pile of silicon and copper works at all!

Thanks for the lesson, Peter. :cool:

[BipolarBill]

I'm wondering - is there any hope that the 64-bit platform will rid us of these legacy systems? I see your comment about the Itanium and GATEA40 and it has me worried...

[Peter M]

Itanium sort of tried that. Doesn't work out. Market penetration: Near zero. Guess why.

AMD Hammer platform maintains hardware compatibility (and, consequently, all of those "neat" restrictions). As discussed above, you need to have that to sell your stuff into the PC marketplace.
All we can be happy about is that at least the legacy I/O is slowly giving way to PCI based stuff like USB and FireWire. Oh, I almost forgot, IDE is ISA too ... on steroids, but still ISA. On the desktop that is ... in the embedded marketplace, customers still want their mainboards to have at least four 16550 compatible :rolleyes: UARTs.

[BipolarBill]

I know - I keep getting questions at USBMan about how to print to USB and telnet through a USB modem from a DOS window. It takes restraint to not scream "Get a 32-bit application!". :p

[Peter M]

Well, everywhere outside the PC business that'd be no problem. If you want to use new gadgets, get a new box and new software, and most importantly, get over it. If you need to run old software, keep the old box.

In the PC world, people get mad when their new 2002 box doesn't run their "special" application from 1989 written for a 9-pin printer model from 1986. There is such a high expectation on that "compatibility" thing, essentially making the big jump impossible. Just look at how long USB took to come up to speed - we've had working USB controllers on mainboards mid-1997, Windows support in late 1998, and now, approaching 2003, we're still not there. PS/2 keyboard and mouse port, game controller, serial and parallel port simply refuse to go away.

You should see how much of a pain in the **** all this is in mainboard design and BIOS engineering.

[BipolarBill]

I can only imagine... :(

[$1500-P4 gamer]

"You should see how much of a pain in the **** all this is in mainboard design and BIOS engineering."

I leave that to you peter. It would stress me out just learning it muchless doing it and learning all the new stuff as it comes out. Nope, not my idea of fun.
;)

[besplin]

I beleive I am running into the same thing when trying to install a Silicon Image chipset 0680 PCI RAID card. I ran MSD and I have 23K of Upper-Memory left BEFORE inserting the RAID card.

When I insert the RAID card the screen goes black, just before Windows loads. I have tried changing IRQ's, no luck. I have not tried disabling the onboard LAN, I will try that next.

Any other ideas on how to get more of the precious UMB to be available? Here's what might be taking it up and my basic configuration:

Mainboard: MSI K7N2-Delta L
AGP Card: GeForce 5200 (128 MB of RAM, I have a BIOS setting saying it is a 128MB card)
1 PCI slot filled with a USB 2.0 card, removing it does not affect the UMB per the MSD program.

Other stuff:
1GB RAM
200 GB drive
DVD burner

After having my Seagate Baracuda 200 GB drive fail (6-months old) and having to painfully recover data from non-destroyed sectors, I'd REALLY like to get my dual 200 GB drives working as a RAID-1? Mirror, if I can get this #$%%@$%$ to work!

[besplin]

I beleive I am running into the same thing when trying to install a Silicon Image chipset 0680 PCI RAID card. I ran MSD and I have 23K of Upper-Memory left BEFORE inserting the RAID card.

When I insert the RAID card the screen goes black, just before Windows loads. I have tried changing IRQ's, no luck. I have not tried disabling the onboard LAN, I will try that next.

Any other ideas on how to get more of the precious UMB to be available? Here's what might be taking it up and my basic configuration:

Mainboard: MSI K7N2-Delta L
AGP Card: GeForce 5200 (128 MB of RAM, I have a BIOS setting saying it is a 128MB card)
1 PCI slot filled with a USB 2.0 card, removing it does not affect the UMB per the MSD program.

Other stuff:
1GB RAM
200 GB drive
DVD burner

After having my Seagate Baracuda 200 GB drive fail (6-months old) and having to painfully recover data from non-destroyed sectors, I'd REALLY like to get my dual 200 GB drives working as a RAID-1? Mirror, if I can get this #$%%@$%$ to work!

[MJCfromCT]

besplin,

Welcome to SysOpt :) In the future you may want to create your own thread instead of reviving a 2 year old one, that way people will be more inclined to assist.

[fishybawb]

Hey, at least someone knows how to use the "search" feature! :p

[BipolarBill]

Have you tried just disabling the onboard LAN boot-block BIOS? You can also try a different video card - as I did.

Try the card in the lowest PCI slot too - as well as the highest.

If disabling the LAN adapter doesn't help, it's not a UMB issue.

[besplin]

I did disable the LAN adapter that is built into the mainboard. That got me an extra 9K of UMB, whoopie! It was not enough to get the RAID card to work.

Next up is trying a different graphic card. I'll remove the AGP 128K GeForce 5200 and try a 32MB PCI card, just for "fun".