S3 Savage4 and ALi chipset problem

[Tonhovsk]

It's an old card by today standards, I know, but I'm curious about it: my savage4 (diamond stealth 530) simply locks, in a motherboard using ALi chipset, on a Pentium 166. In another machine (with a Pentium 166 too) it works well.

Has anybody known some incompatibility between these devices?

[rmanet]

welcome to sysopt :t

could you give us more details? mobo brand, OS, what driver you're using, prior video card in the system or was it onboard, did you uninstall prior drivers as needed in safe mode? Tried another slot? What settings in BIOS for video?

post back so we can help - can't find anything on ALI chipset conflicts but some mobos simply don't like certain video cards - it's happened to me once or twice and never found the real reason and couldn't pin it down to just a chipset conflict

[Tonhovsk]

Ok man, here are the details:

mobo: ATC 5130
OS: Windows 98
driver before this: s3 trio 64 1 MB (offboard)

About the prior drivers, there was no need of worry: I installed a fresh new Windows 98 ( I thought the problem was with windows 95).

Well, I think I didn't try another slot. The card worked perfectly with standard VGA driver; it's when I installed the savage driver that the system locked up before the first Windows screen.

And about the bios, I let it in bios defaults (it's an AWARD bios). IRQ assigned. Is there any important setting for video?

One more thing: to my relief, a voodoo 3 2000 worked perfectly in that system, I think in the same slot ( I'm not sure).

Thank you for the interest!

[Tonhovsk]

In time: my savage is a PCI card, 8 MB.

[Peter M]

Is it a Savage at all, or is it an older chip (Trio, Virge, etc.)?

[Tonhovsk]

It's a s3 savage4, 8 mb of RAM memory.

[Tonhovsk]

Last night I spent some time trying to solve this inextricable problem.

First I tried slot nº 4: nothing new, again the system froze before the first Windows screen.

Then I tried slot nº 3: a little difference, the system froze in the first Windows screen, when showing the box "New Hardware Found".

I noticed that Windows is a bit crazy: in safe mode, the video card IRQ is one: when I let Windows run with standard VGA, the IRQ is another one.

[Peter M]

That would indicate the board's BIOS hasn't given it an IRQ properly, and Windows is trying to reassign it.

First thing to do is set "PnP OS installed" to No in BIOS.

Incidentally, I have an A-trend 5130 right here on my lab desk, being replaced because its L2 cache has died and its IDE channels have become flaky. If there's anything I can assist you with this (like, a manual or the latest BIOS for it), let me know.

[Tonhovsk]

Hey man, I think you guessed: my mobo is an ATC 5130 ( I'm not sure, I'm going to confirm tonight), and the chipset is an ALI - I can find the model with SANDRA, if it helps.
Unfortunately, the "PNP OS Installed" is already set to NO, I think.
A Cirrus Logic modem is installed, using IRQ 11, and I don't use it. Can it be the cause of the conflict?

[Peter M]

No. "IRQ conflict" is an urban myth, not an actual problem - amongst PCI devices. Unless that modem is an ISA card, don't bother pursuing that route.

btw, I didn't guess your mainboard, you told us in your 2nd post right there ;)

Chipset is ALi Aladdin IV+ btw. Does it freeze with the HDD activity light stuck on?

[Tonhovsk]

No. "IRQ conflict" is an urban myth, not an actual problem - amongst PCI devices. Unless that modem is an ISA card, don't bother pursuing that route.

btw, I didn't guess your mainboard, you told us in your 2nd post right there ;)

Chipset is ALi Aladdin IV+ btw. Does it freeze with the HDD activity light stuck on?

I like this statement, this man knows about computers.

Last night I tried all the four slots: same problem. And, answering your question, the HDD activity stays off when the system freezes. And "Pnp OS installed" was set to No.

The motherboard bios version is 1.0, do you think that an upgrade can help? Is it worth a try?

[Tonhovsk]

I like this statement, this man knows about computers.

Last night I tried all the four slots: same problem. And, answering your question, the HDD activity stays off when the system freezes. And "Pnp OS installed" was set to No.

The motherboard bios version is 1.0, do you think that an upgrade can help? Is it worth a try?
Just to be correctly understood: an update (not upgrade) of the bios can help?

[Peter M]

I'll have a look at what BIOS revision is in that half-dead 5130 I have. I faintly recall something like 1.4.
The floppy interface on that board still works, so I'll be able to retrieve it, whatever it is.

The 5130 does btw run processors up to K6-2/500 (2.2V, 6x 83 MHz, retaining 33 MHz PCI). Also, given the added cache tag RAM (socket in upper right corner), also covers at least 256 MBytes of RAM with its L2 cache.

It's quite a good board for its day, actually.

[Peter M]

In case you wonder: I was asking about HDD because the ALi 1543 rev. B southbridge has a known bug, in that it announces UDMA-33 capability, but doesn't actually work in that mode. The board's BIOS "knows" and limits IDE to MWDMA mode 2 (16 MB/s), but an OS driver might try UDMA-33, hanging the machine soon after.
I don't know whether ALi's IDE drivers are still available (www.uli.com.tw). I know for sure that Windows builtin drivers don't know about the bug, while Linux's builtin IDE drivers /do/ know.

[Tonhovsk]

But if the problem was in the ALi southbridge, the system would work with the voodoo3, as it did? (Post #3)

[Peter M]

The IDE bug is nothing to do with your VGA problem.

Just one more thing before we try the BIOS update: See whether you have "Delayed Transactions" disabled. I faintly recall one needs to keep that disabled when the USB controller is enabled.

[Tonhovsk]

"Delayed Transactions" was already set to disabled, so nothing new here.

This weekend I tried the savage in another machine, equipped with a relatively new (2005) motherboard, the ASUS A7N8X-X: the card worked perfectly.

I go the ATC 5130 manual. Don't you think it's a good idea read it from the very beginning, searching for a possible mistake in configuration, or would we go to the bios update right now?

In time: the bios version in my machine is not 1.0 as I said before, it's 1.0 03.

[Peter M]

Mine has 1.0 4.1 or somesuch. I've thrown assorted graphics card at it, no problem so far.

[Tonhovsk]

Ok, man, can you send me your bios version? I'm eager tu put that savage to work on my old system!

[Tonhovsk]

Ok, man, can you send me your bios version? I'm eager tu put that savage to work on my old system!
The edit button could last for more time.

Well, I wanted to correct the orthographic error, but I suppose there is one grammatical too. So, I'm just eager to see that savage working on my old system.

[Peter M]

It'll probably be a day or two until I go back to that machine. Be patient, and supply your email address meanwhile!

[Peter M]

Even meanwhiler, you might want to try this:

http://www.filesearching.com/cgi-bin/s?t=n&l=en&q=ftp.nix.ru/download/drivers/only_from_www.nix.ru/bios/atrend/atc5130

or this

http://www.rom.by/bios/A-trend/5130/5130_ver_1_0_04.rar

[Tonhovsk]

After a long search, I found a bios version 1.0 04 in the internet.

And after a long time struggling with the installation - there was a message: the program file's part number doesn't match your system -, I got to install it.

To my disappointment, the savage stopped again just before the windows first screen.

Disabling the line "display.drv=pnpdrv.drv" in the system.ini file lets the card run through windows, albeit in 16 colors only. Could it be a driver related problem?

[Peter M]

You mean, like, the link I gave above wasn't good enough?

Drivers here:

http://www.s3graphics.com/en/resources/drivers/legacy/software_archive.jsp

[Tonhovsk]

I had some problems with the links: the first links to a ftp, which is blocked by my LAN firewall; the other I couldn't access, and the cause was (now I know) the browser, Opera. With Internet Explorer I got the file.

Its name: 1_0_04.bin. Mine's 5130v104.bin. Are they different? Is it worth a try the former?

[Tonhovsk]

One more question: a fresh new Windows install, with the updated bios, could change things?

[Peter M]

Compare the files.

Whether a fresh install with the new BIOS in place will change things? It's a long shot, but it might. Not very likely though.

I'm now tinkering with mine again as well. I've just found that to get the ISA ethernet card (an SMC EtherCard Ultra 16) working, one needs to keep "Data Merge" and "Byte Merge" chipset features disabled. And I've seen an inexplicable hang with a Realtek PCI ethernet card - before going back to the ISA card.

[Tonhovsk]

Things are becoming complicated. I'll try all I can do. There are a lot of things to be done. I won't give up.

[Peter M]

Yes, those old boards had tons of BIOS settings that required a ridiculous level of knowledge or trial and error to get right.

I'm a BIOS engineer by profession, and I still had to read the ALi chipset datasheets to get them right ...

[Tonhovsk]

Last night I tried latest Diamond A200 drivers, no result. There is still the S3 and the tweaked drivers, but it seems that the problem is not driver related.

Now I'm looking for the IDE drivers. Peter, are you sure that ATC 5130's chipset is the Aladdin IV+? In the motherboard's manual the chipsets listed are M1543 and M1531. Are they the same?

[Peter M]

1531 and 1543 are the combo that had the marketing name Aladdin IV+.

You download the "unified ULI driver" set from www.nvidia.com, and let it install. It'll say "no drivers to install" but it will have installed the essentials by then.

[Tonhovsk]

But there is no M1543 or M1531 in the chipset supported list of the Integrated220 and Integrated213. Only M1543C.

[Peter M]

Nevermind, just install it anyway. It works (I've tried), and it actually does install a few things that replace Microsoft drivers.

[Tonhovsk]

The doubt I had about ALi aladdin IV+ was because I suddenly remembered that the savage 4 was running, before coming to Brasília, in a Pentium 166 with a TX PRO chipset motherboard.

That is, we can conclude that there isn’t a compatibility problem between Aladdin IV+ and savage4. Am I wrong?

This new information helps anyway?

[Peter M]

TXpro was a badge job on Aladdin IV+ on a handful of PC-Chips and Amptron boards back then. I had one of those (an M560) until recently as well, and found it to be generally more compatible with stuff than the Atrend. It's a question of BIOS.

The one thing the Atrend board does better is, it keeps the PCI frequency at or below 33 MHz, even when the CPU bus is at 75 or 83.

That M560 is now running as an Internet gateway server, with a Cyrix MII-333 and a single PCI card - a multifunction monster pulled from an old graphics workstation (twin U2W SCSI, twin Ethernet, and 3D graphics all on one card).

[Tonhovsk]

Last night I observed one thing that I hadn’t noticed before: when I install the driver, the last dialog box, instead of saying “Windows finished to install the driver for this device”, says “Windows did not install the driver for this device”. Next, it asks to reinitialize system. And after reinitializing, that same lock up.

Any clue?

[Peter M]

Maybe it's now time to look at how the card presents itself. Do you have any raw PCI data, from utilities like PCIScope or somesuch?

[Tonhovsk]

Ok, I can try Pciscope tonight, but it’s a so complicated tool. I hope you help me, as always.

According to Microsoft Knowledge Base, it’s a driver problem ( http://support.microsoft.com/kb/127139/en-us).

Actually, my savage worked with hardware acceleration set to “None”.

In Driverguide site, there are hundreds of drivers. Where should I start in?

[Peter M]

The easiest thing to do with PCIScope is do a "Save As ..." and get the resulting file to me for a look. That way, I can inspect the PCI setup of your machine as if I were sat right in front of it.

[Tonhovsk]

Right, it will be the last chance of that savage. Last night I tried two tweaked drivers of "Driverguide": pure deception, it seems that nothing can make that card work with the ATC 5130.

Tonight I'll try the pciscope; my monday was too busy, I had no time to do it.

[Peter M]

Don't rush it, today is a holiday here in Augsburg (and nowhere else, pfbbbllllllt!), so I won't be able to look at it until tomorrow anyway.

[Tonhovsk]

Peter,

I've already got the pciscope files. How can I send them to you?

[Peter M]

You best upload them to a public place and link it here, so I can pick them up from work.

[Tonhovsk]

I'm sorry, but it's been hard to find such a public place on the net. Don't you know one that I can subscribe for now?

[Tonhovsk]

Ok, here are the links:

Uppit.com/k71b

Uppit.com/hyw4

“savage4” was taken with hardware acceleration set to none.

“savage4safe” with full acceleration hardware, but in safe mode.

Thanks a lot.

[Peter M]

I've had a look - out of scientific interest, I've compared your northbridge settings with the datasheet for the 1531.

Everything looks quite normal (although you should be a bit more daring with those RAM timings ;), except for one thing: You got "Primary Frame Buffer" enabled and set to 2 MBytes. Since the Aladdin IV+ supports up to 16 MBytes and your Savage4 has a 128 MByte resource for the framebuffer, you should be disabling this.

Also, why not enable the board's USB controller? USB slot brackets go for less than $2, and they're really useful :)

[Tonhovsk]

Thanks for the suggestions.

Well, I think we lost our battle. The memory timings won’t make savage work, and disabling VGA frame buffer was alread tried.

Unfortunately, the mistery about ATC 5130/s3 savage4 combination remains.

[Crashman]

Wow, I only read page 1 and 4....but anyway, a bunch of motherboards had problems with the Savage4, and received BIOS updates to fix an incompatibility. I don't have the "proof" for you because I'm simply experienced enough to remember the BIOS update documentation.

So it's possible that your particular board never received a needed BIOS update.

[Peter M]

I didn't explicitly mention that, but the PCI setup looked perfectly OK.

It might still be that the Savage4 chip needs a specific workaround (against a silicon bug) applied - but I never heard or saw of one, despite having been a BIOS engineer for a decade.

Still, we need not dispute that /something/ is wrong, and yes, we do have to admit defeat at this point. Unless we get a PCIScope file from an Aladdin IV+ board that DOES work with that VGA card.

[Crashman]

I just remember the update docs specifically mentioning a resolved issue with the Savage4. Remember that I haven't worked on any of these since they were like 4-years old...so remembering that much is a feat!