What board is your W98 machine using?
Personally I find the sis530 board much more interesting, the lack of AGP isn't much of a hindrance for Super 7 rigs imo.
The LX board might be interesting if it can run very low FSBs, but if I were to pick one, I'd go for the SiS board instantly! 😁

Tetrium wrote:

What board is your W98 machine using?
Personally I find the sis530 board much more interesting, the lack of AGP isn't much of a hindrance for Super 7 rigs imo.
The LX board might be interesting if it can run very low FSBs, but if I were to pick one, I'd go for the SiS board instantly! 😁

The Win98 rig is in a stock Dell Dimension (can't remember model--edit: I think it's a Dimension 4100). But it has an Intel chipset, P3-933, Geforce FX5200 AGP (not stock--I added this), 512MB SDR. I think it originally had Windows ME, but I did a clean install of 98SE. No ISA slots on it.

The LX board has multipliers from x2 to x5 (currently at x5) and some sort of onboard ATI graphics (ATI-264VT4). There's also an onboard Crystal audio (CX4237B-XQ3) and a strange 40 pin header labeled AMC. I did test and it powers on fine. Seems to run flawlessly.

I have not pulled the SiS530 apart yet.

You could also set one up as a Windows 95 machine.

I may test the SiS as a pure DOS system. Since both of these came out of HPs, their bioses are crippled with HP variants. Seems like the options are few and far between. I did have time to test memory in the SiS and for some reason it would only work with 64MB or higher modules. If I put a 32 or smaller in, I'd get a 1-3-3-1 beep code.

edit: I realized the beep code means this board only likes single-sided memory modules. And my only single-sided SDRAM modules are 64MB and bigger. And despite there being a jumper to disable onboard VGA, there is no way (that I found) to disable the board stealing RAM for VGA. The lowest setting in the BIOS is 2MB, so that leaves me with a 62MB system. No biggy, but it is annoying that disabling onboard VGA doesn't turn off RAM stealing.

clueless1 wrote:

I may test the SiS as a pure DOS system. Since both of these came out of HPs, their bioses are crippled with HP variants. Seems like the options are few and far between. I did have time to test memory in the SiS and for some reason it would only work with 64MB or higher modules. If I put a 32 or smaller in, I'd get a 1-3-3-1 beep code.

edit: I realized the beep code means this board only likes single-sided memory modules. And my only single-sided SDRAM modules are 64MB and bigger. And despite there being a jumper to disable onboard VGA, there is no way (that I found) to disable the board stealing RAM for VGA. The lowest setting in the BIOS is 2MB, so that leaves me with a 62MB system. No biggy, but it is annoying that disabling onboard VGA doesn't turn off RAM stealing.

Your K6-2 is probably running at the 100MHz FSB, which most 32MB modules and lower cannot run at (these are usually PC-66 and you'll probably need PC-100 rated memory or better). You could try lowering the FSB to 66 and see if the smaller modules will work that way (just keep the CPU multiplier the way it is, it's only for testing purposes anyway).

I had a SiS530 board which I fried by accident *got a replacement years later though) and I don't remember it having any problem with particular standard SDRAM modules, so I'd say a likely cause is the 100MHz FSB, though my SiS530 was a Compaq or PB I think.

Tetrium wrote:

Your K6-2 is probably running at the 100MHz FSB, which most 32MB modules and lower cannot run at (these are usually PC-66 and you'll probably need PC-100 rated memory or better). You could try lowering the FSB to 66 and see if the smaller modules will work that way (just keep the CPU multiplier the way it is, it's only for testing purposes anyway).

Yes, it is running at 100Mhz FSB, but the 32MB modules I'm using are rated for PC100. Googling around the 1-3-3-1 beep code for Phoenix BIOS led me to believe the modules must have memory chips on only 1 side in this mobo, but i'll go ahead and try lowering the FSB to verify. Thanks!

clueless1 wrote:

Tetrium wrote:

Your K6-2 is probably running at the 100MHz FSB, which most 32MB modules and lower cannot run at (these are usually PC-66 and you'll probably need PC-100 rated memory or better). You could try lowering the FSB to 66 and see if the smaller modules will work that way (just keep the CPU multiplier the way it is, it's only for testing purposes anyway).

Yes, it is running at 100Mhz FSB, but the 32MB modules I'm using are rated for PC100. Googling around the 1-3-3-1 beep code for Phoenix BIOS led me to believe the modules must have memory chips on only 1 side in this mobo, but i'll go ahead and try lowering the FSB to verify. Thanks!

My mistake, I thought you were trying the 32MB SDRAM modules with the 2x8 smaller chips (8 on each side), those are typically not for 100MHz, though some of them are. The newer 32MB ones (often just 4 larger chips in total on the module) typically could do 100MHz without issues, but those were often "singlesided" anyway.

Strange, I really can't remember the SiS530 having the requirement of only working with single sided modules?

You're trying with a single DIMM at a time, right? Or maybe it's just some odd HP BIOS thingy, in that case a reflash to a retail BIOS could fix that.

If I'm not mistaken that board has an updated bios that can run k6-2+ and k6-3+ CPUs. Nice thing about that is you can change multipliers using software in dos and that allows you to throttle the plus CPUs as needed depending on how much power you want.

Back in the day I had really bad luck with the combo: socket 7, SiS and integrated video.
They were always buggy and unstable and using an add-in video card didn't help anything.
.

petro89 wrote:

If I'm not mistaken that board has an updated bios that can run k6-2+ and k6-3+ CPUs. Nice thing about that is you can change multipliers using software in dos and that allows you to throttle the plus CPUs as needed depending on how much power you want.

Yeah, I think that is true because when I found the latest bios on the Asus site, it was some versions lower and the highest supported cpu they listed was less than the K6-2 550 (it was a 450, I believe).

How do you change multipliers in dos? That sounds vaguely familiar, I wonder if Phil covered it in one of his videos.

PCBONEZ wrote:

Back in the day I had really bad luck with the combo: socket 7, SiS and integrated video.
They were always buggy and unstable and using an add-in video card didn't help anything.
.

So far the onboard SiS graphics is very fast. I compared it to a modernish GF 8400GS PCI and it is much faster than the GF:

3DBench:
8400GS PCI--127.9 fps
integrated SiS--255.8 fps

PCPBench:
8400GS PCI--85.9
integrated SiS--96.7

Doom:
8400GS PCI--42.5
integrated SiS--109.5

Quake:
8400GS PCI--57.2
integrated SiS--61.6

K6+ chips can change the multiplier in DOS (AFAIK through SETMUL). It doesn't work on regular K6-3.

alexanrs wrote:

K6+ chips can change the multiplier in DOS (AFAIK through SETMUL). It doesn't work on regular K6-3.

SETMUL says it cannot change the mult, only the L1 cache. With L1 disabled it's still too fast for Wing Commander. for fun, I underclocked via jumpers as low as possible (2.5x66=166), then disabled L1 and it was *still* too fast for WC.

On the upside, Descent2 and System Shock are smooth as silk.

Tetrium wrote:

My mistake, I thought you were trying the 32MB SDRAM modules with the 2x8 smaller chips (8 on each side), those are typically not for 100MHz, though some of them are. The newer 32MB ones (often just 4 larger chips in total on the module) typically could do 100MHz without issues, but those were often "singlesided" anyway.

Strange, I really can't remember the SiS530 having the requirement of only working with single sided modules?

You're trying with a single DIMM at a time, right? Or maybe it's just some odd HP BIOS thingy, in that case a reflash to a retail BIOS could fix that.

Well, I retried all my 32MB modules with the bus set to 66 and 3 worked while 4 did not. The 3 that worked at 66 would not work at 100. 2 of the 4 not-working sticks had PC100 stickers on them, but they do have 8 chips on each side. Maybe someone back in the day was trying to sell 66Mhz chips at 100Mhz prices.

I'm not sure yet if I want to reflash to Asus bios. I haven't found an HP bios to flash back if I decide I want K6-2 550 support back. According to Asus, their bios only supports up to K6-2 450. And if I can't get this thing to slow down enough for WC, then there's little point to risking it. It may just have to be a high-end 550Mhz DOS machine with 64MB, or a 366Mhz DOS machine with 32MB.

You can save a copy of the BIOS.

clueless1 wrote:

Tetrium wrote:

My mistake, I thought you were trying the 32MB SDRAM modules with the 2x8 smaller chips (8 on each side), those are typically not for 100MHz, though some of them are. The newer 32MB ones (often just 4 larger chips in total on the module) typically could do 100MHz without issues, but those were often "singlesided" anyway.

Strange, I really can't remember the SiS530 having the requirement of only working with single sided modules?

You're trying with a single DIMM at a time, right? Or maybe it's just some odd HP BIOS thingy, in that case a reflash to a retail BIOS could fix that.

Well, I retried all my 32MB modules with the bus set to 66 and 3 worked while 4 did not. The 3 that worked at 66 would not work at 100. 2 of the 4 not-working sticks had PC100 stickers on them, but they do have 8 chips on each side. Maybe someone back in the day was trying to sell 66Mhz chips at 100Mhz prices.

I'm not sure yet if I want to reflash to Asus bios. I haven't found an HP bios to flash back if I decide I want K6-2 550 support back. According to Asus, their bios only supports up to K6-2 450. And if I can't get this thing to slow down enough for WC, then there's little point to risking it. It may just have to be a high-end 550Mhz DOS machine with 64MB, or a 366Mhz DOS machine with 32MB.

At least you got a few working ones now 😀
You could try to clean the contacts of the other 4 modules with rubbing alcohol, this trick has revived seemingly broken memory modules on more than 1 occasion.

I never really liked flashing BIOS's, so I only done it a few times so I don't know much about it. There might be an unofficial modded BIOS file you could flash, these often add new features like improved CPU support and support for larger harddrives.

The other contact cleaning technique is to use an eraser.

clueless1 wrote:

SETMUL says it cannot change the mult, only the L1 cache. With L1 disabled it's still too fast for Wing Commander. for fun, I underclocked via jumpers as low as possible (2.5x66=166), then disabled L1 and it was *still* too fast for WC.

Non + chips cannot change their multipliers in software. AMD added that capability to K6-2+ and K6-3+ because they were mobile chips, and lowering the multiplier on demand (eg. the BIOS slowing the processor down when using batteries) is a desireable feature. For other AMD and Intel chips - its just impossible. The chips read the multipliers on cold boot or when the RESET line is asserted alonger than a certain amount of time (which the one motherboard I connected the Turbo button to the multiplier pins does not - I have to turn it off and on again to get the new multiplier to work).

Hooking the multiplier pins to the turbo button is a possible workaround. Reducing the FSB (from 100 to 66MHz with a 5,5 multi would give you an underclock from 550MHz down to 363MHz) using the turbo button is another approach - and this one can usually be used while the system is on.

Hi alexanrs,

I'm not aware that there is a Turbo button on these motherboards. Would that reduce the speed even more than I manually (by jumpers) did? I dropped the 550Mhz cpu to 166Mhz by reducing the FSB to 66 and multipler to x2.5, then disabled L1 cache with SETMUL after that and it was still too fast (about 38 in speedsys).