I have that exact same motherboard. I have been busy tinkering too. All I have ATM is a AMD 486 DX4 100 WB running at 120 (3x40) I have been scoping trying to find a 5x86 133 but people on the bay are too proud of them. I have been running with 4x16MB EDO and a Diamond Stealth 3D 2000 (S3 Virge upgraded to 4mb) I have Win98SE installed (for networking) with the unofficial Service Pack. I also am running a 10GB WD HDD... subject to the 8.4GB limitation. I like it, it seems nice, now If I could only increase to 512kb of cache and get that 5x86...

If you happen to have the AMD DX4-100 with 16k wb, you could try to see if your cpu actually has the x4 multi. Theres a chance it actually does 😉

I've never gotten a Am5x86 to power up at 180 MHz. I've tried 4 or so of them. Even running 4v didn't help.

swaaye wrote:

I've never gotten a Am5x86 to power up at 180 MHz. I've tried 4 or so of them. Even running 4v didn't help.

Did you try 5V? I didn't have any success with the cool running ADZ variant, but the one I got to work was a hotter running ADW variant, which seemed to tolerate the 180MHz overclock.

Is there a way to calculate the cacheable limit for the L2 cache on this motherboard? I have 256k with a 32K x 8 TAG and 4 x 32K x 8 in bank 0 and 1.

therelaxer wrote:

Is there a way to calculate the cacheable limit for the L2 cache on this motherboard? I have 256k with a 32K x 8 TAG and 4 x 32K x 8 in bank 0 and 1.

Sorta, but there doesn't seem to be 1 single rule that always works, theres too many factors involved.
All I know is that doubling the L2 cache seems to (usually) double the amount of memory that can be cached.
It also depends on the tag-ram, the chipset itself (Some ali, intel behave differently), write through or write back.

Ideal would be if there was a simple way to determine the amount of cached ram with a bootdisk.
Hmm...I once posted about this, but I can't remember if there was any definitive answer.

Some 386 and 486 board manuals mention the cacheable range.
From what I have read so far, i would say there is a rule:

32kb 8MB
64kb 16MB
128kB 32MB
256kb 64MB
512kb 128MB
1024kB 256MB

These values are for write-through operation. In write-back mode the cacheable range is reduced by half.
I don't exclude additional chipset related restrictions though...

I have found this manual :
http://replay.waybackmachine.org/200202071600 … 3UUD_Manual.htm
Your motherboard supports up to 512Kb cache and 128MB memory. You can use three cache configurations (128Kb, 256Kb, 512Kb) with cachable memory range mentioned above by BastlerMike.

I can confirm BasterMike's words about cacheability. The same goes for Socket 7 - Intel and VIA chips. But s7 ALI is a bit strange about the cacheability amount or maybe they're talking about the same but in a strane manner.

As for the X5 133 overclocking - not all models like 5V. I mean, some don't boot at all at 5V even not overclocked. So you have to either find a 5V tolerant model or solder the board to produce voltage between 4V and 5V (almost none boards have voltage selections in this range).

P.S. can anyone suggest a 486 board with PCI, good memory performance? HOT-433 or there are better ones?

I have just returned to this Biostar mobo, as seen in my original post. I've got it working with the P75 chip, at 200 MHz. (Elsewhere on Vogons, I have got a P75 working at 200 MHz, but that was on a different "SYL8884PCI" mobo.) The amazing thing is, the BIOS settings are completely maxed out on this Biostar mobo, including the Host/PCI ratio set to 1:1!

Oh yes! 3DBench2 gives me 117.6! That's a very good score. That's just using a crappy 2MB 325 Virge card. I ran cachechk(7), and an attached report is included below. I'll mess about with this mobo more tomorrow morning, including getting Quake running, and also Windows 98. That'll require adjusting the BIOS settings, to loosen them up a bit...

ATM, PcpBench gives me 11.7, with the RAM Read setting = 1. If set to 0, it becomes unstable. The other mobo I mentioned earlier, the SYL8884PCI gave me 12.5, so maybe I ought to dig out a better PCI VGA card. Quake timedemo demo1 gives a very impressive 18.3 FPS, in Full screen mode. However, there's some "palette corruption" on screen. That's almost certainly the Virge 325 not liking the 50 MHz PCI bus speed. I'll see if I can find another VGA card...

Speedsys - check out the memory bandwidth!

Edit: CPU/Video gamers performance test v1.4 gives a score of 37.

Edit 2: Doom -timedemo demo3 gives "2134 1077" = 69.350 FPS.

Edit 3: I have Quake working perfectly now. I still get 18.3 FPS, but visually, there is no graphical corruption on screen. The solution was to replace the Virge 325 with a Diamond Viper V330 card. Please note that the excellent and popular Dell Nitro S3 GX card does not work at 50 MHz PCI speed. You won't see a BIOS POST on the screen.

Edit 4: OK people, this is the 3DBench2 score to beat: 118.1. That's using a Diamond Viper V330 card.

Edit 5: I got the board to post at 66 MHz bus speed, with a 3x multi = 198 MHz. The BIOS POST screens says 198 MHz, but it freezes after that. The memory count does not appear on the screen.

Edit 5: I got the board to post at 66 MHz bus speed, with a 3x multi = 198 MHz. The BIOS POST screens says 198 MHz, but it freezes after that. The memory count does not appear on the screen.

This is really stunning result. I tried to boot up cyrix 5x86 100MHz with fsb set to 66MHz (multipler set to 2x) but without positive result. (:

Maybe turning off L2 cache will help? Have you tried this? What type of ram memory are you using (EDO/FPM | 50/60/70ns)?

udam_u, I have managed to get this mobo to work with a P75, 3x multi, 66 MHz bus speed = 198 MHz cpu speed. I ran cachechk7, and it worked. I also ran Speedsys, and it also worked. The Speedsys result wasn't as good as other tests, because I have to set all of the BIOS timings to very slow values. The cachechk7 results were very good, and I'll post the attachment later.

But...there's always a but! After my successful test mentioned in the first paragraph above, I then experimented with different BIOS timings, and different PCI VGA cards, and now I cannot get the mobo to boot up DOS @ 198 MHz any more. I'll keep trying. However, the 3x + 66 MHz option doesn't seem to be as powerful or stable as either the 3x + 60 MHz, or 4x + 50 MHz options.

I am currently using a Number 9 PCI Virge 325 card. Ram used is 1 stick of double-sided 60 ns EDO. I am also experimenting with disabling L2 cache...

udam_u, I notice that you wondered if I was using EDO / 50ns memory. Perhaps I need 50ns memory? I am using 60ns EDO, but if I am trying to run the bus @ 66 MHz, perhaps I need faster RAM? Now, the BIOS POST always freezes when the mobo is about to display the RAM count.

Is 50ns EDO RAM rare? I have a box of EDO RAM somewhere. I might look for it, and then look through all of the sticks, and see if any are marked with -50.

I guess that I must have been lucky earlier on today, with the 3x / 66 MHz = 198 MHz setting. Now, I cannot repeat this test successfully. The BIOS POST always freezes at the point when the memory amount and count is to be displayed.

So, I only have one successfully saved report from that early test. It's cachechk7. The report appears below as an attachment. Looking at it, does it look like the 60ns EDO RAM is being pushed to its limits? If so, I could look for some 50ns EDO RAM...

On second thoughts, 576% seems like a huge value for the L1 reading. Perhaps it's the CPU that can't cope with the 66 MHz bus speed?

Yes, I think that 50ns RAM can help you to obtain stability. 50ns EDO seems to be rare. I have a lot of memory modules but all are signed 60/70ns.

On the other hand this instability might be caused by to high PCI frequency or too slow tag ram. Have you tested 1/2 PCI-FSB divider? I supose that you are still using 15ns tagram?
You can also give CPU a rest, until it cool down to ambient temperature.
I have never used cachechk7 so I don't know what 576% mean (memory speed compared to cache speed?).

Good luck! (:

udam_u wrote:

Edit 5: I got the board to post at 66 MHz bus speed, with a 3x multi = 198 MHz. The BIOS POST screens says 198 MHz, but it freezes after that. The memory count does not appear on the screen.

This is really stunning result. I tried to boot up cyrix 5x86 100MHz with fsb set to 66MHz (multipler set to 2x) but without positive result. (:

Maybe turning off L2 cache will help? Have you tried this? What type of ram memory are you using (EDO/FPM | 50/60/70ns)?

I somehow doubt it's the memory, 60ns should work at 66Mhz without problems anyway.
It's possible that either the CPU or the board are having trouble coping with the very high FSB.
I agree with the 50ns memory being very uncommon. Iirc I only ever found 2 sticks of 50ns EDO 8MB each. But again, I don't think it's the memory.

And about the Cyrix 5x86 100Mhz, these chips were known to be bad overclockers (You should have more luck overclocking the IBM 100's) so it's not surprising you can't overclock a 100Mhz part to 133Mhz. You'll have better luck trying to overclock a 120Mhz part.
There are 133Mhz parts but those are incredibly hard to find!
Those seem to be rarer then the AMD 5x86 150Mhz and 166Mhz parts.

All AMD 5x86 166MHz happen to be fake. And most 150MHz as well (some are real, but there's not to much of those). As for Cyrix - I've heard about 120 but 133 for real? Can you confirm it wasn't a fake?

I think the board can cope with 66 MHz, because I have successfully run it at 66 MHz with an SX-33 chip. With the P75 @ 66 MHz bus speed, I have tried both of the BIOS PCI divider options: 1/2 and 2/3. Also, the tag chip is only 15ns.

I'm really happy with this thing running at 50 / 4x = 200. Perhaps 66 MHz bus speed is more viable for slower chips.

And about the Cyrix 5x86 100Mhz, these chips were known to be bad overclockers (You should have more luck overclocking the IBM 100's) so it's not surprising you can't overclock a 100Mhz part to 133Mhz. You'll have better luck trying to overclock a 120Mhz part.
There are 133Mhz parts but those are incredibly hard to find!
Those seem to be rarer then the AMD 5x86 150Mhz and 166Mhz parts.

You are right. I was able to overclock this CPU up to 120MHz. Peltier doesn't bring any difference. I have never seen 120 and 133MHz version (:

I somehow doubt it's the memory, 60ns should work at 66Mhz without problems anyway.
It's possible that either the CPU or the board are having trouble coping with the very high FSB.
I agree with the 50ns memory being very uncommon. Iirc I only ever found 2 sticks of 50ns EDO 8MB each. But again, I don't think it's the memory.

I think the board can cope with 66 MHz, because I have successfully run it at 66 MHz with an SX-33 chip. With the P75 @ 66 MHz bus speed, I have tried both of the BIOS PCI divider options: 1/2 and 2/3. Also, the tag chip is only 15ns.

I have made some tests. I used 60ns 16MB EDO RAM, SHUTTLE HOT-433 and CPUs listed below:

Intel 486SX-25 -> running at 50MHz (1x50MHz) but both at 60MHz and 66MHz won't boot.

AMD 486DX4-100 -> running at 120MHz (2x60MHz), won't boot at 133MHz (2x66MHz). Also CPU doesn't work with clock set to 150MHz (3x50MHz) so I'm not sure if this CPU is capable to run at 133MHz.

AMD 486DX4-120 -> work fine at 120MHz (2x60MHz), won't boot at 133MHz (2x66MHz). It doesn't work at 150MHz too (3x50MHz) so I'm not sure if this CPU is capable to run at 133MHz as in 486DX4-100 example. ):

I'm going to check Gigabyte and Elitegroup boards based on UMC chipset soon. (:

Regards!

It's an excellent idea to use a variety of different video cards for testing. I use 4. Virge 325, Diamond Viper V330, Dell Nitro, and Trio64v+. They all seem to behave slightly differently, depending on what type of overclocking you are doing. For each speed configuration test, try all 4 video cards.