The Serpent Rider wrote on 2023-03-28, 18:02:

Only specific chipset revision has EDO support.

Which revisions support EDO? I have a 1.8 that is tested working and a 1.22 I have yet to test. It'll be a few weeks before I can get to them to check the chipset revisions

85C496 with "PR" code.

Thanks

...and my mobo has OR & OT so it means no EDO? PR is later?

dj_pirtu wrote on 2023-03-29, 06:45:

...and my mobo has OR & OT so it means no EDO? PR is later?

Yes, indeed. "P" is after "O" in the alphabeht, and SiS used those code in alphabetical order.

dj_pirtu wrote on 2023-03-21, 05:52:

Figured out yesterday that VT8228 clock chip can do 66MHz FSB so tried and it worked. […]
Show full quote

Figured out yesterday that VT8228 clock chip can do 66MHz FSB so tried and it worked.

CPU is late i486DX4/100 WB Enhanced and takes 2x66MHz=133MHz with no problems. (Tried 3x50MHz=150MHz on that QDI V4S471 but didn't make it to DOS)
Cache needs to be slowed down a little and CPU to PCI Post Write needs to be disabled or strange things will happen. If enabled and when in heavy CPU load, screen stops updating and DOS4GW gives some kind of real mode -error. When disabled it just kills PCI-speed (display card, Cirrus 5446 PCI) from 40MB/s -> 20MB/s.

My other system QDI V4S471 with CX5x86 (2x50MHz) is just as fast in memory operations, it's cache timings are fastest and memory 'faster'.

But still, that i486DX4/133 is quite fast.

Anybody else have tried 66MHz clock on this mobo?

Hello,

OK, so 66MHZ is (JP 25 to JP 28): 2-3, 2-3, 2-3, 1-2

Do you also have the 60MHz setting in your notes?

CoffeeOne wrote on 2023-04-30, 17:31:

Hello, […]
Show full quote

dj_pirtu wrote on 2023-03-21, 05:52:

Figured out yesterday that VT8228 clock chip can do 66MHz FSB so tried and it worked. […]
Show full quote

Figured out yesterday that VT8228 clock chip can do 66MHz FSB so tried and it worked.

CPU is late i486DX4/100 WB Enhanced and takes 2x66MHz=133MHz with no problems. (Tried 3x50MHz=150MHz on that QDI V4S471 but didn't make it to DOS)
Cache needs to be slowed down a little and CPU to PCI Post Write needs to be disabled or strange things will happen. If enabled and when in heavy CPU load, screen stops updating and DOS4GW gives some kind of real mode -error. When disabled it just kills PCI-speed (display card, Cirrus 5446 PCI) from 40MB/s -> 20MB/s.

My other system QDI V4S471 with CX5x86 (2x50MHz) is just as fast in memory operations, it's cache timings are fastest and memory 'faster'.

But still, that i486DX4/133 is quite fast.

Anybody else have tried 66MHz clock on this mobo?

Hello,

OK, so 66MHZ is (JP 25 to JP 28): 2-3, 2-3, 2-3, 1-2

Do you also have the 60MHz setting in your notes?

I couldn't find a 66 MHz option on this board, so I'd also like an answer to the above.

EDIT: JP25 yields odd results. Best to leave on 1-2.

MHz, JP26, JP27, JP28
25, 2-3, 1-2, 2-3
33, 2-3, 2-3, 1-2
40, 1-2, 1-2, 2-3
16, 2-3, 1-2, 1-2
16, 2-3, 2-3, 2-3
?, 1-2, 1-2, 1-2
50, 1-2, 2-3, 1-2
20, 1-2, 2-3, 2-3

From the above table, possibly '?' is 66 MHz and my Am5x86 just won't turn on at 3x66 at 3.6 V. For me, I really am looking for 60 MHz, which doesn't exist. I started anothre thread looking for a replacement clock generator, Replacement clock generator for Asus PVI-486SP3 to enable 60/66 MHz FSB?

feipoa wrote on 2023-09-23, 21:42:

I couldn't find a 66 MHz option on this board, so I'd also like an answer to the above. […]
Show full quote

CoffeeOne wrote on 2023-04-30, 17:31:

Hello, […]
Show full quote

dj_pirtu wrote on 2023-03-21, 05:52:

Figured out yesterday that VT8228 clock chip can do 66MHz FSB so tried and it worked. […]
Show full quote

Figured out yesterday that VT8228 clock chip can do 66MHz FSB so tried and it worked.

CPU is late i486DX4/100 WB Enhanced and takes 2x66MHz=133MHz with no problems. (Tried 3x50MHz=150MHz on that QDI V4S471 but didn't make it to DOS)
Cache needs to be slowed down a little and CPU to PCI Post Write needs to be disabled or strange things will happen. If enabled and when in heavy CPU load, screen stops updating and DOS4GW gives some kind of real mode -error. When disabled it just kills PCI-speed (display card, Cirrus 5446 PCI) from 40MB/s -> 20MB/s.

My other system QDI V4S471 with CX5x86 (2x50MHz) is just as fast in memory operations, it's cache timings are fastest and memory 'faster'.

But still, that i486DX4/133 is quite fast.

Anybody else have tried 66MHz clock on this mobo?

Hello,

OK, so 66MHZ is (JP 25 to JP 28): 2-3, 2-3, 2-3, 1-2

Do you also have the 60MHz setting in your notes?

I couldn't find a 66 MHz option on this board, so I'd also like an answer to the above.

JP25 only sets the PCI divisor, such that:
1-2 ==> PCI=FSB
2-3 ==> PCI = 1/2*FSB

MHz, JP26, JP27, JP28
25, 2-3, 1-2, 2-3
33, 2-3, 2-3, 1-2
40, 1-2, 1-2, 2-3
16, 2-3, 1-2, 1-2
16, 2-3, 2-3, 2-3
?, 1-2, 1-2, 1-2
50, 1-2, 2-3, 1-2
20, 1-2, 2-3, 2-3

From the above table, possibly '?' is 66 MHz and my Am5x86 just won't turn on at 3x66 at 3.6 V. For me, I really am looking for 60 MHz, which doesn't exist. I started anothre thread looking for a replacement clock generator, Replacement clock generator for Asus PVI-486SP3 to enable 60/66 MHz FSB?

Are you sure?
This is a picture from my board

There is a setting for 50MHz printed: 2-3, 2-3, 1-2, 2-3
So JP 26, 27, 28 is 2-3, 1-2, 2-3. It is the same as for 25MHz. According to your theory, that would be 25MHz with a PCI clk of 12.5MHz. No, you are wrong, that is a setting for 50MHz.
I am pretty sure that (JP 25 to JP 28): 2-3, 2-3, 2-3, 1-2 IS FSB 66MHz.

CoffeeOne wrote on 2023-09-23, 22:46:

Are you sure? This is a picture from my board PXL_20230923_223224394.jpg There is a setting for 50MHz printed: 2-3, 2-3, 1-2, 2- […]
Show full quote

feipoa wrote on 2023-09-23, 21:42:

I couldn't find a 66 MHz option on this board, so I'd also like an answer to the above. […]
Show full quote

CoffeeOne wrote on 2023-04-30, 17:31:

Hello,

OK, so 66MHZ is (JP 25 to JP 28): 2-3, 2-3, 2-3, 1-2

Do you also have the 60MHz setting in your notes?

I couldn't find a 66 MHz option on this board, so I'd also like an answer to the above.

JP25 only sets the PCI divisor, such that:
1-2 ==> PCI=FSB
2-3 ==> PCI = 1/2*FSB

MHz, JP26, JP27, JP28
25, 2-3, 1-2, 2-3
33, 2-3, 2-3, 1-2
40, 1-2, 1-2, 2-3
16, 2-3, 1-2, 1-2
16, 2-3, 2-3, 2-3
?, 1-2, 1-2, 1-2
50, 1-2, 2-3, 1-2
20, 1-2, 2-3, 2-3

From the above table, possibly '?' is 66 MHz and my Am5x86 just won't turn on at 3x66 at 3.6 V. For me, I really am looking for 60 MHz, which doesn't exist. I started anothre thread looking for a replacement clock generator, Replacement clock generator for Asus PVI-486SP3 to enable 60/66 MHz FSB?

Are you sure?
This is a picture from my board PXL_20230923_223224394.jpg
There is a setting for 50MHz printed: 2-3, 2-3, 1-2, 2-3
So JP 26, 27, 28 is 2-3, 1-2, 2-3. It is the same as for 25MHz. According to your theory, that would be 25MHz with a PCI clk of 12.5MHz. No, you are wrong, that is a setting for 50MHz.
I am pretty sure that (JP 25 to JP 28): 2-3, 2-3, 2-3, 1-2 IS FSB 66MHz.

Here are more settings: I think you were also particpating in this thread:
Re: Test:486 undocumented jumper settings

I could be wrong about JP25 being the PCI divisor. It could be the S3 input. I will need to run some tests with the scope to be certain.

Here is the result from my 50 MHz setting, which 5u3 calls a 48 MHz setting. chkcpu also calls it 50 MHz.

I think jp25 changes the cpu clock from CLK to CLK*2, so for 50Mhz jp25 would be changed while 26-28 remain the same.... note how the jumpers on the sticker are same for 25Mhz as 50Mhz except jp25.
The datasheet for AV9155-02 shows a CLK out pin and a CLK*2 out pin (called CPU and 2x CPU)...and no 30mhz according to the table.....it also has a BUSCLK output of 32Mhz which probably goes to PCI and a /4 for ISA...
assuming the VT8228 is similar (since the jumpers are hardwired) in config....and the VT8225 also has a 1x and 2x clock set by S3.....just observations... 😀

My board is revision 1.8 and it doesn't have any silkscreen or stickers indicating how 50 MHz jumpers would set. I have run all possible jumper combinations for JP25, 26, 27, 28. I measured CLK directly at the CPU's CLK pin, I measured PCI CLK from the PCI slot. The results are attached.

The entries I highlighted in yellow are the ones which I believe are the correct and operational. All others are, non-functional, except for a second 25 MHz setting with a bumpy waveform.

The theory that the VIA PLL follows the convention of the AV9155 PLL seems sound, however the measured results don't follow this CLK*2 convention. The exception is for the 50 MHz setting, which on my board is:

50 MHz FSB
JP25, JP26, JP27, JP28
1-2, 1-2, 2-3, 1-2

If we adjust JP25 to 2-3, we'd expect the 50 MHz setting to double, which it does, but only for the 50 MHz setting.

Thus,
JP25, JP26, JP27, JP28
2-3, 1-2, 2-3, 1-2
has the following waveform, but is NO POST:
Yellow is CPUCLK, blue is PCI CLK.

From this analogy, we might expect the 33 MHz setting, with JP25 set to 2-3 to be the 66 MHz setting, but alas, it is not. It is still 33 MHz and no POST. For the most part, when JP25 is set to 2-3, the PCI CLK and CPU CLK are 180 degrees out of phase. All but one setting with JP25 set to 2-3 would not POST (the other being the second 25 Mhz FSB setting with bumpy waveform).

This is what the proposed 66 MHz setting looks like. FSB CLK is yellow. PCI CLK is blue.

It is still 33 MHz on my system, with asymmetric widths for the peaks/valleys. Does this work at 66 Mhz on your system? Could you provide a scope image?

I have confirmed that JP25 pin 3 is the CLK*2 pin on the PLL (both VT8228N and AV9155). It is unclear why I am not witnessing 2x clocks on the scope or when running the motherboard.

wow yeah something is screwy !! Can you recheck those I boxed in red ? Most of those with jp25 @ 2-3 are FSB to PCI 2:1 but not those....
added: in your chart the 100Mhz is close 2 to 1, makes no sense that the 25 and 33 are not....if the timer can do 100Mhz makes no sense it can not do 50 or 66...

Refer to above oscilloscope image titled, Bogus_66Mhz_FSB_2.JPG. It is really at 33 MHz. Similarly, I looked at the scope image for the other two configurations you've outlined in orange.

feipoa wrote on 2023-09-24, 03:03:

I have confirmed that JP25 pin 3 is the CLK*2 pin on the PLL (both VT8228N and AV9155). It is unclear why I am not witnessing 2x clocks on the scope or when running the motherboard.

I've recently purchased an Asus PVI-486SP3. Motherboard revision is 1.22, chipset is SiS85C496/97 version "OT/OR" and with a Via VT8228 PLL. BIOS is mid 1996, Award.

So, this board is very similar to the one of the OP. I can confirm that my it can output a bus clock of 66MHz. JP25 multiplies the frequency selected with JP26-29 by a factor of two.

This setting, however, is quite unstable. So far, the only CPU I could successfully boot at 66x2 is an IBM manufactured Cyrix Cx586 raising the voltage from 3.45 to 3.6V (CPU will do 120MHz just fine at 3.45V). However, I couldn't complete any benchmark; screen has some random corruption, the system will lock when launching games like Quake or Doom or even some synthetic benchmarks. Only some basic software like CHKCPU or CACHECHK seems to work. I had to disable the L2 cache and use conservative timings for the RAM. I'm using 60ns FPM Kingston modules dated late '97 that flawlessly work in Pentium motherboards even with the tightest timings, so I tend to rule out its a RAM problem.

It's probably a CPU issue, coupled with the fact that the 486 bus/boards weren't engineered to work at such high frequency. The SP3 is however a fine board; it has a single VLB slot and I was surprised that it can take my best VLB card, an ET4000W32 ver B, at 50MHz with 0 wait states. At 66Mhz, however, it won't boot not even setting wait states set to 1. I had to use a PCI card instead (an ET6000).

I didn't test the output signals with a scope, but CACHECHK and CHKCPU report a 133MHz CPU and CACHECHK timings seems to validate that.
I've also tried with a Amd 5x86 @ 66x3 and with an Intel DX4 @ 66x2 but I can't get past the POST screen, not even with a slight overvolt (3.6V).

From this thread, it seems the VT8228PLL is capable of generating some other undocumented frequencies; a shame 60MHz is among them, though... or am I wrong?