SodaSuccubus wrote on 2020-07-12, 07:02:

Pressing the turbo button crashes the machine. I thought maybe it was some kind of sleep mode, but unpressing the button does not restore the PC. My turbo button is a 3 pin, and the header on my DataExpert exp4044 is a 2 pin. Iv tried different position with the pins to no avail.

I'm using a DX4 Overdrive. I have it jumpers for a DX33 (as I think you where suppose to do for ODPRs) although the board does have its own jumpers for a DX4. If that matters.

Jumpering the DX4 Overdrive as DX33 is correct. The way the turbo option works on the EXP4044 I have access to is by dropping the front-side bus from whatever you configured the board to to 8MHz. The clock generator chip on the EXP4044 is supposed to drop the clock smoothly enough that the DX4 is able to keep its internal oscillator at the tripled frequency is sync. If that oscillator gets out of sync (even if just for a short moment), erratic operation or crashes are expected. The EXP4044 expects a toggle switch on the turbo jumper connection, not a momentary push button. You can try booting the board with the turbo jumper pins open and with the turbo jumper pins shorted using a jumper. One configuration is supposed to work at full 100MHz, the other one at 24MHz (3*8MHz). If both configurations work, the problem in your system is the transition between the two FSB clocks.

In case your computer can not reliably switch the FSB clock while operating, a workaround is to use a cache enable/disable tool instead of the turbo button to control system speed.

I had similar issue with a lucky star motherboard when I overclocked amd 100Mhz to 133Mhz with 4X multiplier using multiplier jumper. When configured to 3X multiplier the turbo button slows down the cpu fine but not with 4X because then turbo button just freezes the PC. The cpu is perfectly stable at 133Mhz as long as I don't touch the turbo button.

mkarcher wrote on 2020-07-12, 09:46:

Jumpering the DX4 Overdrive as DX33 is correct. The way the turbo option works on the EXP4044 I have access to is by dropping the front-side bus from whatever you configured the board to to 8MHz. The clock generator chip on the EXP4044 is supposed to drop the clock smoothly enough that the DX4 is able to keep its internal oscillator at the tripled frequency is sync. If that oscillator gets out of sync (even if just for a short moment), erratic operation or crashes are expected. The EXP4044 expects a toggle switch on the turbo jumper connection, not a momentary push button. You can try booting the board with the turbo jumper pins open and with the turbo jumper pins shorted using a jumper. One configuration is supposed to work at full 100MHz, the other one at 24MHz (3*8MHz). If both configurations work, the problem in your system is the transition between the two FSB clocks.

In case your computer can not reliably switch the FSB clock while operating, a workaround is to use a cache enable/disable tool instead of the turbo button to control system speed.

Iv tried booting with just a jumper. It crashes.
Now that I think about it. I think last time it worked, it was with a DX2 and an ISA video card.

Do you suspect all my VLB cards have anything to do with the FSB being unhappy?

I hope I haven't been hurting my CPU and board with all the crashing! 🙁. 😭

SodaSuccubus wrote on 2020-07-12, 14:58:

Iv tried booting with just a jumper. It crashes. Now that I think about it. I think last time it worked, it was with a DX2 and a […]
Show full quote

mkarcher wrote on 2020-07-12, 09:46:

The way the turbo option works on the EXP4044 I have access to is by dropping the front-side bus from whatever you configured the board to to 8MHz.

You can try booting the board with the turbo jumper pins open and with the turbo jumper pins shorted using a jumper. One configuration is supposed to work at full 100MHz, the other one at 24MHz

Iv tried booting with just a jumper. It crashes.
Now that I think about it. I think last time it worked, it was with a DX2 and an ISA video card.

Do you suspect all my VLB cards have anything to do with the FSB being unhappy?

I hope I haven't been hurting my CPU and board with all the crashing! 🙁. 😭

While I never heard of any examples, I can very well imagine a graphics card design that locks up the whole FSB if the FSB clock is too low, so it might be that your VL graphics cards is unhappy at 8MHz VL clock. You should try with a different graphics card to find out. If it turns out that the graphics card is indeed the culprit, I am interested in knowing what card it is.

I do know one example of a VL card disliking 8MHz VL clock: The Tekram DC-680T VL caching IDE controller fails its internal POST at VL 8MHz. It works fine if the VL clock is dropped after the controller firmware has booted, though.

mkarcher wrote on 2020-07-12, 16:15:

SodaSuccubus wrote on 2020-07-12, 14:58:

Iv tried booting with just a jumper. It crashes. Now that I think about it. I think last time it worked, it was with a DX2 and a […]
Show full quote

mkarcher wrote on 2020-07-12, 09:46:

The way the turbo option works on the EXP4044 I have access to is by dropping the front-side bus from whatever you configured the board to to 8MHz.

You can try booting the board with the turbo jumper pins open and with the turbo jumper pins shorted using a jumper. One configuration is supposed to work at full 100MHz, the other one at 24MHz

Iv tried booting with just a jumper. It crashes.
Now that I think about it. I think last time it worked, it was with a DX2 and an ISA video card.

Do you suspect all my VLB cards have anything to do with the FSB being unhappy?

I hope I haven't been hurting my CPU and board with all the crashing! 🙁. 😭

While I never heard of any examples, I can very well imagine a graphics card design that locks up the whole FSB if the FSB clock is too low, so it might be that your VL graphics cards is unhappy at 8MHz VL clock. You should try with a different graphics card to find out. If it turns out that the graphics card is indeed the culprit, I am interested in knowing what card it is.

I do know one example of a VL card disliking 8MHz VL clock: The Tekram DC-680T VL caching IDE controller fails its internal POST at VL 8MHz. It works fine if the VL clock is dropped after the controller firmware has booted, though.

I just tried again, this time with an ISA video and IDE. Still crashed. So it's not VLB causing the issues

Very strange. Anyone else got any ideas?

I thought the 486 turbo just increased waitstates not manipulated the fsb

candle_86 wrote on 2020-07-12, 20:57:

I thought the 486 turbo just increased waitstates not manipulated the fsb

That was my impression too. That or they disable cache.

It wouldn't make sense to change clock speed on the fly, the 486 doesn't support it, that's in the documentation.

candle_86 wrote on 2020-07-13, 13:51:

It wouldn't make sense to change clock speed on the fly, the 486 doesn't support it, that's in the documentation.

I did measurements (including scope measurements) on a copy of the EXP4044 board. This board does change FSB on the fly. You can do that with (most?) 486 class processors, if you do it smoothly enough. The smooth change of FSB frequency is acutally an advertised feature of later 486-class clock gen chips. On the EXP4044, the clock generator chip is a IMI SC462, datasheet here: https://www.datasheetarchive.com/pdf/download … am%2520circuits

SC462_datasheet wrote:

DOZE# DOZE control pin. When DOZE# is high, the clock chip operates in standard mode. When this pin goes low, output frequencies are switched to the frequencies programmed by ST0-ST1 pin. Switching to 16/8/4 MHz, DOZE frequencies occur smoothly to allow tracking by the 486 CPU internal PLL. Switching to 2/1/0 MHz, DOZE frequencies occur quickly, without glitches, to support static CPU's and SMI operating modes. See timing diagems, page 10, for description. This pin has an internal pull-up.

The important point is in the middle of that paragraph: "Switching to [...] DOZE frequencies occur smoothly to allow tracking by the 486 CPU internal PLL". Early 486 documentation did not think of a sophisticated clock generation like this and might quote that the clock must not change during operation, but in practice, this is how the DOZE mode is implemented on most "green" motherboards. I agree that it is uncommon to use the DOZE mode as deturbo mode, as the EXP4044 is the only 486 mainboard I know of pulling this trick.

SodaSuccubus wrote on 2020-07-12, 22:10:

candle_86 wrote on 2020-07-12, 20:57:

I thought the 486 turbo just increased waitstates not manipulated the fsb

That was my impression too. That or they disable cache.

Mine just disables cache

I thought the 486 turbo just increased waitstates not manipulated the fsb

Quite a lot of boards just halved FSB (or further downclocked FSB) on turbo button. For example, Lucky Star LS486-E did that, so you're technically can achieve 12 Mhz 486DX on that board.