No, basically ISA clock is divided by SYSCLK or PCICLK.
Basically the rates can be configured in BIOS.
Use following rates:
For 25 MHz (DX2/50, DX4/75): 1/3
For 33 MHz (DX2/66, DX4/100): 1/4
For 40 MHz (DX2/80, DX4/120): 1/5
For 50 MHz: 1/6
If you don't have a suitable divisor, use 7.13 MHz if available.
ISA clock should be max 8.33 MHz to be not overclocked, or like 8 MHz.

On the 486 the ISA bus is decoupled from the front side bus (if that's what you mean by 'card bus'). Ever since the late-80286 era motherboards have been decoupling the two as CPU clock speeds went ever faster.
The ISA (AT) bus can theoretically run at 12 MHz, even a little higher, but by default it was designed for about 8 MHz. Some expansion cards will cause a failure to boot or otherwise bad scenario if pushed to 10 MHz or higher).
Most motherboard chipsets from this era and onward had BIOS code to configure several "clocks":

CLKIN is the System Clock. This controls the frequency in MHz the CPU is driven at. On 286 and 386 machines, this is half the speed of the main crystal - the CPU divides it by two, hence it was often called CLK2IN.
Incidentally, 486-class motherboards run the CPU at the same frequency as the System Clock - they can do this because they use both edges of the timing signal, which is a square wave.

ATCLK is another clock signal that comes from the clock generator - this one is used by the bus if it's run asynchronously (decoupled from the CPU's speed), otherwise the bus will use the above CLKIN (the System Clock) just like the CPU, except it would use a different divider, e.g. if the CPU were 33 MHz it would use CLKIN/2 for the CPU but use CLKIN/4 for its ISA bus to give it the standard 8 MHz.
On later motherboards that support PCI, this is referred to as PCICLK - the ISA bus speed is then controlled by the PCICLK, e.g. PCICLK/4 if it's a 33 MHz PCI bus.

Happy to be corrected on any of the above, but this is my understanding.

DOSDays

I dis not see anithing like you said

DOSDays wrote on 2023-01-12, 21:17:

The ISA (AT) bus can theoretically run at 12 MHz, even a little higher, but by default it was designed for about 8 MHz. Some expansion cards will cause a failure to boot or otherwise bad scenario if pushed to 10 MHz or higher).

It works fine at 20 Mhz actually. Not with every card though. The older you go - less likely it will work fine.

DOSDays wrote on 2023-01-12, 21:17:

Incidentally, 486-class motherboards run the CPU at the same frequency as the System Clock - they can do this because they use both edges of the timing signal, which is a square wave.

As far as I know, the explanation why 486 CPUs work on 1x CLK (instead of 2x CLK as the 286 and 386) is wrong, although the fact that 486 CPUs work on 1x CLK is correct.

My knowledge is that the 486 CPU contains an on-chip PLL to generate an auxilliary phase-shifte clock to get "extra edges" between the two edges of the clock signal. This is the reason for the "clock stability" specification for the 486 processor: The width of adjacent clock pulses must not differ more than 250ps. In comparison, at 33MHz, a clock pulse has a width of 30.000ps, so the change per clock needs to be less than 1% of the pulse width. Early low-power 486 processors meant for use in battery-powered portable devices, where dynamic frequency changes were required to conserve power thus were still using a 2x CLK input. This kind of 486 processor has never been used in mainstream PCs. Instead, in PC mainboards, power saving has been introduced in different ways: Modern 486 processors with System Management Mode (SMM) also contain a "stop clock" mode. The Intel recommendation is to put the processor into "stop clock" mode, change the frequency, and then exit that mode again. Exiting stop clock mode will re-stabilize the PLL to the new operating frequency. On the other hand, modern 486 mainboards have clock chips that are able to smoothly adjust the output clock so slowly that the 250ps clock-to-clock stability is not violated, allowing arbitrary FSB changes even without putting the processor to sleep from SMM.

AlessandroB wrote on 2023-01-13, 16:15:

I dis not see anithing like you said

You must look in the BIOS. It is a BIOS setting, not a jumper setting.

CoffeeOne wrote on 2023-01-13, 19:38:

AlessandroB wrote on 2023-01-13, 16:15:

I dis not see anithing like you said

You must look in the BIOS. It is a BIOS setting, not a jumper setting.

Some boards do not expose the ISA clock divider in the BIOS setup, but rely on auto configuration. In that case, the BIOS measures the processor frequency during POST, and calculates the FSB frequency from that, and then configures a lot of mainboard timings (RAM speed, cache speed, ISA clock divider) depending on FSB clock. So even if one can not find the ISA speed setting in the BIOS setup (it is supposed to appear in the "Advanced Chipset Configuration"), the divider is still configured by the BIOS code.

The Serpent Rider wrote on 2023-01-13, 16:42:

DOSDays wrote on 2023-01-12, 21:17:

The ISA (AT) bus can theoretically run at 12 MHz, even a little higher, but by default it was designed for about 8 MHz. Some expansion cards will cause a failure to boot or otherwise bad scenario if pushed to 10 MHz or higher).

It works fine at 20 Mhz actually. Not with every card though. The older you go - less likely it will work fine.

I don't disagree, but..
Exceptions prove the rules, sometimes.
There are a few high-speed cards from the days shortly before ISA settled for 8.33 MHz.

For example, I've seen an article about a crazy fast 286 PC in an old German magazine (CHIP?) from the late 80s.

It was a 20 MHz or 25 MHz fast big tower machine, with a fan tunnel made of card board (!) - the article said the staff tested a prototype.
I don't think the beast was running ISA at 8.33 MHz, merely.

I know, this isn't much information. But that's all I remember.
It's been so long. I've read this magazine last time before Windows XP was released.

So some memory boards or SCSI controllers from 1986-1988/89 might be officially capable of operating at 10 or 12 MHz.
The AST Rampage 286 does, at least.
Maybe other models like the QuadRAM or Bocca RAM boards, too.

I used 12Mhz ISA bus on my 386DX40 back in the day. It brought about 50% increase in video performance. On 486 with PCI bus one should stick to 8Mhz ISA bus (as it will be used only by sound card + peripherals). On 486 with VLBus I would still overclock the ISA bus as you will likely end up using an ISA network card. Overclocking ISA on 486 with ISA only is an absolute must.

AlexZ wrote on 2023-01-13, 21:17:

I used 12Mhz ISA bus on my 386DX40 back in the day. It brought about 50% increase in video performance. On 486 with PCI bus one should stick to 8Mhz ISA bus (as it will be used only by sound card + peripherals). On 486 with VLBus I would still overclock the ISA bus as you will likely end up using an ISA network card. Overclocking ISA on 486 with ISA only is an absolute must.

I think the same.

There are occasions were ISA devices are without alternatives, also.

For example, only ISA or VLB/EISA/MCA based VGA cards are compatible with old Super VGA software.
Like Windows 3.1x 800x600 SVGA driver or WfW's 256c SVGA drivers (they can be patched since a few years, but that wasn't an option in the 90s).
Or old graphics drivers made for AutoCAD v2.x, Ventura Publisher etc.
- Okay, in case of professional software there might be VESA VBE drivers (Autodesk software had ADI drivers; there's an ADI VBE driver).
These VBE drivers will work with PCI or AGP/PCIe graphics cards, even.

Another problem is that PCI-based VGA cards nolonger can emulate CGA, EGA or Hercules in silicon.
There's no mode utility, anymore. Software hard-coded for Hercules Monochrome can't be used anymore.
Due to the lack of such an utility, high-resolution text modes, like 130x50, can nolonger be set, either.

Or demoscene productions. The ET-4000w32 series isn't fully compatible to the ET-4000 or ET-4000AX, for example.

In such scenarios, it's worth keeping an ISA SVGA card and operate it on an overclocked ISA bus.