486 Overclocking: Where does the clock come from?

[Pickelhaube808]

This feels like a weird question to ask, but when using the multiplier jumpers on a board to switch between ex. 25, 33, 40, and 50MHz, what is actually generating this base clock? Do the jumpers control an external clock (such as the 14.31818MHz crystal I see nearby on the board) or does the chipset somehow tell the CPU to generate its own clock at that rate?

Say... theoretically, if nothing else was holding it back, could I trick the board into telling the CPU to run at OVER 50MHz for the base clock?

Asking for a friend who has a good chip and some dry ice laying around...

 

[krebizfan]

There should be a single oscillator that all the speeds cascade off of. The various jumpers would set up multipliers of the base clock to give the final clock. Replacing the oscillator with a faster one would run everything faster. Note that the ISA bus could also get pushed to higher clock speed so some ISA cards will not work. There may be other parts of the system that won't be happy at greatly increased clocks.

486 chips that could be stable with a clock at more than 50 MHz were quite rate. I hope you got lucky with your selection.

 

[Makefile]

Some early 486 boards have a can oscillator that has to be changed out in order to run the CPU at a different bus speed.
Later ones use a clock generator chip off of the 14.31818.
There has been past experimentation e.g. of running a 486 chip at 60x2 or 60x3, where possible, to experiment with faster memory timings. Obviously that would be taking advantage of undocumented settings, e.g. where that clock generator chips has more settings than the motherboard manual mentions.

 

[Chuck(G)]

My recollection is that the 50 MHz 5v 486s ran hot as a pistol.

 

[GiGaBiTe]

The DX/50 was also very unstable, mostly owing to motherboards and peripherals not tolerating the 50 MHz FSB. The only motherboards that *mostly* worked fine with the DX/50 were the ISA only boards, because the ISA clock wasn't usually that much higher than normal. The trashy Taiwanese clone boards were pretty much a non-starter. They often had issues at even 33 and 40 MHz due to wafer thin PCBs that were badly laid out.

50 MHz was out of the VLB spec, most VLB cards would become unstable at that clock speed, and it got worse if multiple cards were on the bus. PCI was sometimes also tied to the FSB clock, and few cards at the time tolerated such a high clock speed. It wasn't until late in the 90s when the 66 MHz and higher specs came out, and those often weren't backwards compatible with the original PCI spec anyway.

 

[Pickelhaube808]

Some early 486 boards have a can oscillator that has to be changed out in order to run the CPU at a different bus speed.
Later ones use a clock generator chip off of the 14.31818.
There has been past experimentation e.g. of running a 486 chip at 60x2 or 60x3, where possible, to experiment with faster memory timings. Obviously that would be taking advantage of undocumented settings, e.g. where that clock generator chips has more settings than the motherboard manual mentions.

What is significant about 14.31818...? It's certainly not a clean 2^n like 16.384MHz, and looks more like an irrational number to me. Is there another component that takes this clock and coverts it to 25/33/40/50/66 before sending it to the CPU to be multiplied again (in DX* chips)?

 

[GiGaBiTe]

What is significant about 14.31818...?

It's related to the NTSC colorburst frequency, specifically four times that (about ~3.58 MHz.)

Computers in the 70s and 80s used the NTSC colorburst frequency as a reference and cascaded off it via division or multiplication.

14.31818 / 4 = ~3.58 MHz (NTSC Colorburst). 14.31818 / 12 = ~1.193182 MHz (8253 Program Interval Timer.) 1.193182 MHz x 4 = ~4.77 MHz of the 8088 in the original IBM PC.

As for why it was used, either because the crystals were cheap from being widely used. Or because it made computers easier to interface to televisions, since early computers often used TVs as a video output.

16.384 kHz is used for the RTC.

 

[modem7]

What is significant about 14.31818...?

Put onto the ISA slots. See the diagram at [here].

 

[krebizfan]

Edited since the NTSC origin and cheap nature has been noted.

https://electronics.stackexchange.c...requencies-33-33-mhz-48-000-mhz-from-14-318-m seems like a simplified introduction to how to multiply the frequency to what is needed. Many chipsets were programmed for how to do it.

 

[Chuck(G)]

PLLs are wonderful things. For example, the STM32 MCUs all use an 8MHz crystal (or XCO) for their clock; the ST32H7 MCUs take that 8 MHz and use it to run a PLL to 480MHz internal clock--and, at the same time, uses that same 8MHz to provide USB, HDMI video, ethernet, etc.