386 DX-33 detected as 16 Mhz

[terryfi]

Hi

A while ago I acquired this 386/486 motherboard not posting described in this post https://forum.vcfed.org/index.php?threads/weird-short-on-386-486-motherboard.1244159/#post-1330446.

I replaced the system chipset of the board and now it boots with a 486 DX 33 with no issue.
However with a 386 DX 33 it boots successfully but it is slow and shows 16 Mhz in Sysinfo and CHKCPU.

I checked jumper settings, clocks and voltages of cpu; they are all fine.
I am guessing that 386 might be actually a 16 Mhz part and is rebadged fake, any suggestions?

 

[krebizfan]

386s will run at the generated clock speed. If it was a 16 MHz 386, it would run at the 33 MHz the clock sends until destroyed by being so excessively overclocked. I suspect some sort of undocumented interaction between the various jumpers causing it to keep a more realistic bus speed for the 386 and one of the jumpers would have to be switched to get to 33 MHz with a 386.

 

[Eudimorphodon]

I found a sheet of jumper settings for this board, and there's nothing for clock speed, only CPU type. Does this board intend for you to change the oscillator for different speed CPUs? If that's the case, what speed crystal does it have in it now? It's my vague recollection that a difference between the 386 and the 486 is the input clock for the 386 is actually supposed to be twice the speed the CPU runs at. (IE, a "33mhz" 386 needs an input clock of 66mhz), while a 486's input clock is 1:1 with the oscillator clock. Maybe this board uses a 66mhz clock for a 33mhz 486 by dividing it in two, but if it *doesn't* and feeds whatever CPU is installed directly with the output from the installed oscillator then the oscillator that ran your 486 at 33mhz would only run a 386 at 16mhz.

 

[terryfi]

The clock I see on CPU CLK2 pin is 33 Mhz, there is no clock divider directly from chipset to CPU through some buffer/divider. I have to see what is happening internally in 386 cpu; external clock is 33Mhz. Maybe some initialization or fault in CPU. Last resort would be find another similar CPU and test.

 

[Eudimorphodon]

The clock I see on CPU CLK2 pin is 33 Mhz

Look in the datasheet for the 80386 and reread my post. If you’re seeing 33mhz on CLK2 then the chip will run at 16mhz. It’s working exactly how it’s supposed to. (The name of the pin is ‘CLK2’ *because* it’s divided by two internally.)

Somewhere on that board is an oscillator. It’s probably in a rectangular metal can with its frequency engraved on it. What speed is it? If it’s 33mhz you need to change crystals.

 

[terryfi]

Look in the datasheet for the 80386 and reread my post. If you’re seeing 33mhz on CLK2 then the chip will run at 16mhz. It’s working exactly how it’s supposed to. (The name of the pin is ‘CLK2’ *because* it’s divided by two internally.)

Somewhere on that board is an oscillator. It’s probably in a rectangular metal can with its frequency engraved on it. What speed is it? If it’s 33mhz you need to change crystals.

Yes I didn't read all the message. You are right I need a 66 MHz oscillator for 386; the 33 MHz one came with mobo is only good for 486. The weird part is that chipset only regenerates the clock to get rid of skews and noise but does not provide frequency doubling in 386 mode. The chipset datasheet doesn't mention what oscillator frequencies supported but based on similar mobo photos I can see it works with 66 MHz. Thanks for all the help guys, issue is solved.

 

[Timo W.]

The weird part is that chipset only regenerates the clock to get rid of skews and noise but does not provide frequency doubling in 386 mode.

That's not weird. Every 386 mainboard works this way. You need twice the clock frequency of the CPU to have it run with its rated speed.

 

[konc]

Are you sure it's not in turbo-off mode, either from a jumper or the BIOS?

 

[Eudimorphodon]

Yes I didn't read all the message. You are right I need a 66 MHz oscillator for 386; the 33 MHz one came with mobo is only good for 486. The weird part is that chipset only regenerates the clock to get rid of skews and noise but does not provide frequency doubling in 386 mode.

I would have guessed it more likely the board, if it didn't expect you to switch crystals, would use a double-speed crystal for 486s and divide it, since that's technically easier than doubling. The only picture I could find of the board online that had the speed of the crystal readable had the 386 socket populated and a 66mhz crystal, so, yeah, I didn't know if setting the jumpers for 486 would enable a divider or not. But knowing that your board ran the 33mhz 486 at the correct speed with a 33mhz crystal settles that.

Phil's computer lab has some pictures of the board, also populated with a 386, that show a BIOS setup screen that shows the ISA bus speed divider is selected in software. Make sure to check this after you change crystals, because with a 33mhz 486 is was probably set to 1/3 or 1/4. With the faster crystal you'll want to set that to 1/6 or 1/8, or a lot of ISA cards will get unhappy. (Phil's picture showed a 1/6th divider, which is good evidence that their board has a 50 or 66mhz crystal on it. I had a 33mhz Opti chipset 486 back in the day, might be the same chipset as this. ISA is "supposed" to run at 8mhz, but I used to overclock it to 11mhz with the 1/3rd divider; it significantly improved IDE disk performance. Trying to go faster with the 2/5 or 1/2 dividers made things very unhappy.)