DX2-66 is a clock doubling chip. So it uses 33 MHz bus and 66 MHz only internally. The BIOS might show 33 MHz at boot, but benchmarks should tell the difference. No special setup should be needed. Which benchmark did you run?

A DX2 CPU has clock-doubling circuit inside so it actually accepts 33MHz clock - and the bus is running at that speed as well. It's only the CPU internals that run at 66MHz, but since it does have on-chip cache it can run the code faster once it's fetched.
Can you remove the heatsink and provide a photo of your CPU top? Maybe it's not actually a DX2 after all?

That AMI Super Voyager is a VERY early 486 board. Makes me wonder if everything on it can actually support 33MHz...

According to TH99, J17 needs to be set to 1-2 for speeds over 20MHz. On the photo it looks like it's already set to that, is that the case?

Otherwise - do you have a 40MHz or 50MHz oscillator you could try? If neither works, it's probably something about the can. If at least one does, it's the speed of the 66MHz version something doesn't like.

Edit: the oscillator normally runs at *twice* the system bus speed. So for 33MHz bus, you need a 66MHz oscillator. OP is correctly trying a 66MHz oscillator for a 33MHz bus for a DX2 CPU that doubles that back up to 66MHz.

mpe wrote on 2020-01-28, 22:49:

DX2-66 is a clock doubling chip. So it uses 33 MHz bus and 66 MHz only internally. The BIOS might show 33 MHz at boot, but benchmarks should tell the difference. No special setup should be needed. Which benchmark did you run?

I tried winspeed wich put it at 33mhz or lower, i also tried SYSTEM SPEED TEST Version 4.78 wich put the cpu speed somewhere similar (below a DX2-50 )

if i run CHKCPU it does report cpu internal speed as 66mhz and bus speed at 33mhz.

Maybe the motherboard is just a crappy one? 😀

Deunan wrote on 2020-01-28, 22:52:

A DX2 CPU has clock-doubling circuit inside so it actually accepts 33MHz clock - and the bus is running at that speed as well. It's only the CPU internals that run at 66MHz, but since it does have on-chip cache it can run the code faster once it's fetched.
Can you remove the heatsink and provide a photo of your CPU top? Maybe it's not actually a DX2 after all?

I removed cpu and double and tripple checked its a DX2 😀

dionb wrote on 2020-01-28, 22:55:

That AMI Super Voyager is a VERY early 486 board. Makes me wonder if everything on it can actually support 33MHz... […]
Show full quote

That AMI Super Voyager is a VERY early 486 board. Makes me wonder if everything on it can actually support 33MHz...

According to TH99, J17 needs to be set to 1-2 for speeds over 20MHz. On the photo it looks like it's already set to that, is that the case?

Otherwise - do you have a 40MHz or 50MHz oscillator you could try? If neither works, it's probably something about the can. If at least one does, it's the speed of the 66MHz version something doesn't like.

Edit: the oscillator normally runs at *twice* the system bus speed. So for 33MHz bus, you need a 66MHz oscillator. OP is correctly trying a 66MHz oscillator for a 33MHz bus for a DX2 CPU that doubles that back up to 66MHz.

it should be good for speeds up to 66mhz, the J17 is set correctly. I tried lower speed oscillators aswell but it seems 33mhz is the correct one to use judgeing by other responses.

maybe its just as u say a verry old motherboard and that limits the speed.

dionb wrote on 2020-01-28, 22:55:

Edit: the oscillator normally runs at *twice* the system bus speed. So for 33MHz bus, you need a 66MHz oscillator.

Yes. If only this was a 386 CPU with CLK2 input.

DX2-66 CPU needs 33 MHz on CLK input.

dionb wrote on 2020-01-28, 22:55:

Edit: the oscillator normally runs at *twice* the system bus speed. So for 33MHz bus, you need a 66MHz oscillator. OP is correctly trying a 66MHz oscillator for a 33MHz bus for a DX2 CPU that doubles that back up to 66MHz.

I'm not that familiar with 486 chipsets so perhaps this one does require a doubled clock, but the boards I've seen (and I mean "pure" 486 mobos, not the 386/486 combo ones) use the bus clock as-is.
Also, why would this mobo have a 33MHz oscillator then? The slowest 486 in PGA package I know of is 486SX-20. And that's kinda rare actually. It should have a 40MHz one then, at minimum, and more likely a 50 or 66 by default.

Deunan wrote on 2020-01-28, 23:10:

I'm not that familiar with 486 chipsets so perhaps this one does require a doubled clock, but the boards I've seen (and I mean "pure" 486 mobos, not the 386/486 combo ones) use the bus clock as-is.
Also, why would this mobo have a 33MHz oscillator then? The slowest 486 in PGA package I know of is 486SX-20. And that's kinda rare actually. It should have a 40MHz one then, at minimum, and more likely a 50 or 66 by default.

Most 486 boards with oscillators should have 1x clock. Unlike 386's. At least thats what CPUs expects.

However, my SX-16 feels a bit under-appreciated after your mentioning of SX-20 😀

download/file.php?mode=view&id=75544

RetroFletch wrote on 2020-01-28, 23:06:

I removed cpu and double and tripple checked its a DX2 😀

The reason I'm asking is because there were some reports of AMD DX4 (or DX5?) CPUs being sold as slower DX2. Either not deemed fully stable or just for filling the market gap. These CPUs would actually respond to the CLKMUL signal so an older socket with this signal not wired properly might confuse them. So I was wondering if maybe you have some odd CPU like that.

Since you have access to a scope, you can check the actual CPU clock frequency yourself: http://ps-2.kev009.com/eprmhtml/eprmx/h12203.htm
Test what is being delivered to pin C3, and since you have a 487 socket nearby and it should be 99% wired the same way, you can use that to stick the probe in.

mpe wrote on 2020-01-28, 23:19:

However, my SX-16 feels a bit under-appreciated after your mentioning of SX-20 😀

Huh. I stand corrected then. Perhaps it's some special version, I know Intel manufactured 386DX for a looong time with pitiful (on paper) clocks for military applications. This seems to be a '92 CPU - by that time Intel was mostly out of the woods with the 486 clock issues. I bet this chip can easily run at 25MHz.

RetroFletch wrote on 2020-01-28, 23:05:

if i run CHKCPU it does report cpu internal speed as 66mhz and bus speed at 33mhz.

If you have a DX2-66 in the socket and CHKCPU reports this then I think there is no problem with your crystal. This tool is usually right (unlike whatever other benchmarks say or what's printed on boot). You might try to play with jumpers but likely the performance won't be as good as on a newer board. It should be better than 33 MHz CPU though.

mpe wrote on 2020-01-28, 23:37:

RetroFletch wrote on 2020-01-28, 23:05:

if i run CHKCPU it does report cpu internal speed as 66mhz and bus speed at 33mhz.

If you have a DX2-66 in the socket and CHKCPU reports this then I think there is no problem with your crystal. This tool is usually right (unlike whatever other benchmarks say or what's printed on boot). You might try to play with jumpers but likely the performance won't be as good as on a newer board. It should be better than 33 MHz CPU though.

Yeah its weird.. i only get 13.42 in speedsys wich is pretty low when i compare to other 486 benchmarks and similar ive seen.

If 13.42 is the CPU score in Speedsys then that's indeed quite low. Which CPU did you upgrade from? What was the score like before?

mpe wrote on 2020-01-29, 00:08:

If 13.42 is the CPU score in Speedsys then that's indeed quite low. Which CPU did you upgrade? What was the score before?

Yeah it's pretty much exactly what a dx33 would score 😀 I haven't upgraded the cpu its the one that was in the system when I got it of eBay. Maybe I should just keep an eye out for another motherboard.