clueless1 wrote:

In my mind, I imagine that the button is attached to a pair of wires with a jumper on the end, and the jumper plugs into a "turbo" header on the motherboard. Is that part right? Or does it work differently?

That's how it works for my 386 and 486 boards 😀

I just use a jumper. Some boards are fast with the jumper set, some with the jumper removed.

Can you define the slowed speed, or is it randomly slower?

Yes, early clones, basically 'Turbo XT' systems, generally had a keyboard combination. Many of them also came with a commandline utility to switch speeds (so you could enable the turbo automatically from autoexec.bat or other scripts).
Basically it just switches around some multiplier/divider settings in the chipset, so that the CPU can run at a different speed (while other speeds, such as ISA bus remain the same).

I have a 286, which is sort of a 'hybrid': it supports a turbo button via a header. However, as I found out, the turbo button does not work in all software.
The reason for this is that it is actually implemented in software. The BIOS installs a handler in the timer interrupt, and at every interrupt, it checks the state of the turbo button, and performs a speed switch if necessary.
When software overrides the timer interrupt, and doesn't call the original timer routine, the button no longer works.

On 386/486 boards, the turbo button is generally 'hardwired', so it doesn't need any software routines, and it can't be disabled. These systems generally also do not have a way to switch speed from software.

Perhaps the reason for this is that early cases did not have a turbo button at all. The 286 may be from a 'transitional period', where you may or may not have had a turbo button on the case.
In the 386/486 era, a turbo button was a standard feature on every case, so there was no need for software control anymore.

clueless1 wrote:

Can you define the slowed speed, or is it randomly slower?

It's not random. It's generally a fixed speed, chosen by the chipset designer.
Eg, a Turbo XT will generally run at 4.77 MHz by default, so it matches the speed of a real PC/XT.
On 286/386/486 the non-turbo speed is usually 8 MHz, so it matches the 8 MHz PC/AT.
(The only reason for this feature is backward compatibility: you only want your system to run slower for software that doesn't behave properly at the higher speed. So you pick a speed that gives you the best compatibility).
Some chipsets allow you to modify some multipler/divider settings in the BIOS setup, so you have some control over turbo/non-turbo speeds.

clueless1 wrote:

Can you define the slowed speed, or is it randomly slower?

Well on the Biostar 486 board I'm working on now, using the turbo lowers 3dbench 1.0 from around 26 to 10. Disabling cache lowers it to around 5. Without cache the turbo setting doesn't have any effect.

Thanks guys. Poor choice of words on my part when I said "is it randomly slower?". I meant is the turboed down speed inconsistent from machine to machine. 😀 From the research I did, it seemed like the original intent was for it to hit the 4.77Mhz speed, but it sounds like that wasn't always the case, especially as systems approached 486 speeds. And Phil's Biostar 486 drops it to mid-level 386 speeds, whereas disabling cache drops it into the 286 range.

clueless1 wrote:

Thanks guys. Poor choice of words on my part when I said "is it randomly slower?". I meant is the turboed down speed inconsistent from machine to machine. 😀 From the research I did, it seemed like the original intent was for it to hit the 4.77Mhz speed, but it sounds like that wasn't always the case, especially as systems approached 486 speeds. And Phil's Biostar 486 drops it to mid-level 386 speeds, whereas disabling cache drops it into the 286 range.

Looking at the board I have, every board implements the turbo a little bit different. Some do disable the cache, some don't. Wait states are often inserted and who knows what else they are doing. So yea, you got to benchmark each board individually. That's a good idea anyway as even with cache tricks there can be quite a variation depending on parts you are using.

I recently performed a couple of quick benchmark tests (using Landmark 2.0 & NSSI 0.60) to see what the different results were when I "de-turbo" my machine.
The intention was purely to see what would be a suitable PC to get me as close as possible to "fast" XT speeds or "slow" 286 AT speeds (8 to 10 MHz range).

On my AMD 386-SX 25 MHz, I was able to slow it down to the equivalent speed of a 286 8 MHz (this is now when I add a 1 wait state for the onboard RAM in the BIOS setup, otherwise it's closer to that of a 286 10 MHz CPU).
On my AMD 386 DX 40 MHz, I was able to slow it down to roughly that of a 386 SX 20 MHz (can't quite recall now whether this was after I disabled the external cache in the BIOS setup).
On my AMD 486 DX4 100 MHz, I could slow it down to the equivalent speed of a 486 DX 33 MHz (this is still with L1 & external cache enabled in the BIOS setup).

So, the faster the CPU, the less likely that you will get to the original XT speed of 4.77 MHz.

My 386 SX 25 MHz (in normal mode) is already so slow that my sound card is struggling to properly produce digital sounds in some older demos or games (most likely because of the lack of external cache and the slow CPU speed). Wolfenstein 3D does run nicely on it 🤣.

^^ That's pretty much what I'm seeing also 😀

I thought I read somewhere here that different motherboards achieved this though all sorts of methods, maybe disabling cache, timers, or whatever. basically there was no standard way, therefore results would also not be standard.

Unisys Pw/2 Series 300 (286-8/10mhz) have no Turbo button. The clockspeed can be altered from a keyboard command during boot time. Or it can be set from within the BIOS.