How long it will work is unpredictable. It certainly could work at that voltage, but may degrade in stability over time.

Maybe there is something wrong with the voltage regulator?
Personally, I wouldn't want to overvolt my 25 year old CPU on a permanent basis and potentially shorten its lifespan.

I have a later revision M919 that has the proprietary cache slot and had the same issue getting an AMD 5x86 133 running on it.

I found the same thing in that the chip doesn't seem to get hot. Supposedly there is a pin in the CPU socket you can ground which on supported boards will switch the voltages from 5V to 3.3V. I haven't taken a multi-meter to it, but it's possible that when that pin is grounded it's cutting the voltage to the chip so setting jumpers to 3.3 or 4V could cut the voltage too low for the chip to run.

I haven't used this configuration too much, but it seems fine so far.

Not sure about M919 , but many S3 boards sometimes support power detection sense with DX4 cpus. Thus if you put a 3.3V CPU with low signal on VOLTDET pin (Intel DX4, AMD DX2/DX4, Cyrix 5x86, ...) they automatically switch to 3.3/3.45V irrespectively if you have your jumpers set to 5V.

The only way is to check the voltage by multimeter.

Try to use the 5x86, but jumper it for 75 MHz (3 x 25 MHz). See if it boots then at 3.3 or 4.0 V. My suspicion is the VRM is outputting lower voltage then it should, but it might be enough to run the CPU at such a lower frequency. I have an Asus PVI-486SP3 that exhibited very similar behaviour, any 3.3 V CPU I tried refused to work above ~80 MHz, I would get the same sad ---- on the diagnostic card. Turned out the VRM was only providing about 2.4 V instead of 3.3 V.

Best would be to just measure the VRM output with a multimeter, if you have one.

So, the CPU won't boot at any speed when set to 3.3 or 4v.

mpe wrote on 2021-03-15, 14:05:

Not sure about M919 , but many S3 boards sometimes support power detection sense with DX4 cpus. Thus if you put a 3.3V CPU with low signal on VOLTDET pin (Intel DX4, AMD DX2/DX4, Cyrix 5x86, ...) they automatically switch to 3.3/3.45V irrespectively if you have your jumpers set to 5V.

The only way is to check the voltage by multimeter.

Do you happen to know the pinout/where I can measure the VID?

If you mean the voltage then you can measure between any VSS and VCC pin (highlighted blue and red below) on the picture below. This is the pinout as you see in the socket.

mpe wrote on 2021-03-15, 14:49:

If you mean the voltage then you can measure between any VSS and VCC pin (highlighted blue and red below) on the picture below. This is the pinout as you see in the socket.

S4 on the pinout is the VOLTDET pin. I believe it needs to be grounded by connecting it to one of the VSS pins for the voltage drop to occur.

The CPU will eventually die when ran at 5V, the silicon is not meant to handle it and has breakdown voltages just at this edge.
The onboard voltage regulator is likely dead, I have had to replace transistors etc. before on few boards I have around which brought them back to life.

Honest question, I have one that I've run at 5V (with active cooling) before, though currently running at stock voltage/speed.

Question: Has anyone ever killed one at 5V? I've heard this repeated many times "it will eventually die" but - is there actual experience of this occurring?

Tiido wrote on 2021-03-15, 18:59:

The CPU will eventually die when ran at 5V, the silicon is not meant to handle it and has breakdown voltages just at this edge.

Usually you'll get silicon degradation which may lower overclocking potential or, in worst case, stability on stock speed. Then again, it's a slow process and it won't happen in one day. AFAIK nobody really documented how bad it could be on fairly robust lithography.

I have a m919 and you can check the cpu voltage on either center leg of the two reg chips or (safer) the big solder pad under them against ground. They seem to run in parallel with their outputs combined (old observation on one of my notes but could be wrong).
added: mine has two Vreg chips, have seen some board revs that have just one.
Easier than trying to find a specific pin or via to check on back of board. Best to check it w/o cpu and then with cpu when set to 3.3v or 4v and compare. Should be a bit a V drop with cpu but not more than a few tenths of volt...

macroexp wrote on 2021-03-15, 23:55:

Question: Has anyone ever killed one at 5V? I've heard this repeated many times "it will eventually die" but - is there actual experience of this occurring?

A friend ran his 5x86 at 5V unknowingly and even with proper cooling the CPU started to get unstable after few weeks of use and then stopped working altogether. It definitely happens, it just takes some time. A lot of 3.3V devices can run at 5V for a little while but it is firmly beyond their safe operating area and there will be damage happening.

Done some probing with the multimeter without a CPU installed.

On a VCC pin:

When set to 3.3v the board gives 3.52v.
When set to 4.0v the board gives 4.03v
When set to 5.0v the board gives 5.11v

On the VRM (my board has a single VRM on the right side when looking down on the board from the VLB end):

Regardless of what voltage is set, the left pin is 4.05v, the middle pin is 5.17v, and the right pin is 3.55v. The large pad under the VRM is 5.17v.

Seems that the board is generating the right VCC for the CPU, but the CPU doesn't boot at anything other than 5v.

Already failing CPU is the most likely I guess.

paradigital wrote on 2021-03-17, 11:48:

Done some probing with the multimeter without a CPU installed. […]
Show full quote

Done some probing with the multimeter without a CPU installed.

On a VCC pin:

When set to 3.3v the board gives 3.52v.
When set to 4.0v the board gives 4.03v
When set to 5.0v the board gives 5.11v

On the VRM (my board has a single VRM on the right side when looking down on the board from the VLB end):

Did you ground the VOLTDET pin and receive the same voltages?

SScorpio wrote on 2021-03-17, 12:00:

paradigital wrote on 2021-03-17, 11:48:

Done some probing with the multimeter without a CPU installed. […]
Show full quote

Done some probing with the multimeter without a CPU installed.

On a VCC pin:

When set to 3.3v the board gives 3.52v.
When set to 4.0v the board gives 4.03v
When set to 5.0v the board gives 5.11v

On the VRM (my board has a single VRM on the right side when looking down on the board from the VLB end):

Did you ground the VOLTDET pin and receive the same voltages?

I hadn't remembered to try so have just done that now.

Regardless of S4 (VOLTDET) tied low or not the VCC is the same, follows what is set on the board (3.3v, 4v or 5v respectively).

EDIT:

I've no idea if either of my DX2-66's or my original 486 DX (33) are capable of running undervolted or not, but none of the three will run at 25MHz bus @ 4v on this board either, granted it's a 20% undervolt, but it's also a 25% underclock.

Well I've ordered a cheap Cyrix CPU that's meant to run at 3.3-3.6v and we shall see. If the Cyrix ends up booting with the board set to 3.3v or 4v then I guess my Am5x86 is already dying?

If neither boot, well then I'm just plain confused.

Ok, so Cyrix DX2 v66 turned up and it too will not boot with the board set to 3.3v or 4v. Interstingly though, the VRM behaves differently with the Cyrix installed (I guess it’s VOLTDET pin it grounded internally), the left and middle legs of the VRM show 5.06v, but the right pin only 2.06v or 2.19v respectively when the board is 3.3 or4v.

Looking like I need to do a VRM replacement.

Does anyone know if the passive components for a dual-VRM setup are present on the single VRM variant of the board? Tempted to put a 2nd in.

Cyrix 486 has quite a different pinout to other 486 CPUs. It doesn't have VOLDET pin. That pin on Cyrix is used for cache invalidation as far as I remember.