Well one disadvantage is that it will draw a bit over 1 amp from the 5V line, and machines from that era tended to run primarily off the 5V line; 1 amp in the resistor means 1 amp less for the computer. Make sure that the PS can still deliver the required amps on 5V with the resistor on.

I won't bother with the resistor.

Your 286 already pulls enough power from the 5V rail.

The article you reference is good if you are using a cheap group regulated PSU and only using the 12V line from it for some other purpose (3d printer)

Some ATX --> AT adapters come with one of those resistors in the 3.3V line, but that's because 3.3V is not used at all by AT style computers. I don't think adding a resistor to 5V would help in your case.

Seems a bit weird to worry about the 3.3V rail with the ATX ---> AT adaptor.

First of all, 3.3V is not being used, so who cares if it goes a bit out of spec.
Secondly, old computers on the AT form factor pull very little power anyway, 30-40W being pulled on the 5V rail isn't going to radically unbalance any PSU except for the utter worst, which you shouldn't be using on a retro build anyway.

Thanks guys. This all sounds quite reasonable. I think my PSU quite ok. Not top of the line, but it’s modular and looks to be well built. Was 60€ I think. And it seems to run fine as I said.

These resistors are only ever needed to act as a load for a PSU where a particular rail is not used so it could operate propely. Some PSUs will not regulate the voltages correctly when a rail is unloaded.
In case of that 286 machine this is not a problem whatsoever, that resistor will not do more than waste power and increase temps in the machine.

canthearu wrote:

Seems a bit weird to worry about the 3.3V rail with the ATX ---> AT adaptor.

It is actually not weird at all.

canthearu wrote:

First of all, 3.3V is not being used, so who cares if it goes a bit out of spec.

sure, 3.3V out of spec is not going to be the problem. The other rails going out of spec because of unbalanced regulation however is. Some PSUs will not regulate properly unless all have some level of load on them. Then there are PSUs that have protections that will shut down if there is no load on some rails. It makes sense since any modern hardware will use 3.3V, 12V and 5V and the lack of load on one of them could indicate a fault.

canthearu wrote:

Secondly, old computers on the AT form factor pull very little power anyway, 30-40W being pulled on the 5V rail isn't going to radically unbalance any PSU except for the utter worst, which you shouldn't be using on a retro build anyway.

The main problem with running modern PSUs at low load is that their efficiency drops drastically. As long as the PSU is not 100% chinesium crap, low load (no load is a different story) shouldn't cause major stability or regulation issues.

Koltoroc wrote:

sure, 3.3V out of spec is not going to be the problem. The other rails going out of spec because of unbalanced regulation however is. Some PSUs will not regulate properly unless all have some level of load on them. Then there are PSUs that have protections that will shut down if there is no load on some rails. It makes sense since any modern hardware will use 3.3V, 12V and 5V and the lack of load on one of them could indicate a fault.

The answer isn't to mess around trying to get a crappy power supply to output nicer power, it is to replace it with a PSU that is built properly and isn't a major risk to your valuable retro hardware.

Koltoroc wrote:

The main problem with running modern PSUs at low load is that their efficiency drops drastically. As long as the PSU is not 100% chinesium crap, low load (no load is a different story) shouldn't cause major stability or regulation issues.

Not really a big deal, if your PSU is 60% efficient at 30W output, then you waste an additional 20W of power. Not great, but hardly a bank breaker or the end of the world, especially compared to the waste power of a PSU when running under high load. If you have an 90% efficient PSU, but are loading it to 400W, you are still going going to be losing 45W during the conversion process.