I can't think of any. It might be easier to find the WB variant (P24D) in 3.3V than it is in 5V? Not sure.
As many ran a DX2-66 with no heatsink (some strong views on that) there could be benefit to 3.3V with no heatsink due to the heat issue you mentioned.

There are a number of small things. WB cache is commonly available in the 3.3V, which is a little faster, then also there were supposed to be a few late DX2s with 16kB writeback, which is faster again. Aside from sucking less power in the first place, these also have better power saving modes, sleep and hibernate states, but you have to get a board which supports everything right, and an OS with support or power saving drivers.

Also in the AMD 3.3V some of them are "secret" DX4s, sold as DX2 for market reasons but actually have the 3x multi also, usually completely happy at DX4/75 and rare that one doesn't do DX4/100

The benefits are relative, such as switching from i486DX2-66 to Am486DX2-66V16BGC, you get lower calorific value, a larger 16KB L1 cache, and a WB L1 cache.

But as we all know, Intel and AMD's same frequency 486 processor comparison, Intel's is faster, I don't know if anyone has compared i486DX2-66 and Am486DX2-66V16BGC, which one will be faster?

Am486DX2-66V16BGC has 3x multiplier, so it can easily beat any classic DX2.

ziggy587 wrote on 2024-07-25, 18:30:

Sorry if this question has been asked before... […]
Show full quote

Sorry if this question has been asked before...

I have a motherboard that supports both 5v and 3.3v variants. Right now I have a 5v DX2-66 that I'm happy with. Would there be any worth in hunting down a 3.3v variant DX2-66?

I've recently watched a YouTube build video where the YouTuber noted that he had a 5v CPU and said "it would be a shame not to use a 3.3v CPU since the motherboard supports it" but didn't explain why that is. And I seem to recall reading random things here and there that would leave one to believe that 3.3v is better or more desirable in some way, but I've never read anything explaining specifically why that is.

I do seem to remember (but I cannot find the source now) reading that the lower voltage CPU will run cooler. And that seems more desirable for obvious reasons. But is it really? I mean, the 5v CPU was designed to run at 5v. And if the 3.3v CPU has lower power consumption, does that actually affect anything outside of my power bill? I couldn't find much in Google regarding this, maybe I'm not using the correct search terms. But if I plan on using a heatsink and fan either way, are there any real benefits to using the 3.3v over the 5v CPU?

Well, the lower power consumption and in extension less heat generation is enough for me to generally prefer lower voltage parts. Taxes peripheral circuitry on the mobo like CPU VRM etc less, and allows for more silent cooling. The power bill is nothing to worry about in either case though, the difference is only a few watts. You'd probably have to run a 3.3V 486 for years and years before going break even on the investment.

You could also try running your 5V CPU on 3.3V or 3.45V if possible to see if it works. My 486 is an even older 5V DX33 - there is no 3.3V version - running 100% stable on 3.45V. Running it in a relatively modern SBC that doesn't do 5V at all.

The performance question is kind of moot if you ask me. If you want performance, get a Pentium MMX rig instead. Anything from the 486 era benefiting from it will work with a Pentium MMX aswell. Those can also easily be tuned in small steps using software manipulating the TR12 registers/caches, to cover everything down to slow 386 speeds. If you want it even more finely grained, you can wire up the FSB/multiplier jumper headers to switches aswell.

BitWrangler wrote on 2024-07-25, 19:29:

There are a number of small things. WB cache is commonly available in the 3.3V, which is a little faster, then also there were supposed to be a few late DX2s with 16kB writeback, which is faster again. Aside from sucking less power in the first place, these also have better power saving modes, sleep and hibernate states, but you have to get a board which supports everything right, and an OS with support or power saving drivers.

Also in the AMD 3.3V some of them are "secret" DX4s, sold as DX2 for market reasons but actually have the 3x multi also, usually completely happy at DX4/75 and rare that one doesn't do DX4/100

I'm sorry how can I identify the dx2 66 v3.3 with 16kb writeback?

I actually have 2 dx66 one is SX911 which is the common 5v and then a SX955 which is 5v but writeback.

Yeah, think I got it scrambled in my head, SX955 is the SL enhanced with writeback and is 5V ... and some utils seem to think it has 16kB... don't know a specific part that has 3.3V and 16kB writeback, I had some memory of reading something about it, but it might have been a post 1995 embedded offering or something.

BitWrangler wrote on 2024-07-26, 12:29:

Yeah, think I got it scrambled in my head, SX955 is the SL enhanced with writeback and is 5V ... and some utils seem to think it has 16kB... don't know a specific part that has 3.3V and 16kB writeback, I had some memory of reading something about it, but it might have been a post 1995 embedded offering or something.

I checked all the sspec on ccpu world and apparently there is only one which is 3.3 but it's not writeback. So yeah maybe it's for embdded platforms.

kaputnik wrote on 2024-07-26, 09:52:

Well, the lower power consumption and in extension less heat generation is enough for me to generally prefer lower voltage parts. Taxes peripheral circuitry on the mobo like CPU VRM etc less, and allows for more silent cooling. The power bill is nothing to worry about in either case though, the difference is only a few watts. You'd probably have to run a 3.3V 486 for years and years before going break even on the investment.

You can't tax the onboard VRM any less on a 486 motherboard than you can by using a 5V CPU 😁

> lower power consumption

we are talking like 2 Watts here

Since some 486 motherboards generate 3.3V with a linear voltage regulator, there may not even be any power savings.
As jakethompson1 alludes to, the CPU voltage regulator is entirely disabled when the motherboard is set to 5V.

akimmet wrote on 2024-07-26, 17:13:

Since some 486 motherboards generate 3.3V with a linear voltage regulator, there may not even be any power savings.

Using the same CMOS silicon, a digital component running at the same clock behaves towards operating voltage changes like a resistor: An increase of 50% in the supply voltage (3.3 to 5) also causes an increase of 50% in the supply current. On the other hand, if you supply a 3.3V CMOS chip using a linear regulator, the load voltage goes up from 3.3V to 5.0V, but the current stays the same. So there is some saving by using a 3.3V part and a linear resistor. Furthermore if there are later 3.3V parts that replace earlier 5V parts, it is oftentimes the case that the later 3.3V parts are after a die shrink, which tends to reduce power consumption even more (unless the die shrink causes the insulation layers to become so small that you get excessive leakage currents. Hello 90nm Pentium 4!).

A nice model to understand why it is that way: A digital chip that operates at some frequency has to toggle some internal states between low and high a given number of times per second. Changing the internal state means charging / discharging the gate capacitance of MOSFETs. The smaller the chip, the less electrons fit on the gate. Moving less electrons around means less current. Furthermore, if you charge and discharge the capacitors from 0V to 5V instead of from 0V to 3.3V, you need more electrons per charge/discharge cycle, so again, a higher current is required.

Thanks for all of the replies, everyone! For me and my use, it sounds like I will be sticking with the 5v CPU. 😀