• Please review our updated Terms and Rules here

Do 486 DX2/DX4 really need a fan?

Frankly, I don't think PC electronics suffers from a nightly shutdown as even 365 cycles a year is still very small for thermal cycling stress. And if it can't take the electrical transients, it hasn't been designed correctly. My only exception would be PCs with older hard disks that have the brown oxide coating.

Well, I was not referring to thermal or transients, as much as I was current inrush, as a problem. You see this specifically with poorly made PS's on clones and secondary brand PCs, but it happens more than I care for. The overcurrent suppression is just inadequate, and overtaxes all of the rails.

This may have never happened to you, for which I am happy, but it happens to many, specially when living in an area (like I do) where the mains are always overvoltage (or under) and throwing a lot of spikes (so I guess partially, I was talking transients), like when you know you've got a bad utility, when you can sit there and watch light bulbs popping out on you every few days.

Upshot is, no problem leaving it on if you have good battery backup to do all of the pre-filtering for you...

gwk
 
I'm sure the Evergreen module uses the ADZ version of the AMD 5x86, which has the highest temperature rating. (The ADZ was supposed to be sold as a 160 MHz chip, but AMD feared it would cut into K5/K6 sales too much, so they sold it as a "133 MHz" chip.)

*grabs dead Evergreen upgrade off shelf*

AM5x86™-P75
AM486DX5-133V16BHC

According to CPU World it has an 85C TJunct.

Frankly, I don't think PC electronics suffers from a nightly shutdown as even 365 cycles a year is still very small for thermal cycling stress. And if it can't take the electrical transients, it hasn't been designed correctly. My only exception would be PCs with older hard disks that have the brown oxide coating.

Except you're forgetting about BGA mounted parts. Thermal cycling isn't a big issue on through-hole, DIP, SOP, PQFP and other similar chip mounting solutions, which is almost exclusively used on machines 25 years ago; But it's a massive problem with BGA parts, which almost everything uses today.

Chips don't heat evenly, and can have temperature swings of 30-50C within a matter of seconds. This leads to all sorts of weird stresses on BGA parts, especially if they're mounted using a solder that isn't flexible enough to deal with it. This is why all of those Xboxes, Play Stations, modern graphics cards, laptops and even desktop motherboards have failed, because of thermal stress on BGA parts.

Keeping these machines running 24/7 drastically reduces failure rates because the solder subject to less thermal shock. Instead of having to go from room temperature to 60-70C, you're only experiencing swings of 10-30C in normal use and a bit more on power hungry parts like GPUs.
 
Except you're forgetting about BGA mounted parts. Thermal cycling isn't a big issue on through-hole, DIP, SOP, PQFP and other similar chip mounting solutions, which is almost exclusively used on machines 25 years ago; But it's a massive problem with BGA parts, which almost everything uses today.

Don't forget that Pb-free solder is sticking those BGAs to the PCB. Given that most consumer-level stuff manufactured today has a very short design lifetime, that should surprise no one.
 
I don't think 60/40 would fare any better on a BGA part.

But the disposability of electronics today is certainly a problem.
 
Thanks for the replies to this thread, very interesting reading and far more in depth than I anticipated! I will certainly add fans to some of my fanless systems (including a Compaq Presario, which came without a fan from new despite the label on the CPU! :eek:)

Just need to find some (40/45mm?) that don't sound like angry wasps now...
 
I don't think 60/40 would fare any better on a BGA part.

But the disposability of electronics today is certainly a problem.

Well, apparently one of the issues is that the lead-free RoHS stuff isn't nearly as ductile as the leaded stuff, which is why there are exceptions to RoHS regulations for things such as military and aerospace.

A couple of weeks ago, I visited a wind instrument builder (not a repair tech--he draws his own tubing, for example) He was bemoaning the terrible characteristics of RoHS solder used to build instruments such as saxophones. He said that after a few years, the joints simply fall apart--and that would never happen with the 50-50 solder that was customarily used. I remembered my old Volvo 960 and that I ended up resoldering every relay board in the thing--Bosch was trying to make reliable PCBs using RoHS solder and it wasn't holding up. Then there's the problem of "whiskers"...
 
Metal whiskers are a problem in pretty much every type of solder, leaded or not.

I avoid ROHS solder whenever possible. Ductility is a problem, but the bigger problem is that I can never get it to melt properly. I have a variable temperature soldering iron and no matter the temperature setting, it never melts and makes joints right.
 
... where the mains are always overvoltage (or under) and throwing a lot of spikes ...

I can't say I've ever had a PS fail due to mains issues (outside of my control) but that may be just down to good mains power, good quality PCs and a heap of good luck. I have seen the effect of a HV spike on a microwave oven in a nearby property where it arced from the turntable motor winding to ground and afterwards caused a small leakage to ground via the resulting carbon track. We would not have noticed except the household GFI (called an RCD here) would not hold.
Due to the high cost of power here I minimize what is left plugged into mains but do have a cube PC (Intel NUC) running 24/7. Intel provide a high quality switch-mode wall-wart with the NUC.
 
Except you're forgetting about BGA mounted parts. ... laptops and even desktop motherboards have failed, because of thermal stress on BGA parts.
Yes, true, I did have had a BGA GPU fail on my Dell Inspiron after 5 years.
No doubt good BGA thermal-cycling life is matter of good engineering, but as you say, they certainly they didn't make it easy for themselves by eliminating the mechanical compliance of leads. Decades ago I worked in a company that designed and manufactured automated solder joint inspection systems using x-ray imaging and we built the first 3-D tomography system specifically for BGAs. Being able to directly view solder joints was the first consideration for our customers but once the process parameters were refined, quality control boiled down to process control and x-ray inspection became a superfluous expense.
 

Attachments

  • xy01.jpg
    xy01.jpg
    60 KB · Views: 1
  • Oai-xr01.jpg
    Oai-xr01.jpg
    13.1 KB · Views: 1
they certainly they didn't make it easy for themselves by eliminating the mechanical compliance of leads.

It was out of necessity for them. You can only stuff so many leads on the edges of a PQFP without making it so large it becomes ridiculous, or making the lead pitch so fine that whiskers become an issue.
 
Heat is the enemy of all electronics. Any bit of kit that will run with no fan installed will last longer with a fan--or some other type of cooling. That's not to say that a fan mounted on a heatsink is the best solution. I'd much rather see ducting from the heatsink to a generous PSU fan be used. Small fans just don't move much air and use the heated air from the inside of the enclosure to cool. Better to take in outside cool air to do the job.

I have been wondering about a BTX2 standard that fixes the problems of the original BTX standard for that and other reasons.
 
BTX was a terrible motherboard specification. The only company it benefited was the company that designed it, Intel.

As they couldn't work out how to reduce the power draw and heat output of their P4s, they just redid the whole motherboard so they could use ever more power hungry CPUs.

Systems that used BTX motherboards were often so cramped that you couldn't install a higher end discrete graphics card because the CPU heatsink gets in the way. Some designs put the main power connector right behind the PCIe slot, further limiting cards. Not to mention that the RAM slots were often stuffed near the edge of the motherboard, usually ending up being placed under the optical drive bays and an array of wires, making it an arduous task to add/replace memory modules.

BTX has a place in the history books and it needs to stay there, we don't need it back.
 
BTX was a terrible motherboard specification. The only company it benefited was the company that designed it, Intel.

As they couldn't work out how to reduce the power draw and heat output of their P4s, they just redid the whole motherboard so they could use ever more power hungry CPUs.

Systems that used BTX motherboards were often so cramped that you couldn't install a higher end discrete graphics card because the CPU heatsink gets in the way. Some designs put the main power connector right behind the PCIe slot, further limiting cards. Not to mention that the RAM slots were often stuffed near the edge of the motherboard, usually ending up being placed under the optical drive bays and an array of wires, making it an arduous task to add/replace memory modules.

BTX has a place in the history books and it needs to stay there, we don't need it back.

That is why I talked about BTX2 and not bringing back original BTX. For example, it would be designed for today's CPUs with integrated memory controllers and with single-chip "chipsets".
 
Last edited:
As far as I'm aware, there is no BTX2 standard.

I don't even know what you'd change to make it better because there's only so many ways you can make a motherboard. The memory slots dictate how the motherboard is designed because they need to be as close to the CPU and NB as possible.
 
As far as I'm aware, there is no BTX2 standard.
I know it doesn't exist yet.
I don't even know what you'd change to make it better because there's only so many ways you can make a motherboard. The memory slots dictate how the motherboard is designed because they need to be as close to the CPU and NB as possible.
For one thing, optical drives are becoming obsolete nowadays and I think there are now cases that don't bother with such bays.
 
Last edited:
Thanks for the replies to this thread, very interesting reading and far more in depth than I anticipated! I will certainly add fans to some of my fanless systems (including a Compaq Presario, which came without a fan from new despite the label on the CPU! :eek:)

Just need to find some (40/45mm?) that don't sound like angry wasps now...

Well, for what it's worth, if a modern system with a CPU that has ten times the power dissipation of that DX2/DX4 can be made fanless, then the older kit can be, too. My favorite example of a fanless modern PC is the Aleutia R2 (Core i3, 16GB RAM, up to 1TB SSD, and the coolest case ever ); see http://www.aleutia.com/products/r50-fanless-server/
 
Well, for what it's worth, if a modern system with a CPU that has ten times the power dissipation of that DX2/DX4 can be made fanless, then the older kit can be, too. My favorite example of a fanless modern PC is the Aleutia R2 (Core i3, 16GB RAM, up to 1TB SSD, and the coolest case ever ); see http://www.aleutia.com/products/r50-fanless-server/

Not 10 times, but about double. The i3 model used has a TDP of 15w though uses considerably less at idle; the DX2 and DX4 models range from 5w to 8w. That case is a heat sink with a few ports. One could do the same type of case with a 486 but the design would have to start with solving the cooling issues and accepting typical cases and motherboards will not be suited for the task.
 
Not 10 times, but about double. The i3 model used has a TDP of 15w though uses considerably less at idle; the DX2 and DX4 models range from 5w to 8w. That case is a heat sink with a few ports. One could do the same type of case with a 486 but the design would have to start with solving the cooling issues and accepting typical cases and motherboards will not be suited for the task.

Ok, double. :) Although at the 5W end of things it's triple.


Aleutia also has the M200 with an E3-series Quad core Xeon that's fanless. http://www.aleutia.com/products/m200-fanless-server/ That Xeon has a TDP of 45W, and the total case load can be up to 75W, fanless.

Something like a modern laptop's heatsink and heatpipe with radiators could easily be retrofitted to a 486 in a baby AT case and would likely cool passively with no problem. I just bring up the Aleutia fanless PC's as a recent example of what can be done in the fanless space.
 
Back
Top