Question Why no fins on VRM heatsinks nowadays?

[WelshBloke]

I would trade all those RGB for the return of proper copper heatsink.
View attachment 6706

This sort of thing get hideously clogged with dust really quickly. Most people aren't cracking open their cases every 6 months and blowing out the dust.
Heat sinks with flat aluminium are going to cool the same in a year without any cleaning.

 

[Markfw]

Well, my DQ6 pictured above was running non-stop for years when I took that pic. No clogging there, same type heatsinks.

 

[Furious_Styles]

This sort of thing get hideously clogged with dust really quickly. Most people aren't cracking open their cases every 6 months and blowing out the dust.
Heat sinks with flat aluminium are going to cool the same in a year without any cleaning.

I generally follow the rule of cleaning once a year. And maybe blowing out some dust filters during that year if it gets bad. But I don't see how it will get clogged with dust unless you are in some kind of environment that's really bad. Or you just never clean, which I'm sure applies to some people.

 

[WelshBloke]

I generally follow the rule of cleaning once a year. And maybe blowing out some dust filters during that year if it gets bad. But I don't see how it will get clogged with dust unless you are in some kind of environment that's really bad. Or you just never clean, which I'm sure applies to some people.

Go crack open some computers at work if you can, ours are hideous. Opening them and blowing the dust out is not the norm, they just sit and cook and get replaced.

 

[Markfw]

Go crack open some computers at work if you can, ours are hideous. Opening them and blowing the dust out is not the norm, they just sit and cook and get replaced.

Well, in any data center I have seen, its almost like a cleanroom with the AC and filtration.

All the desktops are so small now, I don't even know if they have internal fans.

 

[mindless1]

It is funny how nowadays motherboard manufacturers treat heatpipe or properly filled heatsinks as a "luxury items" and they put them only on the premium boards.

Why is it funny? After board manufacturers started using 2oz copper foil on the PCB, ultra low RSD 'fets and more stages, it became easy to hit over 100W TDP without 'sinks on the fets, unless the board was really crowded which is sometimes the case on premium boards due to more/longer-slots and chips.

They're more about overclocking than anything. Well that and looking fancy, but it's barely even functional.

If they really wanted to keep the 'fets cool, they would tombstone them (leaded type TO-126/TO-220/etc, not SMD) and mount their heat spreader to the heatsink, not copper foil on the PCB, and then would barely need any heatsink surface area at all to exceed the performance of the best, traditional top-mounted 'sinks.

 

[Furious_Styles]

Go crack open some computers at work if you can, ours are hideous. Opening them and blowing the dust out is not the norm, they just sit and cook and get replaced.

Where is the 21st century janitor when you need him?

 

[Jimzz]

The MOSFETs and other parts used in the VRMs have improved a lot, making them more efficient than they used to be. They barely need heatsinks at all, leaving the motherboard manufacturers free to design heatsinks that look good rather than ones that cool well.

I disagree. If you don't push your computer/use more than a quad CPU then yea the VRMs are fine as is.
But with 6, 8, or even 16core chips made for desktop now the VRMs get very hot. Some have even been tested to exceed 100 degrees under load.

VRMs are the biggest thing I look for on boards now. The other things are so similar the biggest difference is the power sup[ply system first and BIOs setup 2nd. Some board go above and beyond on VRM design/layout and heatsinks. While others try to lie about how many phases they have, have bad heatsinks or none at all, and some just use such low quality parts that it can really shorten the life of the board.

 

[Mr Evil]

...Some have even been tested to exceed 100 degrees under load...

That's not too much for a MOSFET. Most of them are rated for at least 150°C, and fancy new Silicon Carbide ones will survive 200°C.

 

[Jimzz]

That's not too much for a MOSFET. Most of them are rated for at least 150°C, and fancy new Silicon Carbide ones will survive 200°C.

If all board makers used good VRMs and built boards that can handle that temp I would agree. But I think we both know thats not the case. I believe most board makers say to keep the VRM temps below 100.. I think the issue is more on the board side then the VRM side my guess.
Let alone Intel for example getting very fast and loose with TDP ratings is not helping either.

 

[IntelUser2000]

Thermal issues are likely going to get worse as HPC guys are quoted as saying we'll soon see accelerators using 1KW(that's 1000W for you).

Performance demands continue to rise while power reduction from new processes are decreasing. You can mitigate some with a more efficient architecture, but there are limits.

The unexpected slowdown from process shrinks are what delayed the first Exaflop computers from 2018 to 2021. The original target was 1EFlop using 20MW. Now its unlikely we'll see under 40MW in 2021 with the same performance target.

It's still progress. And compute isn't the only thing improving. Our cooling systems and methodologies have improved noticeably over the years. 10 years ago you needed the CPU to be 3W or lower to make the system fanless. Now some premium systems can cool a 9W chip.

 

[Kocicak]

Gigabyte and Asrock have nicelly finned heatsinks on TRX40 boards, Gigabyte ones have very fine pitch which depend on supplying proper external airflow (quite strong to be able to blow through the fine pitched fins), while Asrock have more coarse fins with its own air flow , I believe that Asrock solution may be a better one.