Slashdot Mirror
A Look At The World of Heatsinks
A reader writes: "There's an interview with Glenn Summerfield, Senior Sales Associate for Alpha Novatech (USA) that talks about heat sinks and where some of it is going." Alpha Novatech is one of the big boys in the field of heat sinks - the responses do have a bit of "salestalk" for Alpha Novatech, but seeing industry thoughts on watercooling vs. aircooling and such is interesting, nonetheless.
A bit of salestalk? That guy was full of hot air!
blah, blah, stop groaning
ALPHA: I can't go in to too much detail, as it's a highly proprietary process. However, the copper is embedded in the base at the same time that the fins are formed during the forging process. This takes place under a tremendous force.
That's their way of saying they've got Superman locked in the basement crushing heatsinks with his fists. They're holding him hostage with Kryptonite.
Free Superman!
Seriously though, here's the megacorp that just got some free but arguably useless press.
Obviously while sucking is good, blowing is better!
*narf!*
Fogey mode: You used to see stories like this posted to Slashdot all the time back in the day. Back then, there was no Katz or fluffy BS, just hardcore tech geekiness and Microsoft bashing (yeah, well, some things never change).
Reading over this article was like seeing an old friend again.
Every year during my review, I just pray the words "slashdot.org" aren't mentioned.
or at least for right now.
... kinda forgot to return the processor) and guess what? Neither one of us really noticed much of a difference. That's 133 MHz difference, a full 33MHz FSB boost too, and for just about everything that we did, we didn't notice a difference.
... well frankly I don't see the point other than just the coolness factor of having glow-in-the-dark coolant running through your PC =)
So I have a TBird 1.5GHz machine, and it hits 100% processing capacity for 0.02% of the time that it's in use. Save the time that I do mpeg compressing, but that's rare and the difference in speed that a watercooled system would give me over this would be negligable.
Using an expensive watercooler solution (well, expensive compared to a $30 air cooler that will work perfectly fine) to squeeze out a few extra megahertz, is that really worth all of the hassle?
Even at work I had an P3/800 (don't start with the intel bashing) and did a lot of work daily including compiling, and was upgraded to a P3/933 (at the expense of my friend who was away for a week, and I
If you're just so keen that you need to go from 33FPS to 33.8FPS in your Quake3 games, then, well, your choice... But is there really any good practical applications where the cost of a watercooled solution is worth the price? Keep in mind that you're comparing not the marked speed of the chip, but the speed that you could overclock to with air vs the speed that you can overclock to with water. I'd be surprised if you can see more than a 5MHz FSB difference there, even if 10MHz difference at a 15x multiplier (which means that you're already at 1.5GHz) you can gain 150MHz, which will do what for you? Practically nothing. Now add int he factor that you could just add the money of the watercooled solution to the cost of the chip to get the next higher up model, and
And before someone starts on the noise levels, we have a Dell 1.5GHz P4 at work that you have to put your ear right next to the bloody box to hear anything at all, they are VERY quiet. They have 2 fans too (1 case that blows across a heatsink on the proc through a tube, and 1 for the power supply). They are both thermally throttled and the hard drives are mounted on a thin strip of rubber too. Trust me, these babies are SWEET boxes and QUIET too.
So, who can explain to me why this would be worth it? I'm curious to know =)
-- Sean
If God gave us curiosity
I would have liked to see more focus on using the heatsinks without fans. I wish we could get away from moving the air past the heatsink with a big-ass whirring fan.. Between the fans and drives in my systems I can barely hear the phone ring. I've started putting foam and dynamat in my boxes to quiet them down, but I'm thinking it would be cheaper and simpler to just duck-tape the foam to my ears..
air and light and time and space
I read several reviews when building a Tbrird 1.333 system for where I used to work to compare to the latest and greatest Cow-boxen p4's we were getting.
Most of the Reviews said the FOP38 was the most common found to cool Ghz Tbirds, but if you can find an Alpha, get it!
Well, I did (pcnut.com, if you are interested) and it works like a beauty...even had a disclaimer "caution: loud!"...mild understatement.
Under a full load for 5 hours (decoding/converting DVD's to DivX, highest thread priority) the max temp of the processor was 111 degrees F. Some of the newer ones average 110 to 115...not bad for an "oldie but a goodie".
Tho I do admit some of the reviews now use thermal diodes under *and* above/near the top of the processor.
Another reason to go for the alpha is it won't crack those poor durons/tbirds...not so much a problem with the Athlon XP, or so the common wisdom/thinking goes because the new chips use a fiber composite (IIRC) instead of ceramic. So if the chip is under excessive pressure, the base will flex.
Just a satisfied customer here...I plan on getting 2 for a Dualie system I want to build.
MP or XP Tboids (heh, I like saying it tha way, mad props to N.Y. for enhancing the english language!) or whatever their new names will be (sledge/clawhammer?).
Heat sync technology has had to keep pace with the current procs (and the proc's current..err..watttage), so far so good, but a 1lbs heat sync on the P4...oye.
We can only hope they keep up the good work making better syncs, ducts, fans lest things all go up in smoke.
Moose
Have you read the moderator guidelines? Well, have you, PUNK? (and I want a Karma: Gnarly option)
It's a good thing they're looking into better heatsink design. Because what if one happens to fall off, guess what happens?
And you can't blame my AMD for that! It was the heatsink not being there!!!
Keep in mind the noise comes from the FANs in your system (well, okay, the HDs, too, but mainly the fans). There _are_ fans that are quiet! Oftentimes the HSF fan is the loudest one in the system, too, so check your system carefully.
The neat thing about high-end coolers like the Alpha PAL8045 is that the heatsink itself is so efficient that a 'whisper' fan with low airflow can still effectively cool a CPU. I just won a free Athlon XP 1800+ (1.53gHz) last week, and I'll probably be doing just that - a PAL8045 with a quiet CPU. I don't plan to bother overclocking the CPU, as it's already freaking fast.
I went to some sites that specialize in cooling products, and when you buy an HSF at some of them, they let you choose the fan that comes with it, which is nice.
And if you're 'stuck' with a really loud HSF - just replace the fan itself - they usually cost less than US$10.
I recently set up a server at work and had a bunch of old pentium class PCs to do it with. Swapping parts I was able to put together a P233 with 256 MB SDRAM w/2 HDs (a whopping 3 and 2 GB) FIguring that this server was going to be running 24/7 I decided it would be good to add some more cooling capabilities to the unit. .3 A fan and a medium riser on the voltage regulator next to it. Looking at my spare parts, i had another CPU fan of .2 A, which I decided to mount directly on top of the other fan on the CPU. So Now I had a double-decker fan system on the CPU, that was loud but pushed some serious air down onto the CPU. Across from the CPU was a fan on the front of the case with a duct that pointed towards the CPU heatsink. The crappy case design was choking this fan and it was hardly moving any air. So I cut away the grill on the case (the plastic front cover would act as a plenum and a guard) and removed the duct. This gave me enough room to install 2 fans (power supply size- about 3"across) again both blowing towards the CPU. Under full load the CPU would only rise about 10-15 degrees from ambient.
With the extra parts I had, I removed the small stock aluminum heatsink and replaced it with a heatsink with 3 inch riser. Then the voltage regulator (I think this was what it was- it was a "clip in" unit that got very hot and had its own small heatsink), which was located directly next to the CPU, was so hot that I replaced it's small 1" square heatsink with a medium sized heatsink from another PCs CPU. Now I had a big riser on the CPU with a
What I am wondering is just how effecient my design is. You rarely see any mods with dual fans. Maybe the noise is bothersome, but this was a server that was going to sit in a corner so I don't care. So, would two fans (presumably the stronger pulling through the weaker is the best) mounted together both blowing in the same direction be effecient? What do fellow slashdotters think?
There was recently a physics presentation at my University about thermoacoustics and its ability to be used for heatsinks. The basic idea behind it was that a thermoacoustic engine could be made to take the heat from the processor and convert it into sound, dissipating the heat more effectively than conventional heatsinks. The group working on the project already has a number of prototypes and showed some of them at the presentation and they were quite impressive! These 'engines' are already being made smaller than a penny in order to fit a number of them on a processor to increase cooling ability. And if you're worried about the sound coming out of them (the heat is converted to sound), the engines are converting it to high enough frequencies that it's undetectable to the human ear. They also told us that they're working on converting the sound back into electricity, perhaps to be used to cool the processor even further. I can't wait until these are commercially available...
Once upon a time software engineers would go through great lengths to optimize their code. Hardware engineers would work closely with the software folks to develop efficent and useful fast paths. Oft-cursed quality assurance teams would spend months hunting down elusive bugs and areas of poor performance. Physical equipment was both elegent and overengineered.
Today we have copper heatsinks that have undergone more engineering than the typical Formula One racecar.
Nevermind that we have to reinstall Windows every eight months or constatnly watch Bugtraq regardless of our platform.
Hey... ya know what...
I have a water cooled Athlon and aside from the cost and complexity (true.. not fun) it just looks friggin cool...It's completely worth it to me when people see it and stare for that extra 30 seconds...
The best question I here is "What they hell is that?"
And sorry but, I just have to add...
It's gone from suck to blow!
Way back in the dark dawn of the desktop PC age, I was a programmer and service tech for the local importer of Intertec Superbrain computers.
These were a CP/M based machine with two Z80 processors (the second one was dedicated to disk I/O but configured so that the main CPU was placed in a busy-wait loop while the IO occured. Obviously this was a crappy hardware solution to a problem caused by an inability to write decent firmware on Intertec's part.
Anyway -- these machines were originally designed for the US market, so the PSUs were all 110V. Around these parts the mains voltage is 230V so they included a 230V-110V transformer with machines shipped here -- and it was mounted inside the all-encompassing case that also incorporated the screen and keyboard.
Cooling on the machines was by way of a weak fan that exhausted down onto the table beneathe the machine. It was barely adequate for the 110V machine so when the extra heat from the transformer was added to the thermal input -- the machines began to overheat.
The manufacturer was useless -- offering no suggestions and losing all interest in supporting the product.
The solution was pretty simple -- use a bigger fan.
However -- there was a rather unfortunate side-effect. When you turned on the computer, the fan-blast would blow every single piece of paper off your desk. Funny as hell -- the first time.
Although attempts were made with the fan reversed so that it blew up into the machine, a couple of machines expired after a sheet of paper found its way under the case and got sucked up against the fan grill -- effectively stopping all cooling.
Originally the Dreamcast was supposed to be liquid-cooled. We were pretty excited to open up the case and check that out -- no doubt it would involve hundreds of tiny valves and pipes and pumps and very small migrant laborers to work them. However, Sega seems to have engineered the Dreamcast to run without overheating and scrapped the liquid-cooling -- we saw no evidence of it when we poked around.
Instead, heat is distributed out through a large metal plate that acts as both shielding as well as a heat sink. A sizable fan runs when the system is on to circulate air -- it's both effective and a little noisy. We've had no overheating problems with the Dreamcast, even after extended 12-hour or more sessions.
- PlannetDreamcast
I've purchased various super-silent heatsink/fan combos before, but they never remain silent for long. After about a year of leaving my PC on, the fans get louder and louder. I hope someone can find an simple, economical, non-fan alternative.
No disrespect to your boss, but he is wrong. If you want high speed with a high transistor count you are going to pay for it in heat. It is a valid tradeoff. A designer made the decision to disipate heat with a heat sink and fan so that the chip could run faster. Now there is room for improvement. The Intel chips produce less heat then Athlons. The PowerPC chips are, last I checked, much better still. But look at the very power hungry, liquid cooled, and very fast old CRAY machines. Those machines were fast for many reasons, but a big one is the crazy research Cray did in liquid cooling. The Cray engineers knew that if they wanted fast they were going to pay for it in heat. If one of your design criteria is low power/low heat they yes the chip is poorly designed. But in this case the criteria is high speed and the tradeoff is 30-80+ watts of heat.
A neighbor in an apartment building I was in in
1999 built a water-cooling extension for the heat
sink in his PC by soldering small copper tubes
between the fins of the heat sink that was already
there and running tubes from them to a jar of
water that was above the board.
As anyone who has worked with pumpless solar water
heating knows (I myself read an article on this in
popular science in the early eighties) to have
thermal circulation you want the reservoir of cool
water to be above your umm, solar panel.
postercommently diagram available
A passive water cooling system can be built on
the same principles: substitute a heat sink with
tubing attached to it for the solar panel and
there you are.
During periods of intense load you can throw some
ice into your cooling tank -- but then there is a
danger of condensation.
As long as the whole system is at ambient temp.
or above, condensation will not happen.
It is easy to imagine a server farm with cooling
hoses running from each machine to a large central
cooling tank with lots and lots of fins on it
or more aggressive cooling strategies. Once the
heat is out of the enclosure, size considerations
are no longer as important.
Rack mounted automotive radiators, anyone?
There's all those distributed computing projects to choose from.
Ok, so let me rephrase.
Why is spending $80-$150 for the watercooled solution that will give you 100MHz more that you will spend helping other people get 8% more performance out of your computing for distributed computing which will result in a 1.0 * 10^(-12) percent completion of their data set so that you can get a silly performace counter a smidgem higher worth it? =)
Funny actually, I computed that at home it costs me $4CND/mo more to run distributed net than it does to not to (just from electricity). So why am I spending $4 to help someone else win $10000 for cracking something? =)
If God gave us curiosity