General Garcia is dead now, but there are other Garcias - Elbert Hubbard
WHAT DO YOU get if you mix 190 proof booze, liquid nitrogen, and overclockers? World records and a lot of smoke.
Booze and Crossfire bridges - how to party
AMD is going for a bunch of overclocking records today, and it designed some interesting hardware to get it there. Of the bunch, by far the coolest things were the clear acrylic tubes filled with alcohol and liquid nitrogen. Why use a standard copper cylinder when you can make one that boils and smokes, and you can mix drinks in it when you are done?
Normally, overclockers use a copper or aluminum-copper cylinder instead of a heatsink on the CPU, and fill it with liquid nitrogen (LN2). The LN2 boils off cooling the CPU to sub-zero temps, and that allows it to run faster. Now, the nutballs are all going to liquid helium, and that requires different types of pots.
One of the side projects to come out of the helium transition was the clear acrylic pot, and it looks great. If you take an acrylic tube, mount it to a copper heat plate, and put that on the CPU, you get a nice tube that almost instantly turns to ice when you pour LN2 or helium in. To prevent this, you put a second tube around that, and fill it with ethanol as a thermal buffer. It looks like this.
Hot pots
The fun begins when you pour LN2 into the middle tube, it starts violently boiling off, and cooling the CPU. The easiest and most common way to get near pure ethanol is go down to the local liquor shop in a bad neighborhood, and buy a bottle of Everclear. Pour it between the tubes, and it prevents the outer tube from icing over. Somewhat. Under full boil, it looks neat.
Booze and boiling
As of this writing, Chew, the overclocker running the rig, had the CPUs down to -146C, and the record runs were just starting up. Once the Helium rig gets down to its running temperature, records are going to be falling, that is for sure.
In the end, what more do you want? Some 6+ GHz Phenom II 955 CPUs, overclocking records, liquid nitrogen, high end graphics cards, and 190 proof booze. That is how geeks really party. µ
I have played with acrylic pipes before and they always shattered on me when I poured a sub zero liquid into them. Maybe they used something stronger, like Polycarbonate instead.
Oh, boy, are they going to be surprised when they start using liquid Helium. Assuming they can get enough, it's a superfluid, with damn near zero viscosity and density, which means it will flow upwards, also very slippery. Not to mention insanely cold.
At some point, the chip carrier and substrate is going to shatter, not to mention the silicon itself.
If the chip gets cold enough, it'll go superconducting on them, which will completely screw things up, as common logic circuits are not designed with superconduction in mind.
Well, you can't get Everclear in NY. At least not legally. The best I have found is Devil Springs at 160 proof (80%). It works for me, but I am just using it to sterilize wine making equipment. I have accidentally breathed MUCH more than I have drunk.
The Mad Winemaker
I would think a modified Dewer flask would be better than Everclear. The vacuum would significantly reduce the contact by the N2 with the warm outside, except for where it contacted the chip.
Why not toss the whole motherboard into the liquid N2?
I could go into why Everclear is only 190 proof and not 200, but it would take some 'splainin' and I don't feel like it. It's something we went over in my undergraduate chemical engineering education.
The nice thing about going to a college with chemical engineering and chemistry students is that you could always get your hands on pure ethanol for use at parties. Awesomely wicked stuff.
"Since 95.6% ethanol and 4.4% water form an azeotrope (meaning that simple distillation cannot remove any of the remaining water), 191-proof spirits are the maximum proof that is available from the distilled beverage industry. As Everclear is a neutral grain spirit, it is relatively low in congeners."
Amazing stuff really. It's a nice demonstration of how chip performance is heat limited, not electronically limited.
Also, it's a testament to AMD's chip fab process that a Phenom can take such thermal abuse. Just imagine what a strain all those bumps and the underfill are under when the whole thing's cooled to -146! Most milSpec components are -40 to +85, so what's the spec of the Phenom?
I bet an Nvidia chip would just fall to pieces given their current track record.
I doubt it, liquid helium has been used in extreme overclocking attempts for a bit now (I think as far back as '06-'07 at least) and AFAIK it hasn't caused any unexpected issues.
Helium 4 only turns superfluid below –271 °C which it is never likely to reach.
I don't think the chip will start superconducting either. Although some semiconducters can superconduct: Diamond and SiC are examples, I don't think anybody has observed superconductivity in silicon [yet].
Chew*
Well, first off there is an enormous difference between liquid helium and superfluid helium.... helium is a liquid below 4.2k and a superfluid below 2.17k, which may not seem like a lot, but it is. Second silicon is not a semiconductor, it can be turned into a superdonductor i believe if it is doped enough, chip silicon no though. Plenty of applications use liquid HE though, without it being HE-II and crawling up the walls and stuff.
There's a very large difference between the two. Normally, when a component is rated with a wide operational range, this is for *continuous use*. The expectation is that the milspec chip can be chucked in a hot tank, in a hot desert, and still work.
Extreme overclocking makes about as much sense as drag racing : it's fun for a short while, but absolutely no use in the real world.
If it keeps people happy, fine, but personally I see it as utterly pointless unless it results in generally usable products or some sort of genuine cost/benefit tradeoff.
Yes silicon is a semiconductor:
http://www16.wolframalpha.com/input/?i=silicon
http://en.wikipedia.org/wiki/List_of_semiconductor_materials
was a typo, meant superconductor. Thanks
Extreme overclocking does have a purpose, right off the bat, it shows the flexibility and limitations of a given CPU design and how much room there is to grow provided the cip makers has success reducing power consumption and heat generation...
No, Dave is right. Extreme overclocking using LN2, Helium or DICE serves no purpose other than bragging rights and marketing for the respective chip companies. Clocks achieved using any of these temporary methods are so far beyond the thermal designs of the chips that they will never be achieved using conventional cooling. Cooling that can be run 24/7, like watercooling gives a more realistic idea of the maximum clocks that can ever be achieved with a silicon respin.