On August 31, a team of AMD-sponsored overclockers cranked a Bulldozer-based AMD FX processor up to an unearthly 8.429GHz, setting a new world record. This week, The Reg sat down with the leader of that processor-torturing team to find out exactly how they did it. "This is an extreme technology sport," Simon Solotko, AMD's …
It wasn't any secret
All you had to do is watch the video where they first cooled with liquid nitrogen to get to -190C or so and then switched to liquid helium to get to -223C.
It was a very neat exercise in overclocking. Kudos to AMD's staff, engineers and Bulldozer architecture. Consumers are the winners in the end.
but want to taunt at top of article. Remember when Intel said the P4 would scale to 10ghz? Ha and AMD or IBM will beat them to it.
In many laboratory facilities, liquid nitrogen - or even helium - is more or less available "on tap". There is only one substance more volatile than liquid helium, and that is liquid helium-3. Using helium-3 it would be possible to get down to 3-ish K. However, only one helium atom in about 8000 is helium-3, and although isotope separation is comparatively easy , the small amount of material available makes liquid helium-3 prohibitively expensive.
 The two isotopes differ so radically in low-temperature physical properties that they become immiscible and separate into distinct layers below 1K.
Not really an expert in cryogenics are you. He3 does boil at 2.7k or so, but you can get down below 1.5k with He4 just by pumping on it. In low pressure liquids evaporate at lower temps. Using the same trick with He3 gets you below 0.3k. So with a simple tweak to their apparatus they could go from 4K to ~ 1K which is a huge reduction in thermal energy. I'd be curious to see how much better they could do at 1K. It may not help much or it may help a lot. Also at 1K Al is likely a superconductor ( Bulk Tc is ~ 1.2K) so if they are using AL in packaging they could have some interesting behavior.
Ye - Moore's law continues...
If some small scale quality cooling could be arranged for consumer devices would that double processing power just like that? Hehe
@ Eddy Ito
Heh. I don't think anyone has ever worked on a bread based cooling system...
Probably because it'd toast the processor.
<- The one with crumbs in the pockets.
Because it costs a lot of bread!
Helium resource is limited - use it wisely guys
And the point of this is??
Helium should be priced x100 to stop wasters like this, party balloons etc.
That's genuinely interesting...
we are going to loose it whether we use it or not
Naughtyhorse: Get back to mom's basement...
...and when you get that brain tumour 20 years from now from playing too much CoD on your overclocked processor let's hope that there's still enough helium left for an MRI scanner to save your life.
Collect your Darwin award as you skulk out.
"Helium should be priced x100"
Doesn't know how price finding works.
Helium is luckily not an exclusively government-managed resource, thus available without papers in triplicate, 20-month waiting period and fingering by the TSA.
Apparently buyers and sellers think that the current price at the current quantity on market is correct. So who is to say that it should be 100 times higher?
Destroy All Monsters: try reading the article in the link
(i) The world price of helium is artificially deflated by an act of the US Congress *forcing* the sell-off of the USA's strategic reserve of helium by 2015, i.e. flooding the market regardless of demand and price;
(ii) We *cannot* make helium, terrestrial reserves result from the radioactive decay of elements in the earth's crust: we have almost exhausted 4.6 billion years of production in about 100 years;
(iii) Helium is essential for all sorts of things that you probably approve of (if only you took the time to find out): MRI scanners, the LHC, keeping deep sea divers safe, making rocket engines viable, geoscience and solar telescopes.
Oh but yeah...you wanna have cheap helium-filled balloons at your overclocking party, silly me!
Cannot make Helium ?
Check nuclear fusion and we can make helium, if we need to!
If only there were some common source for helium! At night!
There's the Solar Wind, the Van Allen Belt feeder thing, and jolly French Reactor Companies. Even so, running the air stills takes a bit of to-do.
Perhaps you should get this team (and the suffering enthusiasts) a nice thermoelectric cooler to work around 70 Kelvin, or less likely encourage the homebrew substitution of nanotubes for copper?
Given the temperatures for liquid nitrogen and liquid helium, I rather doubt that they'd be using a cupped bare hand to avoid spillage!
At least, unless it was his spare hand...
Silly, but not fake.
If you get spattered with liquid nitrogen or helium, the end result will generally be that it quickly turns into a gas before it can get a good thermal contact with your skin. Now, if you were to plunge your hand into a dewar of the stuff, that wouldn't be a good idea. Small splashes though, are likely not dangerous.
Leidenfrost effect protects the skin if you're just splashed with a small amount. The LN boils immediately on contact with your skin created a thin layer of gas that insulates the skin for a short period.
In fact, you can pour liquid nitrogen over your head, if you have a mind to, with no ill effects. The rapid evaporation forms an insulating layer of gas to keep your skin from freezing (for a little while).
I had a bald-headed chemistry professor that put on a well-attended demonstration during exams week every winter. That was one of his less dangerous acts. His "explosions" rattled the windows in the back of the lecture hall.
I remember university science days where people were invited to dip their hand into (and rapidly out of) a dewar of liquid nitrogen. If you do it briefly you'll come to no harm, and the feeling of having your fingers in boiling liquid is weird, to say the least.
It's a very nice article. I wish they'd have determined the highest _stable_ overclock with the liquid nitrogen, say, by running prime95 for an hour with no warnings/errors, although I suspect an hour's worth of liquid nitrogen is expensive!
Well done to AMD!
as expensive as good beer
No, liquid nitrogen is not expensive. I suspect that having done this once, AMD will redesign with a closed loop cooler, and do it repeatably (a bit slower). Anybody (engineers please) know how to build a liquid nitrogen heat pipe ?
This is an ok feat but hardly a great one.
There should be a standard burn in time using prime 95 or similar.
Would also be interested if non-x86 cpu's were involved in a more formal event. Particularly ibm's power chips which already run at clock speeds in excess of the x86 mob.
Not expensive at all
You'd be surprised how cheap liquid nitrogen is. In any good-sized city, it's cheaper than milk. It's the dewar you carry it around in that's a bit spendy, but the liquid itself? Dirt cheap.
For a party a while ago, a friend of mine carried a small (~gallon-sized) dewar into an industrial gas company and offerd to buy some liquid nitrogen to make ice cream with. They laughed at the tiny quantity he was asking for, and just gave it to him. Another friend of mine, who does low-temperature physics in a magnet lab, literally has it on tap.
The energy it takes ...
... to pump all the latent heat out of helium is huge. Moolah, many, much.
Also very tricky stuff to contain -- much more so than other liquid gases. Scary.
Depends on how much you are buying but ~£4/L - this stuff is used all the time with superconducting magnets. Chemistry NMR instruments hold ~~ 100L
Incidently the LHC uses ~120000 kg or ~~ 650000L I believe.
On another point (although I wouldn't recommend trying this) a stream of liquid nitrogen droplets will bounce readily off the skin without harm due to the boiling liquid being insulated by a layer of gas. As the skin cools however then the problems start. We used to use these types of materials without gloves as any finding its way down a glove could easily produce a nasty burn.
"Although the liquid helium–cooled AMD FX hit 8.429GHz, it didn't stay there long. "Seconds," Solotko said. But at that clock rate, the chip churned through tens of billions of cycles – long enough to take a screenshot of the system running CPU-Z, few seconds and a screen grab from cpu z"
Not impressive at all once u relies it was only stable for a few seconds. And what was all his crap about intel CPU's topping out in the 5's ?? I've seen intel CPU's do better than that on water.
Liquid nitrogen is very cheap (few £ for tens of litres) so to give a stable overclock a go would still be dirt cheap from a LN2 point of view. ie once the chip is down to temp, it doesnt take a huge amount of LN2 to keep it there...
Also, I suspect the hand is there in the vid to prevent small splashes ending up on parts of the mobo that don't want cooling and thus also prevent erroneous cooled parts getting condensation on which is another big problem at these temps. The temp differnce between your hand and a drop of LN2 is so high that when it touches it immediately boils off and you barely even notice. More than that and you know all about it as it burns you with prolonged contact....
Pay attention folks
The intent of the OC'ing exercise was not to benchmark the CPU or run Prime 95. It was as other similar events have been - just to see how high the CPU could be overclocked and function. Some CPUs... like Intel have an issue with extreme cold and do not function well. The OC'ing exercise was to demonstrate for those who like extreme OC'ing, that the new Bulldozer architechture based AMD chippies do not have a cold function defect like other chippies.
Perhaps the most noteworthy comment from the AMD folks was that Zambezi chips will easily run over 5 Ghz. on good air cooling.
If that's the case
Why don't they release some ultra high clock speed parts and really challenge the i7?
"Liquid nitrogen turns to ice at -209.86°C"
Only if you're an alchemist.
It turns to solid nitrogen at -209.86C. Thats completely different stuff to ice, which is solid water. H2O you know.
The common usage for ice is any substance that freezes below 0.
Usually ice is a contraction of water ice - as opposed to ammonia or nitrogen ice that are commonly used in astronomy.
Or "Dry Ice"
Which is solid CO2, I think?
Then I guess by your 'common usage' definition, icebergs aren't made out of ice, as salt water doesn't freeze 'til about -20C.
Surely if you an alchemist...
...then it would turn to gold?
Alchemy revealed not to have been anything to do with gold: it was the quest to make early processors run faster by freezing them!
AMD will be releasing fast Bulldozer CPUs
It should be obvious that AMD will be releasing fast Bulldozer based CPUs or they would not have made the comment about 5 GHz. on air.
I was taught at school that liquid helium behaves pretty weirdly - crawling up the sides of containers, etc. Is this why the cyclic flow thing was required ?
I mean, I grant you, I was also taught at school that if the earth stopped spinning we'd all float off to space (albeit in primary 3 - 8yo)
@stu 4 re: Helium weirdness
It has to be cooled to below just liquid, and there are specific requirements for it to work.
Google says below 2.17kelvin for He4 and 0.002491 K for He3 - so the chances of finding it in an open container are quite slim...
I tried this at home kids...
... and managed to play 3 seconds of crysis 2 at almost acceptable frame rates - awesome! ...
I have, however, somehow managed to lose two of my fingers during this extreme overclocking event.
"I have, however, somehow managed to lose two of my fingers"
You should sue the guys who made Crysis 2. Sounds to me that game is too hard on the player's fingers.
Have you counted the kids?
If you're going that cold...
Why not build something out of superconductors, which are viable at those temperatures*.
Should be able to clock much faster..
* That's assuming that building a cpu out of superconductors is as easy as building them out of semiconductors.. buiilding the damned thing isn't my field :p
It's strange that the article says that they couldn't measure the temperature below 77 K. Its not very difficult. There are specialised diodes, thermocouples, or for £10 one can buy a platinum resistance thermometer which will work to below 20 K.
They should ask some one in a University physics department, NIST, or NPL for help
Once upon a time....
... back in about 86?... I was a bonded messenger and ran paperwork and such around Tucson, AZ. One fine day I had a pickup at a medical office and they handed me a thermos with the lid loose - warned me not to tip it over and don't tighten the lid. So, in my infinite young wisdom, I drove to my next two stops with the thermos set between my legs before I delivered the thing. I couldn't help but ask what was in the thermos....
Would you believe STD samples in liquid nitrogen? I freaked. Bastards.
get some more clocks if you cover the cpu and desk in thermal gel and give alot more space for heat to dissipate
Great. Shame the video was so stylised (purple lights, tricksy camera, baseball hats). Makes it hard to take the effort seriously.
its no worse then what thermal gel heatsinks will be like, when they take over from water blocks and pumps, and be standard like fan heatsinks, and able to take 200c for cheap thermal gel
- Mounties always get their man: Heartbleed 'hacker', 19, CUFFED
- Batten down the hatches, Ubuntu 14.04 LTS due in TWO DAYS
- Samsung Galaxy S5 fingerprint scanner hacked in just 4 DAYS
- Feast your PUNY eyes on highest resolution phone display EVER
- AMD demos 'Berlin' Opteron, world's first heterogeneous system architecture server chip