Because, unlike carbon nanotubes, they are commercially available, particularly the diamond composites, whereas carbon nanotubes are still very much in research. 8-ball
Probably cheaper than dimond and dimond composites CuSil is a very good conducter. 8-ball Was being sarcastic. I wouldnt classify dimond as redily available. CuSil is easyer to get than dimond and substancially cheaper
CuSil doesn't work. It has been stated that CuSil's thermal conductivity is lower than that of copper.
Agreed. There was a lot of hype about it, but the manufacturer admitted that the data they were basing their claims on originated from very old research with several flaws. 8-ball
were could i buy 1 *walks up to the company in taiwan* *askes the secterary if i could buy one* *ok that will be 5.3 million dollers, cash or credit* *credit! *
Hmm.. I don`t think this processor-technology is for computers after all.. But it could be in the near future http://www.opentechsupport.net/forums/archive/topic/16976-1.html ...but how do we cool a AMD 81000+ ?
That'll be for a single device such as a FET, or more likely, a BJT. That speed isn't really impressive though. Consider this, InP HBTs (same devices as the transistors in the RF stage of your mobile phone) have shown that they can readily get to Ft values of 200GHz plus (that's being very conservative). InP is a III-V material that has incredible high frequency and power characteristics. However, III-V materials have never been used for processors and never will be. Currently, processors like the P4 are made on 12" Silicon wafers, which probably cost around $70/wafer. I don't know how large the die size on a P4 is, but it's probably a couple of cm2, so you can probably get a couple of hundred die from it (rough guestimate). The largest wafer size you can get with InP is 4" and it cost at least $500/substrate before you do anything with it, plus defects are much higher than in silicon. Reading that link, I'm imagining that this diamond material will have similar problems (defect problems is one of the issues they state). I imagine that the diamond material is also prohibitive due to price like InP, GaN and other novel materials. More than likely, this material is being touted as an alternative to GaInP, InP, and SiGe RF devices that are used in telecommunications. If you notice, the company who developed it are Nippon Telegraph and Telephone Corp. Anyway, another thing holding processor speeds back at the moment is interconnect technology, beside the speed of the devices themselves. I have heard that silicon will be with us until at least 2010, and more than likely until 2015 as the leading edge technology. After that, who knows. Some say self-aligning technologies at the molecular level such as proteins. Imagine a Bio computer . What would you do? Buy some seed, plant it, feed it and watch it grow? Or perhaps buy an egg? What if you went on holiday and forgot to water or feed it? Maybe the watercooling of the future will involve getting out your watering can? . Or then again, maybe I'm being daft.
Diamond for sure. It's an amazing heat conductor, it looks cool, and it could be made for a low price if synthetically made. Here's the reason. Find a diamond piece of jewlery. put it in a cold location until you know the 2 materials (air and whatever) have made thermal equilbrium. Then, take the jewlery out, and touch the diamond lightly, as to not give it too much heat, and then the metal. The diamond will return to the ambient temperature much faster than the metal, which is a good thing. Edit: The speed of processors could also be increased by using light instead of electricity. Some company in Israel tried it. You wouldn't use diamonds though, because light travels through diamonds very slowly (becuase of their density), which is why diamonds bend the light so well! Plastic could be used, although it might melt under the temperatures inside a CPU die!
yar! diamond it is. silver too! surprised no one has made mention of quantum computing, that does away with the speed race as right now the fastest quantum computer runs at 7mhz but has the processing power of a computer many factors faster.. to make a long story short, instead of the standard computers that do 1 instruction very fast; quantum computers get around that by doing many instructions at once (just slower). then again you have to control them with radio frequencies and so far they been used to calculate the factors of 12..
Quantum computing is a long way off. Besides, the way it works is ideally suited to some things, but not others. The first applications will be in the military for encryption and so on. Silicon will be with us for at leats 10 years yet. 8-ball
well.. heres my possibly slightly dumb suggestion. why the hell not just grow a bloody gr8 silicon crystal on the back of the die? u can then cut it up into the 'comb' shape, and fashion a block to sit around it. silicone sealent around the bottom, and woo hoo! no more 'die-heatsink' interface, no more AS5, no more having to re-seat ur heatsink every 2 months cos the paste has dried up and gone bad.
Think about it this way. Processors come from single crystal silicon boules. This is then sliced up into many wafers, polished, and identified according to the orientation of the crystal within the wafer. A LOT of the cost of a processor is determined by how much silicon it takes up. A large silicon boule can be worth several million dollars. You WANT to get as many processors out of it as possible. Besides, the conductivity of silicon is only ~150W/m/K, roughly one third that of silver. 8-ball
Lets see.. According to my chem. teacher the Cp (specific heat in calories) of a element is equal to 6.2/X when X is the atomic mass of the element. Platinum .03178 cal/C* Copper .09756 cal/C* Silver .05747 cal/C* Basicly Platinum is the best material as it takes less calories to raise it one C*. Also the heavyer the element, the better it can absorb. Maybe you could pump mercury though your water system .
8-ball... u make a very valid point there, without knowing it. Silicons conductivity is CRAP! there is very little point putting a darn good heat-transfer material on there (diamond, gold, a wormhole into deep-space ) when the transfer thru the die its-self is so poor. Were fighting a loosing battle untill they decide to make CPUs outa something else, that DOES conduct well. A chain is only as strong as its weakest link, and u can only pull out 300 watts of heat if the material ur trying to pull it out of lets you!
Rekarp, the specific heat capacity of a material does not determine the equilibrium temperature difference. This depends on the materials thermal conductivity. The specific heat capacity only determines how quickly the equilibrium temperature delta is reached. 8-ball
small correction m8, a silicon wafer costs less than 100 pounds, thats for the finished polished wafer with an oxide layer ready to go into the manufacturing process, the major cost is in the capital equipment required to produce IC's( between 1 - 5 billion dollars to equip a fab depending on critical dimension required) & profit margins. The first Pentium CPU's cost 19 dollars to make, current ones around 50 max. Thats how Intel makes approx 5-6 billion dollars a year profit, large margins on large volumes. They have a huge amount of manufacturing capacity.
Diamond is used for heatsinks. it is used industrially to cool high power lasers and stuff. CVD (synthetic) diamond costs about £1 per 5 grams (£1/carrot) and thats when ordered in bulk, and very small particle sizes. As you can imagine, larger particals, would cost more as they are harder to produce/more rare (for natural diamond). Needless to say, unless your fkn rich, any talk of this is utterly pointless. I dont care how cool you think it would look - anyway, if you were crazy enough, you would probably only use a diamond insert the same size of the die, much bigger would be getting a bit pointless.. ... So you wouldnt even see it. Ofcourse a more cost effective way - you could use diamond and silver together (or copper if your running out of cash... ). using a grit/mesh size of 25/inch, about 50 grams of diamond, with about 200 grams of silver, could work quite well. Mix silver powder, and the diamond, put it in a press at about 1.5 tons. next put it in a hot press, about 800°c at 200 bar pressure (this would also shape it how you want it.. channels or cups etc...) for 20 mins. Then a copper or brass block could be wielded around this insert to make the base. materials would be about £125 and Id assume thermal conductivity would be better than pure silver, if only slightly, but how much I wouldnt know... Thats all a guess btw. I have no idea how thermal conductivity is affected from pressing. I would assume that powdered silver, pressed into solid, even under high pressure and heat, would not be as good as solid pure silver (ofcourse assuming the pressing doesnt get hot enough to melt it... which would effectivly make it the same). Btw my gpu block is made from pressed silver powder, no diamonds though. the silver was 99.5% pure, in its powdered form.
