Comments on: What is wrong with Moore’s Law?

By: The rapture is coming, look busy » Doctor Recommended

The rapture is coming, look busy » Doctor Recommended — Wed, 04 Jun 2008 07:57:01 +0000

[...] where will they be? Can the Future Do Without Economic Logic? Seth Lloydâ€šÃ„Ã´s Million Megahertz CPU What is wrong with Mooreâ€šÃ„Ã´s Law? « Indiana Jones: just download it [...]

By: 10 years from now, where will they be? » Doctor Recommended

10 years from now, where will they be? » Doctor Recommended — Sat, 31 May 2008 17:36:56 +0000

[...] also: What a difference 36 years make GPU versatility Seth Lloydâ€šÃ„Ã´s Million Megahertz CPU What is wrong with Mooreâ€šÃ„Ã´s Law? Specialization, Centralization, and the Future of Chip Integration What do Botnets and GPGPUs have [...]

By: If you build it, they will byte you » Doctor Recommended

If you build it, they will byte you » Doctor Recommended — Mon, 26 May 2008 16:13:18 +0000

[...] the past year or so I’ve argued that a better gauge for measuring processor performance is through metrics like the SPEC suite. Furthermore, I’ve mentioned that watts should also be [...]

By: What a difference 36 years make » Doctor Recommended

What a difference 36 years make » Doctor Recommended — Wed, 12 Dec 2007 18:52:35 +0000

[...] I stated in What is wrong with Moore’s Law, this is a good illustration of engineering innovation in the form of extreme miniaturization. [...]

By: Marshall Sontag

Marshall Sontag — Thu, 01 Nov 2007 17:52:36 +0000

Moore's law wasn't the first, but the FIFTH paradigm to provide exponential growth of computing! Screw Moore's law, and embrace Kurzweil's Law of Accerating Returns!

By: You Will Never Live To See The Day When A One (1) Terabyte Hard Disk Is Sold On Shelves » Doctor Recommended

Wed, 24 Oct 2007 10:57:09 +0000

[...] also: What is wrong with Moore’s Law? « A gusty storm or just a mild shower? [...]

By: more

more — Mon, 15 Oct 2007 07:32:50 +0000

CPU’s frenquency stop groving since Pentium 4. CPU frenquency is by speed of light, becouse CPU lenght is about 0.01 meter, light this length can travel per 33 ps and therefore frenquency can’t be higher than 30 GHz, and becouse chip is quadratic then electrisity must travel to eages chip, and therefore frenquency can’t be higher than 15 GHz. But microprocessors have many conveyers, then making many cycles, and programs who showing you processor speed don’t showing how much cycles was maded in conveyer. So longer conveyer, the lower frenquency show you program of processor. But usualy processor have about 6 or so conveyers and therefore frenquency is about 2.5 GHz. So processors frenquency is limited by chip/microprocessor size. And frenquency limit is reached from Pentium 4 and now chipmakers only play with conveyers length and to get more speed from bigger frenquency with the same crystal size is imposible.

By: What do Botnets and GPGPUs have in common? » Doctor Recommended

What do Botnets and GPGPUs have in common? » Doctor Recommended — Sat, 15 Sep 2007 08:23:14 +0000

[...] also: What’s wrong with Moore’s Law? FLOPS, MIPS, Watts and the Human Brain Seth Lloydâ€šÃ„Ã´s Million Megahertz CPU Specialization, [...]

By: Alex Temal

Alex Temal — Sun, 09 Sep 2007 13:18:12 +0000

This got me to thinking… how do FLOPS compare to transitor counts in the real world? Well, I don’t have any charts on that, but I did find one interesting piece of information while looking for such a chart. Wikipedia’s FLOPS article has the following stats :

* 1997: about US$30,000 per GFLOPS; with two 16-Pentium-Proâ€šÃ„Ã¬processor Beowulf cluster computers, [7]
* 2000, April: $1,000 per GFLOPS, Bunyip, Australian National University. First sub-US$1/MFlop. Gordon Bell Prize 2000.
* 2000, May: $640 per GFLOPS, KLAT2, University of Kentucky
* 2003, August: $82 per GFLOPS, KASY0, University of Kentucky
* 2006, February: about $1 per GFLOPS in ATI PC add-in graphics card (X1900 architecture) – these figures are disputed as they refer to highly parallelized GPU power (see above)
* 2007, March: about $0.42 per GFLOPS in Ambric AM2045 [8]

It seems to me that that beats Moore’s law, whereas the transistor counts shown on the Moore’s law article page seems to follow Moore’s law closely. Moore’s law says that transistor counts double every two years. Well, the cost per useful unit of work that each transistor does is obviously decreasing faster than half per two years. In 2003, the cost was 2% of the cost in 2000. In 2006 the cost was 1% of the cost in 2003, and obviously a few hundredths of a percentage of the cost in 2000.