Global network computers

Abstract

Embodiments useful for a network of computers are presented. In an embodiment, a microchip includes a plurality of dies. Each die is made by a separate fabrication process and assembled into a package with the separate die sections connected directly.

Description

[0001] This application receives the benefit of provisional application 60 / 418,177, filed Oct. 15, 2002. This application is a continuation-in-part of International Application No. PCT / US02 / 29227, filed Sep. 16, 2002, which in turn claims the benefit of U.S. Provisional Application Nos. 60 / 322,474, filed Sep. 17, 2001, and 60 / 323,701, filed Sep. 21, 2001. This application is also a continuation-in-part of U.S. patent application Ser. No. 09 / 935,779, filed Aug. 24, 2001, which receives the benefit of priority from provisional applications 60 / 308,826, filed Aug. 1, 2001, and 60 / 227,660, filed Aug. 25, 2000. U.S. patent application Ser. No. 09 / 935,779 is a continuation-in-part of U.S. patent application Ser. No. 09 / 571,558, filed May 16, 2000, which receives the benefit of priority from provisional applications 60 / 134,552, filed May 17, 1999, 60 / 135,851, filed May 24, 1999, 60 / 136,759, filed May 28, 1999, and 60 / 135,852, filed May 24, 1999. U.S. patent application Ser. No. 09 / 935,779 i...

AI Technical Summary

Benefits of technology

[0022] However, the confluence of all three of the established major trends summarized above--supercomputer-like personal computers, the spread of parallel processing using personal computer general purpose microprocessors (particularly massively parallel processing), and the enormous increase in network communications bandwidth—enables a solution to the excessive idleness problem of personal computers and the possible end of Moore's Law. The solution may achieve very high potential economic savings once the basic infrastructure connecting personal computers with optical fiber is in place in the relatively near future.

[0023] The solution is to use those mostly idle PC's (or their equivalents or successors) to build a parallel or massively parallel processing computer or computers utilizing a very large network, like the Internet or, more specifically, like the World Wide Web (WWW), or their equivalents or eventual successors like the Grid or MetaInternet (and including Internet II and the Next Generation Internet, which are under development now and which will utilize much broader bandwidth and will coexist with the Internet, the structure of which is in ever constant hardware and software upgrade and including the SuperInternet based on essentially all optical fiber transmission) with extremely broad bandwidth connections and virtually unlimited data transmission speed.

[0025] This makes possible efficient external parallel processing (and multitasking), including massively parallel processing, in a manner paralleling more conventional internal parallel processing, called superscalar processing.

[0029] The metacomputing hardware and software means of the Grid (or MetaInternet) provides performance increases that are likely to at least double every eighteen months based on the doubling of personal computers shared in a typical parallel processing operation by a standard PC user, starting first with at least 2 PC's, then about 4, about 8, about 16, about 32, about 64, about 128, about 256, and about 512, for example. After about fifteen years, for example, it is anticipated that each standard PC user will likely be able to use a maximum of about 1,024 personal computers for parallel processing or any other shared computing use, while generally using for free the Internet or its successors, like the Grid (or MetaInternet). At the other end of the performance spectrum, supercomputers experience a similar performance increase generally, but ultimately the performance increase is limited primarily by the cost of adding network linkages to available PC's, so there is definite potential for a huge leap in supercomputer performance.

[0030] Network computer systems as described above offer almost limitless flexibility due to the abundant supply of heretofore idle connected microprocessors. This advantage allows “tightly coupled” computing problems, which normally are difficult to process in parallel, to be solved without knowing in advance how many processors are available (as is now necessary in relatively massively parallel processing), what they are, and their connection characteristics. A minimum number of equivalent processors (with equivalent other specifications) are easily found nearby in a massive network like the Internet and assigned within the network from those multitudes available nearby. Moreover, the number of microprocessors used are almost completely flexible, depending on the complexity of the problem, and limited only by cost. The existing problem of time delay is solved largely by the widespread introduction of broad bandwidth connections between computers processing in parallel.

Problems solved by technology

Despite these tremendous improvements anticipated in the future, a typical PC is already so fast that its microprocessor is essentially idle during most of the time the PC is in actual use, and the operating time itself is but a small fraction of those days the PC is even in use at all.

Nearly all PC's are essentially idle during roughly all of their useful life.

Because the reliability of PC's is so exceptionally high now, with the mean time to failure of all components typically several hundred thousand hours or more, the huge idle time of PC's represents a total loss; given the high capital and operating costs of PC's, the economic loss is very high.

PC idle time does not in effect store a PC, saving it for future use, since the principle limiting factor to continued use of today's PC's is obsolescence, not equipment failure resulting from use.

Moreover, there is continuing concern that Moore's Law, which holds that the constant miniaturization of circuits results in a doubling of computing power every 18 months, cannot continue to hold true much longer.

Indeed, Moore's Law may now be nearing its limits for silicon-based devices, perhaps by as early as 2010.

Images