If the exaggerated clock of the CPU is carried out, CPU will be moved by FSB higher than regular FSB.

If it continues raising FSB with an exaggerated clock, the tolerance limit of a circumference circuit may come by CPU of low magnification previously from the tolerance limit of CPU. In order to improve this state, the "Underclock" technology which lowers the frequency of PCI or AGP of operation is this reconstruction.

However, PCI and AGP clock are made to be to synchronize with FSB by the specifications of a chip set. Since it is made to operate, without synchronizing it by reconstruction, the word of "being asynchronous" is used.

Since it is the usage besides regulation of a chip set and it may not operate with the performance of IC on the kind of chip set, OS, and boards, such as for example, a video board, please approve of this asynchronous operation.

Although the Intel BX tip had appeared in the main board of IBM PC/AT compatible machine at the time of that it will convert first and having considered, it is specified that FSB and PCI of BX chip set of Intel are synchronous clocks. Since FSB is support (66MHz and 100MHz) in the case of BX tip, PCI is support of only the 1/2 and 1/3.

However, if the main board at the time of having considered reconstruction is seen, PCI has [ FSB ] one fourth of the products of FSB above 124MHz. When it turned out that this 1/4 operates, I doubted whether IC which moves by 1 / 2 - 1/4 of FSB would move truly synchronizing with FSB. (probably -- one fifth -- also coming out -- even if it thinks that it moves, I think so) Moreover, he doubted also of AGP also from moving by halfway (a phase being reversed to FSB) clock called two thirds at the time of FSB100MHz.

Moreover, it is asynchronous, and when using by FSB100MHz and it considers that the clock of PCI can be raised and I/O access can be carried out early, I want for it to be effective also to the clock rise to which PCI had become the neck, and to have come to do, when moving. Therefore, it will experiment.

At first, it experimented in PCI by 34MHz fixation by very ordinary single CPU main board called P6F91i Scorpion of Freetech. However, although it operated well, since FSB did not operate by 133MHz, it returns.

Next, it experimented by dual CPU main board called P6B40D-A5 of Tekram. Since this board had failure that a clock will stop when contact of the box of the button battery on a board was amusing and turned off the main power supply, although 26000 yen excelled those days, since it came to the head, it plunged into the new main board at a change and reconstruction.

Since then, since every day of infernal reconstruction started, it carried out (smile).