lukasz.bromirski.net

aviate, navigate, communicate

proszę taktownie zegarować!

during the last 30 years, processor speeds has increased from millions of cycles to billions - multiplied by multi-core and special mechanisms that increase the efficiency of working with ‘boring’ cores.

Pentium 66 processor from 1993 contained 3.2 million transistors, which is anyway quite a value, considering they are packed into a space comparable to that of four dices - and contains one main unit. available today Xeon E5-2699v4 has 22 cores operating at nominal frequency 2.2GHz clock and 7.2 billion transistors. ‘imagine that’!

everything, however, whether we work with one transistor or billions - requires the proper clock. at this scale of integration and operating frequencies, the error limit is unbelievably small. which means - and it should not be a surprise to anyone - that sometimes manufacturers shoot themselves in the foot

it looks like my problems with NAS Synology 1815+ are associated with factory processor defect of Intel. while in the mainstream systems sometimes this type of problem can be worked around (usually) by upgrading the processor’s microcode, in the embedded systems there is no chance. in this particular case, the problem relates to Atom C2000 processors and signals LPC_CLKOUT0 and LPC_CLKOUT1. it looks like the same problem may also apply to some [Cisco] products (http://www.cisco.com/c/en/us/support/web/clock-signal.html) and probably - hundreds of other manufacturers (and millions of products). according to STH portal, looks like, that silence and mystery is simply the result of the NDA that Intel has forced on users of this chipset (or other similar systems). anyway, there are already first signs from other places and from other manufacturers that something ‘can’ go wrong.

cyber security? AI? yes, but well clocked.


Share