Any bug on a piece of hardware automatically earns a place in the news, not only for its rarity, but also for its potential for damage. In many cases, failures can be corrected or contained through a BIOS and/or firmware update, however, this time it will not be enough. Microservers, routers, firewalls, NAS units, and other similar devices have been dropping like flies in recent months, and all arrows point to a problem in the Atom C2000 processor series that Intel launched in 2013.
The story began with Intel's earnings report for the last quarter of 2016. The Santa Clara giant had a fairly solid financial year, but the so-called Data Center Group recorded a 14 percent decline due to a certain "quality problem" in one of its products. Additional information revealed that the problem manifested itself through failures detected in a specific series of processors, and although the report suggests a "minor correction" in the design, this particular group was forced to create an additional reserve (read: "emergency fund") to deal with the situation.
The affected series is the Intel Atom C2000, which has variants in two, four, and eight cores. These SoCs are used in highly integrated platforms, from microservers and NAS units to firewalls, routers, and other network devices. At first, the failures did not suggest a silicon problem, and the affected companies maintained a strange radio silence when it came to confirming or denying the relationship between the bug and the processors (perhaps due to a confidentiality agreement forced by Intel), but there was one consistent fact: equipment stopped working after about 18 months. At that point, engineers began to suspect possible degradation, which led to failures in the LPC clock bus (Low Pin Count). When the LPC_CLKOUT0 and/or LPC_CLKOUT1 signals (not necessarily both) fail, the system is unable to restart correctly.
In other words, the Atom C2000 are a time bomb. Can it be fixed? Now that the problem has been identified, the answer is obviously yes, but the process is not simple. The Atom C2000 chips are BGA, so any repair will require direct work on the motherboard, or its complete replacement. Aaeon, ASRock, Cisco, Dell, HP, Infortrend, iXsystems, Lanner, NEC, Newisys, Netgate, Netgear, Quanta, Seagate, Sophos, Supermicro, Synology, and ZNYX Networks are some of the companies that use the C2000 in their products. So far, the "damage control" has been disappointing. We hope they implement long-term replacement programs. The affected hardware is hardly cheap, to say the least...
