At the recent reporting conference the head of AMD Lisa Su explained that it would be impossible to count on this without the ability to steadily increase the supply of products. It is possible to level out the impact of shortages through long-term supply planning.

At least, this is the wording used by AMD CTO Mark Papermaster in his comments at the Web Summit conference, quoted by Reuters. “Our supply specialists have worked through the issue in such a way that supplies for months and years to come are carefully planned,” the spokesman said. “Everyone has had to pay more attention to the supply chain, but we’ve been doing that since the beginning of the pandemic,” as Mark Papermaster added.

However, he mentioned another factor that adds weight to AMD’s negotiations with suppliers: “We use advanced process technology and are a big buyer in that area, which definitely helps ensure supply stability.” In the automotive segment and consumer electronics production, component supply problems concern products made with more mature process technologies. In this sense, AMD is partially protected from problems for the reason that it uses advanced lithography.

Translated with www.DeepL.com/Translator (free version)

The post AMD avoids significant impact of deficit on business appeared first on ZED.

But now they are reportedly developing SiliFuzz, which will detect defects in central processing units.

The principle behind SiliFuzz is to analyze processor performance by running pre-prepared test data collected through emulators and disassemblers. This is one of the varieties of phasing – the processor is loaded with “random” calculations, the result of which is checked at the output. If there are discrepancies, the processor is considered defective.

The system is designed to detect, first of all, electrical defects in the chips, which could have occurred during manufacturing, installation, operation, etc. Particular attention is paid to them, rather than to logical errors in the CPU itself. In this case, the tests themselves do not use any low-level debugging mechanisms, allowing to use them on “combat” systems.

In fact, the developers’ task is to create a system that could regularly test every core in every Google server with minimal impact on its performance. In its current form, SiliFuzz chooses the moment when the load on a particular machine is not so great, and consistently tests groups of four threads (2 cores with SMT) for no more than two minutes. So far, the developers are focused on x86-64 processors, which are massively used by Google itself.

The main purpose of the project is to automate the detection of hidden defects, which lead to incorrect calculations and are much more dangerous than usual crashes and crashes because a single small deviation in the chip operation leads to the accumulation of an array of errors. For example, it turned out that some CPUs sometimes returned incorrect F2XM1 (x2-1) call results, while others occasionally gave different FCOS calculations from the correct ones. In the latter case, the difference is less than 0.0000003%, but even that can be enough of a problem.

As noted, about 45% of defects found with SiliFuzz are not tracked by other tools. In the future, the developers plan to scale SiliFuzz, increase the speed of the program, as well as improve the overall quality of the work.

The post Google develops SiliFuzz project for mass detection of hidden defects appeared first on ZED.