Last week, Gamers Nexus, a well-known overseas technology channel, exposed that Intel's 13th generation Core and 14th generation Core desktop processors had process defects, resulting in a variety of system crashes on desktop PCs using these CPUs.
Intel was also quick to respond, recently announcing in the community that it had found the cause of the problem, and said that the fix is expected to be available in mid-August. At the same time, Intel clarified that the cause of the problem was not a process defect, but an abnormal processor voltage, which was beyond the safe range.
CPU instability has been a problem for a long time
As early as the middle of last year, there were some users on hardware forums who reported that the 13th Gen Core high-end processors often crashed in high-load scenarios, such as games.
There are also some netizens who said in the forum that their i9 13900K began to appear frequently after a year and a half of use, and it could not be solved after replacing the memory, and finally it could not be turned on completely, and the detection found that the CPU was broken.
In April of this year, a number of technology media began to report that in the same month, Intel and motherboard manufacturers added Intel Baseline Profile to the BIOS, that is, Intel's benchmark configuration, which is said to solve the game stability problem of 13th and 14th generation Core CPUs.
However, it was found that the processor performance loss was 8% to 10% after using this benchmark configuration, which is equivalent to improving stability by reducing clocks and limiting power consumption.
According to statistics, the fault phenomena caused by this problem include the sudden blue screen or crash restart of the system, the POST cannot be passed, and the partially invalid core needs to be disabled before it can be used.
In July, the impact of processor crashes is getting bigger and bigger, and it is reported that some game companies have also encountered the problem of CPU crashes, and the servers used by game companies have some high-frequency needs, so in addition to traditional server CPUs, many consumer-grade flagship CPUs are also used, such as the 13th and 14th generation Core i9 series. Crashes are said to be particularly severe in servers, where CPU crashes are not avoided despite the extremely conservative voltage and CPU frequency settings on the server motherboard for system stability.
Australia game developer Alderon Games even said that its internal Intel Core 13th and 14th generation processors crash rate as high as 100%.
Regarding the instability of the Core 13th and 14th generation processors, Intel has been updating the report on the official forums, but it has not been completely resolved. For example, in June, Intel said that the problem was in the microcode and provided new source code for users to update, but soon Intel said that updating the source code could not solve the problem, and the reason is still under investigation.
Until recently, Intel said in the community that the immediate cause of the problem was the presence of faulty microcode, causing the CPU to constantly request higher voltage levels.
It is speculated that there is a problem with the ALD process of the chip, which leads to the oxidation of the copper via
According to the exposure of Gamers Nexus, they speculated that the collapse of Intel Core 13th and 14th generation processors was due to a process defect in the wafer manufacturing process, and the ALD (atomic layer deposition) process defect caused the copper vias inside the CPU to oxidize, resulting in high impedance, causing the CPU to fail.
ALD is an advanced thin film preparation technology that deposits thin films layer by layer on the surface of the substrate by introducing different chemical precursors alternately, achieving atomic-level precision control and extremely uniform film thickness.
In logic ICs such as CPUs, the number of transistors represents the performance of the CPU to a certain extent. The thickness and quality of the gate dielectric layer between the gate electrode and the transistor channel have a direct impact on the performance of the transistor, and ALD technology can deposit uniform and consistent thickness of high dielectric constant materials to reduce leakage current and improve transistor performance.
One of the purposes of the ALD process in the chip is also to create a complete protective layer around the copper vias, and if there are problems with the process, such as porosity or cracks, then these defects can become a path for oxygen to penetrate the copper vias, leading to oxidation.
Oxidation of copper vias is a serious problem in semiconductor manufacturing, which can lead to an increase in resistance, which in turn affects current transfer efficiency and the overall performance of the chip.
Interestingly, Intel also mentioned in its recent response that the manufacturing problem of copper through-hole oxidation had occurred in the early days of Core 13th generation processors, but it was resolved as early as 2023, and said that it was not related to the problem of CPU instability.
Indeed, the problem of excessive voltage mentioned by Intel is also more consistent with the performance of processor crashes, which mainly occur on desktop versions. Because laptop processors have a larger power limit and a lower frequency than desktops, it is possible to avoid crashes due to excessive voltage.
Intel is facing a crisis of confidence, and domestic CPUs are also making new progress
Although this incident was not confirmed to be a hardware defect, it may be the worst crisis of confidence in history for Intel.
This presents an opportunity for competitors to improve their reputation. In particular, AMD recently said that it found that the first batch of products did not meet quality expectations in the new product review, and decided to take measures to correct the error through more rigorous testing and screening, and postponed the release time of Ryzen 9000 series CPUs. AMD also emphasized that this is "so that every Ryzen user can enjoy the highest quality experience."
In terms of domestic CPUs, Loongson recently announced that its 3C6000 server CPU tape-out was successful, and the measured performance of the previous generation Loongson 3C6000 processor has doubled, reaching the level of Intel Xeon Silver 4314 processor.
Loongson 3C6000 single chip integrates 16 LA664 cores, 32 threads, and supports dual-channel, four-way, and eight-way direct connection. At the same time, it supports Loongson Coherent Link technology, which can form an on-chip interconnection through Loongson Coherent Link to form a version with 32 cores or more cores, so as to expand the multi-chip interconnection to support multi-server solutions.
Write at the end
Recently, Intel's situation is indeed a belly enemy, earlier suspended the investment in the France R&D and design center, shelved the construction plan of the Italy factory, and the wafer foundry business is not going well, AI chips have to face such a strong opponent as Nvidia, and also face AMD in the CPU field more and more courageous. Coupled with the crisis of confidence generated by the Core 13th/14th generation processors, the road to Intel's strategic transition led by Gelsinger may be extremely difficult.