Chinese Chips Are Artificially Slowing Down To Avoid US Export Regulations

In the past, countries protected their domestic industries with import tariffs. This gives the home side a price advantage for companies operating abroad, but the practice has somewhat fallen out of fashion in the past few decades.

Today, governments are more creative, using lavish export controls to protect their interests. To that end, the United States instituted export restrictions on powerful computing devices. In response, Chinese designers try to artificially slow down their hardware to avoid this rule.

Companies like NVIDIA and AMD had to rework certain products to comply with US regulations against Chinese exports. The A100 data center GPU was banned from export, so NVIDIA developed the inferior A800. Credit: NVIDIA Press Site

The new export rules come as the United States government grapples with the rise of China’s military, both in size and technological sophistication. Regulation limit the export of advanced integrated circuits, but the regulations do not stop there. Hardware, software, and other manufacturing equipment needed to design such hardware are also subject to the rules. The often-cited goal is to slow or stop the development of advanced military equipment that the Chinese government could use or sell to other countries. Alternatively, it could be painted as an attempt to protect the profits of existing players in the semiconductor market.

Chips capable of an “aggregate bidirectional transfer rate over all inputs and outputs” exceeding 600 GB / s, not counting volatile memory, cannot be exported or re-exported to China, under the new rules. Advanced manufacturing equipment used for electroplating, chemical vapor deposition, and other chip production processes are also banned from export. Just to cover the whole language, software packages for the design, manufacturing, or use of these chips or associated hardware are also subject to sanctions. Companies can apply for a license to export such materials to China, however, as with many such restrictions, there is a presumption that such a license will be denied. Other limitations apply to chips that exceed the performance limits of machine learning and powerful supercomputers.

In addition, with respect to the export of goods not subject to the above restrictions, “US persons” must have a license if the goods will be used in the “development” or “production” of ICs in China meeting certain criteria. This includes chips that use a non-planar architecture, or are made at a technology point of 14 nm or less, as well as NAND memory with 128 layers or more, and DRAM made at a point of 18 nm or less. The category of “US persons” is also broad, including US citizens, permanent residents, and companies and legal entities established in the US, even when operating abroad.

Chinese tech giant Alibaba and smaller startup Biren Technology have found themselves struggling with restrictions. Along with various lesser-known Chinese chip firms, they have invested heavily in designing new chips for powerful computing applications. These include new chips to rival GPUs from companies like Nvidia and AMD, and processors for machine learning applications.

But Alibaba and Biren are fabless, outsourcing the actual production step. Many of these firms have their designs manufactured by Taiwan Semiconductor Manufacturing (TSMC), considered the world’s leading silicon manufacturer.

Some of the latest designs from this company violate the new export rules, in terms of data rates or other factors. While they are slated for production in Taiwan, US export regulations remain in effect. That’s due to the fact that the majority of semiconductor manufacturers around the world rely on US-made equipment and software. If foreign manufacturers start shipping such designs to China, they will quickly be cut off from the US equipment and software needed for the job. China has been spinning up its own semiconductor production facilities, but they are now several years or decades behind the cutting-edge, so they cannot produce such advanced designs.

Biren Technologies has touted its new datacenter GPUs as outperforming NVIDIA’s A100 offering. Since the latter is not legal for export to China, having a replacement at home is key. Credit: Biren Tech News Site

Biren Technology is at risk of running over the limit and its BR100 GPU, which is intended for machine learning applications. The initial statement quoted a figure of 640 GB / s, exceeding the stated limit. Since then, the company’s website has listed the bandwidth of the card in various figures 512 GB/s To 448 GB/s. According to some researchers, the company could disable part of the BR100 chip to push the limit, while potentially allowing it to be reactivated later.

Alibaba’s own business is facing similar problems. The company has been working on an advanced machine learning chip for AI work at TSMC’s 5 nm technology point. Reportedly, the team is exploring reworking the design to avoid regulatory issues, but this is an expensive exercise that will take months and millions of dollars.

Engineers complain that the rules are not clear, as there are different ways to calculate the two-way transfer rate. Regardless, many have tried to reduce processor speed by rule. The key is to remain low-key, according to a source speaking to Ars Technica. Some companies have released press materials in public for chips with transfer rates that exceed regulations, alerting authorities to monitor shipments of such parts. In cases where the chip’s capabilities are not yet widely known, engineers have more potential to work with fabs to find redesigns that can pass regulations.

Part of a Trend

This is not the first time that US export regulators have tried to take the wings of a Chinese technology company. Huawei’s semiconductor arm, HiSilicon, fell under the previous set of export rules in 2019. The initial sanctions were placed on the company due to backdoors allegedly found in Huawei’s communications equipment. These rules stop Huawei’s access to software and hardware from companies like Intel, Google, and Qualcomm. However, Huawei persevered with its own chips and apps, with Chinese buyers propping up the company’s sales as international business faltered. HiSilicon is quickly becoming the number one supplier of smartphone chipsets in China.

From there, the US government went up the chain, making it illegal to supply HiSilicon with equipment for its semiconductor fabs. That knell of death for the company’s flagship smartphone chipsetand it was quickly overtaken by other companies in the market.

Access to the world’s best silicon factories is key to building high-performance chips. Now, the US holds the key, thanks to a monopoly on the supply of advanced manufacturing equipment and design software. China will persevere on spinning up its own capacity, in much the same way as it has pursued production jet engine itself and other technologies. However, by the same token, it’s a long, slow road to walk, and an expensive one to boot.

Banner image: “Silicon Wafer” by Enrique Jiménez

Leave a Reply

Your email address will not be published. Required fields are marked *