The 1,600-Year-Old Capital China Forgot and Hungarian Scholars Still Visit

At the edge of the Mu Us Desert in Jingbian County, Shaanxi, there is a white ruin. It does not sit inside a city like the Xi’an city wall. It does not offer the staged spectacle of the Terracotta Army. It is a mass of rammed-earth walls standing in loess and sand, silent like a chapter deleted from the book and left on the ground.

Most Chinese travelers have never heard of it. The names that fill itineraries are Xi’an’s city wall, the Terracotta Army, the Big Wild Goose Pagoda, Mount Hua, and the Hukou Waterfall. Tongwancheng is not on that list. Yet scholars in Hungary have traveled thousands of miles to see it.

The contrast is a clue. Why would a ruin nearly invisible inside China’s travel imagination draw people from the other end of the Eurasian steppe?

The answer is in the word “only.”

Before Tongwancheng, the Xiongnu were a pastoral political power. Their centers were mobile: the chanyu’s court, the royal camp, the seat that moved with grass and water. Archaeologists have found Xiongnu fortified sites, but many were “private towns” built to hold captives, not true capitals. In 413 CE, Helian Bobo conscripted 100,000 laborers to build Tongwancheng. It was the first time in Xiongnu history, and the only time in the surviving archaeological record, that the Xiongnu built a fixed political heart in the form of an agricultural-style capital. The name means something like “Unify Ten Thousand States.”

So the claim that Tongwancheng is the only surviving Xiongnu capital site carries weight. In the Eurasian steppe, Xiongnu remains are mostly tombs and horse gear. Tongwancheng is the only settled settlement ruin of Xiongnu civilization still standing above ground. It is the skeleton a civilization left behind when it turned toward settlement.

And yet, in the narrative of Chinese architectural history, this site barely has a place.

On Zhihu, someone asked whether China has any architectural remains earlier than the Tang dynasty. Answers worked through Han que stone pillars, ceramic tomb models, stone house models dug from graves, and pre-Qin rammed-earth mounds. No one thought to mention Tongwancheng: a large ruin entirely above ground, with almost no later expansion and almost no restoration, still there after 1,600 years.

The gap comes from taxonomy. The classic narrative of Chinese architectural history centers on timber-frame construction. The earliest surviving timber buildings are Tang dynasty structures such as Nanchan Temple. For anything earlier, the discussion turns to Han que, mingqi tomb models, and tomb images. Tongwancheng is a massive rammed-earth city wall. Architectural history placed it in the “city site” or “military defense” category, outside “architecture.”

So a Xiongnu capital that once stood tall became, in public memory, a pile of earth; in academic classification, a city site; and almost never “Chinese architecture.”

That is where its value lies. Tang and Song timber buildings have been repaired and rebuilt many times, layer upon layer of later information pressed into them. Tongwancheng’s rammed-earth walls were partly demolished in the early Northern Song, when the outer wall was ordered destroyed. After that, the site was left largely as a ruin. It is an early capital skeleton with almost no later expansion or restoration. Its “original flavor” is survival. Time dried it out, and little else touched it.

The height needs a distinction. Historical texts say the city wall was about eight zhang high. That figure refers to the full defensive height at construction, including the wall and parapets, roughly 18 to 25 meters depending on the Wei-Jin measure. The southwest corner tower remnant still visible today is 31.62 meters high. It was a bastion at the corner of the wall, built to strengthen defense. Its survival shows the scale of the original engineering. For comparison, the Forbidden City wall in Beijing is about 9.3 meters high. Even allowing for differences in measure and in the function of a corner tower versus a wall, the textual height and the surviving corner tower make clear that Tongwancheng in the fifth century was an oversized creation, beyond the normal scale of later cities. The records of “steamed-earth construction” and “a cone could not penetrate it” point to a mixture of sand, lime, and clay rammed into a material close to stone. Technology, power, and ambition built this city out of that mixture.

There is another layer to the setting. Yulin, the broader municipality that contains Jingbian, also contains Shimao. Shimao is a prehistoric stone-walled site about 4,000 years old, with an area of more than 4 million square meters, belonging to a complex society of the late Neolithic. It has nothing to do with the Xiongnu. Yet together, Shimao and Tongwancheng form a distant echo on the same land: stone and earth, prehistoric and imperial, two architectural traditions and two civilizational forms. One is a prehistoric center of power built of stone. The other is the last capital of a steppe empire built of rammed earth. Placed together, they make the Loess Plateau look like a cold-storage unit for civilizations.

The coldness and heat around Tongwancheng come down to selective focus in historical narrative. For eyes trained on centers such as Chang’an, Luoyang, and the Forbidden City, a white earthen ruin at the edge of the desert is hard to place in the canonical sequence of “Chinese architecture.” For scholars looking for evidence that the Xiongnu settled, and for people who trace their ancestry to steppe empires, these white walls are the only physical evidence. Their value lies in the final chapter of Xiongnu history and the history of the Eurasian steppe, outside the classic catalog of Chinese architectural history.

Hungarian scholars come for the only built trace of a steppe empire’s settled phase. Tourism is beside the point.

If a traveler passes by, know this: the ruins are a city.