No surprise: Two Hygon C86 3185 beat the Ryzen 5 5600X

Every now and then we report on CPU developments from China and with the Kaixian KX-6780A we have already had a closer look at one of them. Youtube channel from china EJ . devices He posted a video about such a platform where two Hygon C86 3185s were installed on the used motherboard.

Hygon processors based on the Zen architecture emerged from a Chinese joint venture with AMD, which has since been discontinued. Processors with four to 32 cores – the latter, of course, are intended for server use. The processors are manufactured in 14nm. The processors are largely identical to the first generation Ryzen and EPYC processors. Rather than being built into a base, it is usually welded.

Hygon C86 3185 is an octa-core processor with a base clock of 2.0GHz and a boost clock of 3.4GHz. As the basis of Zen, a 4 MB L2 cache and a 16 MB L3 cache are used accordingly. Accordingly, one of these processors is slightly slower than the Ryzen 7 1700X, which you can certainly call a little outdated. The RAM is probably another limitation in terms of performance. With DDR4-1866 this was much less than was possible even with the first generation of Ryzen – that is, DDR4-2667. Accordingly, limitations on overall system performance are expected due to slow memory.

When compared to the Ryzen 5 5600X with six cores of Zen 3, two Hygon C86 3185 with a total of 16 cores had a similarly tough time. The current Ryzen model is nearly twice as fast in single-threading performance. After all, dual Chinese processors can beat the Ryzen 5 5600X by about 10% when the performance of all eight cores is required. Keep in mind, six versus 16 cores. Some gaming benchmarks with the GeForce RTX 3080 Ti showed fairly slight differences, but the graphics card was decisive here, especially at higher resolutions.

Once again it becomes clear which development steps modern processors have gone through within a very short period of time. Even if China still relies on outside help for processors, in other areas, such as supercomputing, they have long reached eye level.