Loongson’s 3A5000 is the most promising domestic Chinese CPU we’ve seen so far. Compared to the Zhaoxin KX-6640MA and Phytium D2000, Loongson’s 3A5000 is a wide core with a better balanced backend and a better cache hierarchy. But it suffers the same fundamental issues as the other two in its quest to be a general purpose CPU. Loongson’s LA464 simply cannot deliver performance in the same class as any recent Intel or AMD architecture. Compared to its western counterparts, LA464’s architecture is smaller, the L2 and L3 caches are worse, and the DDR4 memory controller is embarrassingly bad. Even though Loongson has gotten their cores up from 1 GHz to 2.5 GHz, no one runs desktop or even laptop CPUs at clocks that low. Because of its massive clock speed deficiency, Loongson can’t even get in to the same performance ballpark as recent desktop CPUs. It even struggles against Neoverse N1 running at 3 GHz.
This is a far more detailed looking at these processors than we posted a few days ago.
I think in the short term if they can produce these in volume using domestic fabs, like SMIC (which can produce sub-10nm chips), this will fill a gap for at least government use. The performance will be about the same as an 8 year old Intel processor or something, but serviceable to run basic Office Suite (ie libreoffice) software and light web browsing. The question is how quickly can they iterate and, without access to EUV to get node size below ~7nm, what are the performance limitations they have. It’s been very interesting to watch. The NSA must be having a field day with these looking for exploits.
1 – https://www.tomshardware.com/news/smic-mass-produces-14nm-nodes-advances-to-5nm-7nm
As bad as the 3A5000 may seem, it’s still lightyears ahead of anything we europeans have to offer…