Recent content by prosty_mirek

CNS was rushed. Centaur states before that because of schedule, they don't implement SMT and uop cache. Indeed it had frequency, power and temperature issues, but not necessarily because of architecture, but from in silicon implementation. In 2019 it was only engineering sample. On that time VIA...

I have found something interesting. From "Shanghai Lianhe Investment Co., Ltd. 2023 Public Issuance of Technology Innovation Corporate Bonds (Second Series) Prospectus for Professional Institutional Investors" (PDF file, translated by ChatGPT): KX-6000 is HX002. KH-40000 is HX003. KX-6000g is...

Zhaoxin is not on Entity List. And what if it will be? All that AMD-Thatic-Hygon JV is on it now, and AMD is still able to produce Hygon CPUs. Despite sanctions, they modified zen1 to 'zen2 level' (Hygon no.4 core) and Hygon filed some patents. If Zhaoxin really uses 7/6nm, then only to bump...

Your description is basically a copy-paste from Yongfeng:

So that Schrodinger's CPU is zx-g based now? Or they changed arch behind zx-f, once again? Because previous zx-f leaked results (from 2020?) was definitly CNS. Yongfeng is already different than CHA. It is missleading. Kx-7000 is few years late now. There are public traces of kx-8000, not kx-7...

Looks like Kx-7000 was skipped.

So Zhaoxin skipped CNS, AI cores, that CHA concept (though owning patents), and all that post 16nm nodes, to sneak peek regression on CPU+GPU integration. SMIC 7-ish nm needs small chips to better yields? That kx-7000 GPU looks big to cores. Zx-1080 is nice improvement from c-960 on...

My bad. I was mean exactly that. Zhaoxin owns patents that Henry invented. https://patents.google.com/?q=Neural&assignee=Zhaoxin&language=ENGLISH&num=100 Rumor has it, that Ncore was very difficult to program.

Process, clock and IPC of KH-40000 are sameish as CNS. From patches we know that KH-40000 and Yongfeng, are the same model and have ID of Centaur. Looks like Zhaoxin cutted Ncore (AI), and pack more L1 cache per core and 2x cores per socket.

According to GeekBench entry KX-6000G is LuJiaZui. Looks like it have TDP issue - @3,4ghz worked 3158mhz max, and no model, so propably ES. Same performance as KX-6640MA @<3ghz. It have new GPU ZX C1080. https://browser.geekbench.com/v5/cpu/compare/7767362?baseline=15619785 Some new Linux...

First signs of KH-40000. Same model (0x5b) as YongFeng. https://github.com/Kitware/CMake/blob/e5f5e05981b9c2db2a9e73b9308267de179ea4d4/Source/kwsys/SystemInformation.cxx Uses ZEN's software optimisations for fixed AVX2 implementation. Also first sign of family 8, if I read this correct...

Zhaoxin have it's own CNS sample since 2019. Calls it family 6. Maybe because of ES/unchanged core, or it is a software/GB bug. https://browser.geekbench.com/v5/cpu/526995.gb5 Zhaoxin's CNS clock was lower (2ghz) than Centaur's (desired 2,5ghz), which was already lower clocked than ZX-E, and...

New Zhaoxin's CPU architecture name is "Yongfeng". It is family 7 model 91 (0x5b). https://github.com/google/cpu_features/pull/218/files CNS is family 6. Lujiazui/kx-6000 is family 7. WuDaoKou/kx-5000 is family 7. Zx-c/isaiah is family 6. So, family 6 = small cores, family 7 = big cores?