Processador CPUs Testing a Chinese x86 CPU: A Deep Dive into Zen-based Hygon Dhyana Processors

[Dark Kaeser]

Ultimately we ended up with two systems to test – one commercial ‘Hygon Dhyana’ and one enterprise ‘Hygon Dhyana Plus’.

By and large, as we could determine, the core layout is identical, with the same cache sizes, TLB sizes, and port allocations – there were no differences at this fundamental level. The CPU still offered 64 KB 4-way for the L1 instruction cache, 32 KB 8-way for the L1 data cache, 512 KB 8-way for the L2 cache, and 8 MB 16-way for the L3 cache, identical to the Zen 1 core.

Para além da óbvia diferença a nível dos algoritmos de criptografia usados no CPU, também há diferença a nível do número de instruções (dependendo das instruções) por cada ciclo, o que não deixa de ser surpreendente...

Cryptography Changes
For the cryptography changes, these are detailed in the Linux kernel updates. The updates revolve around AMD’s secure encryption for virtualization features, or SEV. Normally with an EPYC processor, SEV is governed by the cryptography protocols defined by AMD, in this case RSA, ECDSA, ECDH, SHA, and AES. In order to generate the right keys, SEV uses these methods. However, in the Hygon Dhyana designs, SEV is built to use algorithms known as SM2, SM3, and SM4.

Slowing Down Some Instructions

In our testing, we found that while integer performance is similar between Hygon and EPYC, certain floating point instructions, namely DIV and SQRT, are not pipelined in the Hygon CPU. This means throughput and latency is reduced. A lot of simple MMX/SSE instructions have reduced throughput:

Perhaps the biggest change however was one that even differed between the server ‘Dhyana Plus’ processor and the consumer ‘Dhyana’ version. Random number generation, which is a key backbone in a lot of stochastic and financial processes, is severely reduced on the Dhyana Plus. The key instructions, RDRAND and RDSEED, have various reasons for being slow/fast. Here’s the comparison:

https://www.anandtech.com/show/15493/hygon-dhyana-reviewed-chinese-x86-cpus-amd

 

[Dark Kaeser]

Há dias alguém descortinou umas Instinct Mi50 "a.k.a. Vega 7nm" na versão mezzanine, que supostamente a AMD nunca produziu, e com o branding da Sugon

Chinese Server Maker Sugon Has Its Own Radeon Instinct MI50 Compute Cards

When AMD launched its Radeon Instinct MI50 and MI60 compute accelerators based on its 7nm Vega GPU back in late 2018, the company only introduced them in add-in-board form-factors. But as a post on a Chinese forum by YuuKi_AnS shows, there are (or at least were) mezzanine-style Radeon Instinct MI50 cards for HPC applications.

The boards in question are said to be samples of the so-called Hygon DCU that is based on the Radeon VII chip enhanced with HPC-specific instructions and made using TSMC’s first-generation 7 nm (N7) process technology. The GPU — which is marked as a ‘Pre-Wukong Sample’ — reportedly carries 16 GB of HBM2 memory, just like the Radeon Instinct MI50 in a card form-factor.

Considering the fact that we are dealing with something that is marked as a ‘sample’, it is likely that the mezzanine board is indeed a pre-production device made in 2018 for evaluation.

https://www.tomshardware.com/news/c...as-its-own-radeon-instinct-mi50-compute-cards

 

[Dark Kaeser]

Aparentemente a China iria ter um 3º sistema exascale baseado num Hygon, mas este limitou-se a um licenciamento do Zen que não teve continuidade, e entretanto o plano original terá sido cancelado

The intended processor was supposed to be a new version of Sugon’s Hygon CPU, said Kahaner, but due to restrictions imposed by the U.S. government, it is no longer clear what computing platform will be used. Hygon Dhyana was a China-made AMD x86 CPU, licensed through AMD’s joint venture with THATIC, a Chinese holding company that was added to the U.S. Entity List in 2019 (see related coverage). At ISC21, it was reported (by Depei Qian of Beihang University) that the Sugon exascale system in Shenzhen would use a new version of the Hygon CPU connected to a Hygon DCU accelerator (“like a GPU”), cooled using Sugon’s custom liquid immersion technology.

In addition to the exascale systems, there are a number of large multi-hundred petaflops systems that are either already in production or are coming online very soon, said Kahaner. According to multiple trusted sources that HPCwire spoke with, two such systems were provisionally submitted to the Top500 in 2019, and benchmarked at ~260-petaflops and ~315-petaflops Linpack, but they were withdrawn from the list before it was published.

None of these systems – neither exascale nor so-called pre-exascale – have made it onto the Top500 list. And while China has its own version of the Top500, called the Top100 (released Nov. 12, the Friday before SC21), none were to be found on that list either.

https://www.hpcwire.com/2021/11/24/...iled-at-sc21-two-operational-and-one-delayed/

Mas pelos vistos há pelo menos 3 sistemas da Sugon para o National Supercomputer Centers (NSCCs) da China.

 

[Nemesis11]

Review de uma Board com 2 Hygons C86 3185:

Ligar a tradução automática do youtube.

São basicamente 2 Ryzens Zen1, com 8 cores @ 2 Ghz Base / 3.4 Ghz Turbo com um TDP de 70W. RAM é DDR4-1866.
Como são 2, ao todo tem-se ali 16 Cores.

Comparando o benchmarks com o 5600X:

Processor	Cinebench R20 Single-Core	Cinebench R20 Multi-Core	Cinebench R23 Single-Core	Cinebench R23 Multi-Core	Blender (BMW Scene)	x264 HD Benchmark	PCMark 10
Ryzen 5 5600X	598	4,536	1,536	11,717	3:33.06	60.5	12,089
C86 3185 x 2	304	5,065	655	13,214	2:44.65	40.5	7,618

 

[Dark Kaeser]

https://twitter.com/i/web/status/1503753921168293889

https://twitter.com/i/web/status/1497780270593810432

 

[Dark Kaeser]

Nos últimos dias tem havido uma variedade de Hygons a aparecer em benchmarks, incluindo modelos sem HT

https://twitter.com/i/web/status/1503312088403529728

https://twitter.com/i/web/status/1503313581558968327

 

[Dark Kaeser]

Tenho de separar os posts pois o fórum só permite 2 posts de media

https://twitter.com/i/web/status/1503318539050098689

https://twitter.com/i/web/status/1503316051911749637

 

[Dark Kaeser]

Não sei, mas há aqui qualquer coisa de errado neste resultado em single...

https://twitter.com/i/web/status/1669624427175108610

 

[ímpar]

China’s Loongson 3A6000 CPU Reaches Core i5-14600K Performance In Clock-To-Clock Test, Overclocked To 3 GHz​

https://wccftech.com/china-loongson...nce-clock-to-clock-test-overclocked-to-3-ghz/