Zhaoxin may not produce the best CPUs for gaming, but the leading Chinese fabless semiconductor enterprise is undoubtedly preparing to unleash a highly impressive server chip. Zhaoxin has unveiled its next-generation Kaisheng KH-50000 processors, which the company describes in its press release as "presenting a 'technological gift' on the eve of the 76th anniversary of the founding of the People's Republic of China."

The KH-50000 utilizes Zhaoxin's latest Century Avenue architecture, named after a famous road in Shanghai. The company is fond of naming its architectures after famous locations within Shanghai because, after all, Zhaoxin is a joint venture between VIA Technologies and the Shanghai government. Century Avenue is the current architecture used by the company for its mainstream KaiXian KX-7000 processors; consequently, it is logical for Zhaoxin to align its latest server processors accordingly. Although Century Avenue is an internally developed architecture by Zhaoxin, many speculate that Century Avenue derives from Centaur Technology's CNS core, prior to the company's split from VIA Technologies in 2021.

Zhaoxin utilizes a chiplet design for the KH-50000, similar to AMD's Ryzen and EPYC processors, with a greater emphasis on the latter, given the high number of cores. A chiplet design would enable Zhaoxin to push the core boundary on the KH-50000, effectively matching AMD's EPYC 9004 (codenamed Genoa) series that tops out at 96 cores. Zhaoxin has planned two variants of the KH-50000: the flagship 96-core SKU and a more affordable 72-core SKU, both of which lack simultaneous multithreading (SMT). The KH-50000 represents a monumental leap forward for Zhaoxin, as it provides 3X more cores than the company's existing KH-40000.

Zhaoxin's photograph of the KH-50000 reveals that the chipset layout exhibits minor differences from that of AMD; the core design remains consistent, however. The gargantuan I/O die is centrally positioned on the processor, encircled by four clusters of compute dies. Each cluster contains three compute dies, totaling twelve. Each die incorporates eight cores and 32MB of L3 cache. When assembled, the resulting processor comprises a 96-core configuration with 384MB of L3 cache.

Zhaoxin Kaisheng KH-50000 Specifications

The clock speeds on the KH-50000 aren't too shabby and fall in line with what you'd expect from a server chip. The 96-core variant has a 2.2 GHz base clock and 3.0 GHz boost clock. Since the 72-core chip has fewer cores, Zhaoxin could push the base clock to 2.6 GHz but maintained the same boost clock.

Although the company has taken the wraps off the KH-50000, it didn't reveal the TDP or other power metrics for the upcoming server chip. The thing with a chiplet design is that Zhaoxin can effectively utilize older process nodes for the KH-50000. Sanctions don't hurt as much if you don't care about power consumption.

For comparison, AMD has historically kept its top EPYC chips around the 300-350W range, and that's with SMT. Nonetheless, the chipmaker has recently pushed the power envelope up to 500W, which is understandable when its EPYC processors are maxing out at 192 cores.

In addition to core count, the KH-50000 advances the development of Chinese server processors. It now supports up to 12 channels of DDR5-5200 RAM, allowing for a maximum of 3TB of memory, in contrast to the 2TB supported by DDR4-3200 on the KH-40000. Zhaoxin has added Compute Express Link (CXL) interconnect support. Furthermore, the expansion capabilities of the KH-50000 have enjoyed an upgrade to include 128 PCIe 5.0 lanes and 16 PCIe 4.0 lanes, compared to the 128 PCIe 3.0 lanes available on the KH-40000.

The SATA and USB ports experienced a slight decrease in numbers when comparing the KH-50000 to the KH-40000. However, Zhaoxin has upgraded the latter to support the latest USB 3.2 Gen 2 specification.

Zhaoxin, one of China's top fabless semiconductor companies, has announced the company's next-generation KaiXian KX-7000N and Kaisheng KH-50000 processors at the World Artificial Intelligence Conference (WAIC) 2025. The former is a new consumer chip for AI PCs, while the latter is a server chip with specifications that rival AMD's EPYC portfolio.

The Kaisheng KH-50000, the successor to the Kaisheng KH-40000, will feature up to 96 cores, aligning with AMD's EPYC 9004 (codenamed Genoa) series. The forthcoming server processors provide three times the number of cores compared to Zhaoxin's current models. Nevertheless, the company has not disclosed whether the Kaisheng KH-50000 will incorporate simultaneous multithreading (SMT) or continue to operate without SMT, as is the case with the Kaisheng KH-40000.

While the Kaisheng KH-40000 employs Zhaoxin's Yongfeng architecture, the company has not disclosed which architecture is utilized for the Kaisheng KH-50000. However, considering the significant upgrades in specifications and features, it is reasonable to assume that the Kaisheng KH-50000 adopts a new architecture. Zhaoxin just didn't want to reveal which one yet.

Zhaoxin Kaisheng KH-50000 Processor Specifications

In terms of clock speed, Zhaoxin may achieve a range between 2.2 GHz and 3.0 GHz for the Kaisheng KH-50000. This may be higher or lower than the Kaisheng KH-40000, which operates at 2.5 GHz. The increase in core count is also associated with an expansion of the L3 cache. Zhaoxin has officially confirmed that the Kaisheng KH-50000 is equipped with 384MB of L3 cache, a capacity that is six times greater than that of the Kaisheng KH-40000.

You may be aware that AMD's Genoa processors include 384MB of L3 cache. Indeed, the Kaisheng KH-50000 could potentially be an identical counterpart to Genoa. Similar to Genoa, the Kaisheng KH-50000 provides 128 PCIe 5.0 lanes and support for 12-channel DDR5 ECC memory, in addition to embracing the Compute Express Link (CXL) interconnect.

The Kaisheng KH-50000 supports ZPI (Zhaoxin Processor Interconnect) 5.0, allowing partners to install two or four chips on one motherboard. This enables systems with up to 384 cores.

KaiXian KX-7000N

A sudden flood of blacklisted Chinese chipmakers in both the GPU and CPU arenas are preparing to IPO over the coming months. The development comes as the rise of AI and U.S. sanctions against China's semiconductor and supercomputer sectors have had a drastic effect on the development of China's semiconductor industry.

On the one hand, companies like Huawei initiated projects to develop advanced fab tools that could eventually replace equipment produced in the U.S. and Europe. On the other hand, over half a dozen GPU companies emerged to develop processors for AI and gaming. Coincidentally, they are all looking forward to going public and making plans for their IPOs.

There are multiple private GPU developers in China, including Biren Technology, DenglinAI, Innosilicon, Illuvatar CoreX (Tianshu Zhixin), MetaX Technology, Moore Threads, and Zhaoxin. Some of these companies — Innosilicon, Moore Threads, and Zhaoxin — are formally focused on gaming GPUs (though it does not mean that their GPUs cannot be used for AI or other data center applications), others — Biren, DenglinAI, Illuvatar CoreX, and MetaX — primarily develop processors for AI and supercomputer applications.

For now, all of these semiconductor companies are private, but for many, this will change soon.

Some Chinese GPU companies pursue IPO

Perhaps the most promising China-based GPU developer, Biren, remains privately held but has been actively working toward becoming a public company for the last couple of years. It is undergoing the required regulatory guidance process in China, a prerequisite before listing. The timeline and venue (such as Shanghai's STAR Market or Hong Kong) have not been finalized publicly.

Moore Threads is also still private, but it has taken formal steps toward a domestic IPO. In late 2024, it restructured into a joint-stock company, a standard step for firms preparing to go public in mainland China. It is expected to list on one of China's major exchanges, pending regulatory approval.

China’s financial regulator has disclosed that MetaX has finalized the IPO preparation process known as regulatory tutoring, reports TechNode. Over a span of five years, MetaX completed eight investment rounds, securing several billion yuan in total capital. Its estimated market value reached 10 billion yuan (approximately $1.38 billion), according to the 2024 Hurun Unicorn Report.

Zhaoxin, a joint venture between the Shanghai government and Via Technologies, is better known for x86 CPUs with integrated graphics. The company recently applied to list on the Shanghai STAR Market, expected to be completed by mid-2025, which makes it one of the more advanced IPO candidates among China's semiconductor companies. A public offering could occur within the next 12 to 18 months if approvals proceed smoothly.

Biren Technology, Moore Threads, and MetaX are pursuing IPOs primarily to secure substantial funding for continued product development and scaling in a highly capital-intensive industry. All three companies are developing advanced AI and GPU hardware, which demands heavy investment in R&D, silicon validation, software ecosystems, and talent, especially as they aim to rival global leaders like Nvidia.

Beyond funding, public listings also serve strategic and reputational goals. IPOs enable these companies to demonstrate transparency, regulatory compliance, and long-term commitment, thereby positioning them as credible suppliers in China's push for self-reliance in semiconductors. For Moore Threads and MetaX in particular, an IPO can elevate their visibility among enterprise and cloud clients, while Biren may use the opportunity to cement its role as a national AI accelerator champion.

Some developers choose to stay private

When Zhaoxin, a joint venture between Via Technologies and Shanghai Municipal Government, introduced its KaiXian KX-7000 series of x86 processors in late 2023, it implied that the new CPUs will offer tangible performance improvements over its predecessors as they were based on a brand-new microarchitecture. While the new chips are probably faster than their predecessors, they are still years behind AMD and Intel CPUs.

PC Watch this week put Zhaoxin's eight-core KaiXian KX-7000/8 processor (Century Avenue, 8C/8T, 4MB L2, 32MB L3, two DDR4-3200/DDR5-4800 channels, 24 PCIe 4.0 lanes) into a test against AMD's six-core Ryzen 5 5600G (Zen 3, 6C/12T, 3.90 GHz – 4.40 GHz, 3 MB L2, 16 MB L3, two DDR4-3200 channels, 16 PCIe 3.0 lanes, 65W, released in 2021) and Intel's quad-core Core i3-8100 (Coffee Lake, 4C/4T, 3.60 GHz, 6 MB cache, two DDR4-2400 channels, 16 PCIe 3.0 lanes, 65W, released in Q4 2017). While these processors belong to different years and were even positioned differently when released, they operate at the same clock speed, which gives an idea about potential of the latest Zhaoxin microarchitecture (allegedly developed by Centaur Technology). 

PC Watch tested Zhaoxin's KaiXian KX-7000/8 in a variety of benchmarks, some artificial, others are more real-world applications.

When it comes to single-thread CPU performance via CPU-Z benchmark, the KX-7000/8 scored 335.9, significantly lower than the Intel Core i3-8100 (422.2) and AMD Ryzen 5 5600G (615.3). Since the KX-7000/8 has eight cores, it achieved a score of 2517.2 in multi-thread CPU-Z benchmark, outperforming the Core i3-8100 (1618.4) but falling short of the Ryzen 5 5600G (4790.5).  

To a large degree, such behavior defines the capabilities of the KX-7000/8: it can beat rivals in workloads that take advantage of eight cores, but it can never beat them in applications that require superior single-thread performance. 

In Cinebench R23, the KX-7000/8 scored 584 points in single-thread benchmark, which is significantly lower than both the Intel Core i3-8100 (953) and AMD Ryzen 5 5600G (1,435). In the multi-core Cinebench R23 test, the KaiXian KX-7000/8 achieved 3,595, roughly on par with the Core i3-8100 (3,585) but far behind the Ryzen 5 5600G (10,387).  

When it comes to PCMark 10 Express, the Zhaoxin KX-7000/8 exhibits lower performance across all categories compared to the AMD Ryzen 5 5600G and Intel Core i3-8100. The Chinese CPU struggles particularly with productivity, spreadsheet, and app scores, which pretty much indicates limited multitasking capability in heavy workload scenarios. While it may be sufficient for very basic tasks, it is outclassed by both competitors in all aspects. 

Just like its competitors, the KaiXian KX-7000/8 has a built-in graphics core, though its performance is so low that it does not really make sense to discuss it, but to focus on performance of the CPU in gaming applications when it is equipped with an external Radeon RX 6400 graphics card.

In 3DMark, the Zhaoxin KX-7000/8 demonstrates competitive performance with the Core i3-8100 in two scenarios (e.g., Time Spy and Night Raid) and outperforms it in one case (Night Raid), but generally falls behind of the Ryzen 5 5600G in all tests. But this is rather an exception than a rule. 

The KX-7000/8 is significantly behind both of its rivals in the Dragon Quest X benchmark even when it is equipped with an external GPU, which highlights its low single-thread performance. The same applies to the Final Fantasy XIV benchmark: the new CPU is barely a good choice for gaming. 

Where Zhaoxin's KaiXian KX-7000/8 can actually shine are sequential read and sequential write SSD speeds as unlike its rival, it has PCIe 4.0 lanes and therefore it can work with PCIe 4.0 x4 drives at full speed. 

While it is hard to draw conclusions based on one review and a limited number of benchmarks, it looks like Zhaoxin's latest microarchitecture is still years behind those of AMD and Intel in terms of single-thread performance. It is also behind rivals in terms of power consumption and efficiency. Nonetheless, now that Chinese government agencies and state-controlled companies are told not to buy PCs with foreign processors inside, they will buy Zhaoxin-based machines with not so many questions asked. 

The Zhaoxin KX-7000, one of China’s latest desktop CPUs, has finally received its first comprehensive review from PC Watch. The review shows that old budget chips from Intel and AMD are still firmly in the lead.

Made by Shanghai Zhaoxin, the KX-7000 originally debuted in December. It boasts eight cores, a boost clock of up to 3.7 GHz, support for both DDR4 and DDR5 memory, PCIe 4.0, and the company’s latest C-1190 integrated GPU. The older KX-U6880A, the fastest member of the KX-6000 desktop family, also had eight cores but only a maximum clock speed of 3 GHz and lacked support for PCIe 4.0 and DDR5 memory.

Like AMD’s Ryzen CPUs since the 3000 series, the KX-7000 uses a chiplet design: one chip for CPU cores and another for I/O functions. Zhaoxin hasn’t disclosed what company fabs the KX-7000 chiplets, nor what node they’re fabbed on.

Although the KX-7000 slots into Intel’s LGA1700 socket, it’s almost sure that the Chinese CPU won’t work in 600- or 700-series motherboards since Zhaoxin has likely wired its chips differently than Intel. Instead, using the LGA1700 socket seems to be a move to ensure broader cooler compatibility, an important consideration for a socketed processor.

Perhaps the most notable characteristic of the KX-7000 and other Zhaoxin chips is that they’re based on the x86 architecture, just like CPUs made by Intel and AMD. Zhaoxin inherited its x86 license from VIA Technologies when it partnered with the local Shanghai government to form the KX-7000 CPU designer.

PC Watch tested its KX-7000 on a generic, unlabeled motherboard that supports DDR4 instead of the newer DDR5 made by ASUS. Strangely enough, only DDR4 clocked strictly at 3,200 MHz by default could get the CPU to boot for the first time. According to PC Watch, this limitation can be changed in the BIOS. The publication raised concerns over the firmware and software support, particularly access to drivers, which the reviewer only got access to thanks to “determination.”

As for performance, the KX-7000 made gains over its predecessor but couldn’t quite catch up to old, low-end CPUs like the Core i3-8100 and Ryzen 5 5600G. In Cinebench R23, the KX-7000 could only tie the 8100 in the multi-threaded test and was only 60% as fast as the Intel chip in the single-threaded test. The 5600G, meanwhile, was about three times faster in both single- and multi-threaded performance.

The C-1190 iGPU, which PC Watch claims, is based on the Arise GT10C0 graphics card made by Chinese graphics company Glenfly. Given that Glenfly’s latest Arise 1020 didn’t impress, it’s unsurprising that the review found the KX-7000’s iGPU to be pretty slow, scoring a paltry 42 points in 3D Mark Time Spy to the 8100’s 469 and 5600 G’s 1440. The C-1190 was similarly behind in other 3DMark tests.

When equipped with an RX 6400, AMD’s cheapest and slowest graphics card, the KX-7000 could finally match the 8100 in 3DMark and Final Fantasy XIV. However, it wasn’t nearly as good in Dragon Quest X, an older game that relies more heavily on single-threaded performance.

The KX-7000 cannot battle modern Intel and AMD CPUs, and it’s unlikely Zhaoxin chips will be able to go head-to-head with Western x86 CPUs soon. However, performance is beside the point for China’s semiconductor industry, and self-sufficiency is the more important goal as the US government gradually places more trade restrictions on China. Earlier this year, the US government blocked Intel from selling Meteor Lake CPUs to Huawei, making domestically produced chips more appealing concerning supply chain stability.

Six major Russian supercomputer centers located across the country have teamed up to form the 'Distributed Scientific Supercomputer Infrastructure' consortium, reports the Russian Academy of Sciences. This new alliance aims to pool resources for the performance needed to process modern scientific workloads. In total, the six supercomputing centers have around 900 servers and these fleets cannot be upgraded due to the U.S. sanctions. To that end, pooling resources is largely a forced measure. Meanwhile, DSSI has a plan to collectively upgrade their hardware.

Six supercomputing centers join forces

The consortium's consolidated infrastructure includes 900 servers with a total peak performance of 1.5 FP64 PetaFLOPS and scientific data storage systems exceeding 15PB. This infrastructure supports distributed systems for the collection, storage, and analysis of scientific data across various regions. To put the number into context, 1.5 FP64 PLOPS performance trails the slowest supercomputer in the Top 500 list, which has Rmax performance of 2.13 PetaFLOPS. Members of the DSSI are providing computing resources and support to 240 organizations across the country and by working together they will be able to offer more compute performance to their customers as the latter's requirements increase. 

The consortium does not include Russia's most powerful state-owned supercomputer, the Lomonosov 2 located in the Lomonosov Moscow State University, which has Rmax performance of 2.48 FP64 PFLOPS. It should be noted that the MSU introduced its MSU-270 supercomputer with a peak computational power of 400 'AI' PFLOPS last Fall and this one also does not participate in the DSSI project. Commercial supercomputers, such as Yandex's Chervonenkis, Russia's most powerful machine with Rmax Linpack performance of 21.53 FP64 PFLOPS, will also not be a part of the consortium. 

The DSSI consortium brings together multiple players from different regions, including the Far Eastern Federal Research Center in Khabarovsk, the Institute of Automation and Control Processes in Vladivostok, the Matrosov Institute in Irkutsk, the Institute of Computational Mathematics and Mathematical Geophysics in Novosibirsk, the Krasovsky Institute in Yekaterinburg, and the Space Research Institute in Moscow.  

"We need to move away from competition between collective-use centers and instead focus on uniting their resources and expertise," said Alexey Sorokin, director of the computing center of the Far Eastern Branch of the Russian Academy of Sciences and a member of the consortium's Coordinating Council. "There is a growing demand for a distributed infrastructure that connects technological sites across different regions of the country via high-speed communication channels. These sites should be equipped with modern computing systems of various architectures and specialized data storage systems. It is impossible for even the largest organization to tackle this complex task alone. It makes sense to leverage the groundwork laid by existing supercomputer centers, which have already proven their efficiency. This approach will undoubtedly lead to more efficient use of financial resources and, just as importantly, foster the development of scientific teams in the regions, particularly in HPC space."  

And have upgrade plans

Due to sanctions imposed by the U.S. after Russia invaded Ukraine, Russian state-owned entities have a hard time acquiring supercomputer hardware, such as the latest processors from AMD, Intel, and Nvidia. While smuggling exists (and largely prospers), it is likely that Russian institutions cannot get enough new machines to modernize their fleets.  

Pooling in resources will enable Russian scientists to get access to higher performance. Also, the DSSI consortium aims to develop a comprehensive project to upgrade scientific supercomputer centers for collective use with modern compute hardware. After discussions with all interested participants, the project will be submitted to the Ministry of Science and Higher Education of the Russian Federation and the Russian Academy of Sciences. 

Given the sanctions imposed by the U.S., EU, Taiwan, South Korea, and other countries, it will not be easy to upgrade all of the aforementioned datacenters with modern Western hardware. But perhaps the plan is to use parts that can be obtained without any curbs: including usage of cut-down components like Nvidia's HGX H20 or hardware from Chinese developers. However, it is evident that the Russian scientific community wants access to higher performance.

Zhaoxin, the Chinese tech firm behind CPUs like the Kaixian KX-6000 and KX-7000 series, has announced a flurry of computer designs from demostic vendors. It says companies including Lenovo, Tsinghua, Ziguang, Ruijie, Honghe, and Seewo have a broad range of Kaixian processor-powered devices ready. The firm shared the “big news” in a blog post about the recent 83rd China Educational Equipment Exhibition.

Quite a wide range of systems were showcased at the educational event. There was a mix of “PC terminals, notebooks, smart blackboards, and interactive large screens based on Zhaoxin's independent processor platform,” on the show floor according to the company.

Probably the biggest Kaixian KX-7000 series processor partner is Lenovo, which showed off a “high-end flagship,” desktop PC dubbed the P90z G1t. Inside this nondescript 17-liter black tower resides a KX-7000 series processor with eight cores, running up to 3.6 GHz. The standard configuration also comes with 16GB of RAM (expandable to 128GB), and a 512GB M.2 NVMe SSD (with spare capacity for more drives). This desktop tower also has 10M/100M/1000M Ethernet, 8x USB 3.0 Type-A ports, and one USB 3.0 with a Type-C connector. Optional discrete graphics and wireless networking can be configured.

Lenovo also showcased a more compact model, dubbed the M90z G1s. Specs look very similar to the above, and it can also fit a discrete graphics card if the iGPU isn’t powerful enough for the use case. The only difference we could see on the spec list is that this system is more compact at 8.2 liters. We will assume the M90z G1s will face more restrictions in the amount of extra storage devices it can house, and perhaps the size of the discrete GPU which can be fitted.

Elsewhere in the Zhaoxin press release about the education industry design wins we can see Tongfang and Ruijie showcased 15 liter and SFF PCs. They all seem to use the same processor: the Zhaoxin Kaixian KX-7000 series processor, with eight cores, and clocks up to 3.6 GHz. Sadly, there’s no further information or pictures showing devices like laptops and smart blackboards.

As an x86 platform, the above-outlined systems should be able to run Microsoft Windows or various Western Linux distros. However, China’s current thrust is in the direction of both hardware and software independence. With that in mind, Zhaoxin highlights “UOS, Galaxy Kirin, Zhongke Founder, and other operating systems,” are all happy to reside on these new platforms.

The above design wins are evidence of newfound momentum behind Zhaoxin and its new KX-7000 series processors. It sounds like many more are in the pipeline, with domestic brands like Unisoc, Shengteng, Lianhe Donghai, Haier, Datang Gaohong, Baoxinchuang, Zhiwei Intelligent, Weibu, Yidao, Medijie, Boli Intelligent claimed to be preparing Kaixian KX-7000-powered devices. Moreover, Zhuoyi and Hengwei are making systems for the government, industry, and enterprise.

Benchmark evidence we have seen indicates that the Kaixian KX-7000 series processors can perform over twice as fast as their KX-6000 predecessors. From a wider perspective, though, these Chinese CPUs are probably only capable of being competitive with AMD and Intel CPUs from the mid-to-late 2010s.

A leaked Geekbench 6 result for Zhaoxin's new KX-7000 CPU shows that it's over twice as fast as its old KX-6000 series KX-U6780A. The benchmark score indicates that not only has Zhaoxin substantially improved clock speeds, but also instructions-per-clock (or IPC) with its latest CPU architecture. However, the KX-7000 is only about on par with competing AMD and Intel CPUs from the mid-to-late 2010s.

Zhaoxin isn't the only CPU designer in China, but it is unique given that it has a license to the x86 architecture that AMD and Intel use for their processors. The company announced its KX-7000 CPUs late last year, and claimed they would be twice as fast as the prior KX-6000 series. The spec sheet indicated roughly a 20% frequency gain and significantly faster DDR5 and DDR4 memory support (KX-6000 only supports DDR4).

Those improvements alone probably wouldn't double performance, but the purported architectural advancements may make up the difference. Zhaoxin boasts that KX-7000 and its latest Century Avenue architecture feature a new front-end, introduce out-of-order execution, and a better cache and memory system. 

KX-7000 adds 32MB of L3 cache to the mix for a total of 36MB combined L2 and L3 cache, compared to KX-6000's 8MB of L2. Plus, KX-7000 supports faster DDR5-4800 and DDR4-3200 RAM while KX-6000 only supports stock DDR4-2666.

In the table below, we've compared the KX-7000 chip's score to a KX-U6780A result, as well as the aggregate score of the Ryzen 5 2500X and the Core i3-8300, according to Geekbench. For the KX-U6780A, we selected the median result benchmarked in Windows (just like the KX-7000 was), though there were both significantly faster and slower scores.

Though KX-7000 is apparently rated for up to 3.7 GHz, Geekbench says that it only hit 3.3 GHz. Nevertheless, the CPU was almost three times as fast as the KX-U6780A, which is the second-fastest processor in the KX-6000 lineup. Although we don't know what kind of RAM the KX-7000 sported during benchmarking, we can conclude that KX-7000's architectural improvements are largely responsible for the uplift in performance, assuming that this is a legitimate result.

Similarly performing AMD and Intel CPUs include the quad-core Ryzen 5 2500X and Core i3-8300. These chips had a pretty similar multi-core score, but pulled ahead substantially in the single-core score. These AMD and Intel CPUs have a clock speed roughly 20% faster than the KX-7000, but are about 40% and 60% faster respectively, implying that much of the performance gap is down to architectural design.

While this isn't an amazing position for KX-7000 to be in, the new CPU supports as many PCIe lanes at 4.0 speeds and as much RAM as more recent-gen AMD and Intel platforms. Zhaoxin does lack PCIe 5.0, but given that only SSDs at the moment are taking advantage of PCIe 5.0, it's not a massive disadvantage.

Although KX-7000 clearly won't be breaking any records, for China it's a big improvement. Technological autonomy is a priority for the country, and while performance leadership would be nice, it's not necessary. With performance roughly on par with CPUs like the 2500X and the 8300, KX-7000 could power decent PCs made entirely from Chinese hardware.

Although Zhaoxin's KX-7000 processors based on the company's latest Century Avenue microarchitecture promise a significant performance uplift compared to their predecessors in general-purpose applications, its integrated graphics does not impress. According to Benchleaks data, performance is lower than that of a decade old iGPU.

Zhaoxin's eight-core Kaixian KX-7000/8 processors integrate the company's latest ZX C-1190 iGPU design, which is said to be DirectX 12/OpenGL 4.6/OpenCL 1.2-capable and provide four-times the performance of its predecessor. This may not be enough to be competitive on the modern market though as it scores just 2024 points in Geekbench 5 OpenCL benchmark (via @Benchleaks).

To put the number into context. Intel's slowest integrated UHD Graphics 730 GPU benchmark scores are around 7500 points. AMD's lowest-end discrete desktop Radeon RX 6300 graphics card scores 26,127 points in the same test. Even Intel's 10-years old HD Graphics 4400 scores 2500 ~ 3000 points in this benchmark.

Of course, an OpenCL benchmark does not necessarily reflect gaming capabilities of a GPU, but it gives us an idea of how the compute capabilities of one graphics processor stack up against compute capabilities of another. Perhaps, Zhaoxin will be able to boost performance of its ZX C-1190 iGPU with better drivers, but it does not look like this is the case here.

Zhaoxin's integrated graphics processor is not supposed to be a performance champion. It packs four compute units operating at 700 MHz and while we do not have any information about exact specifications of these CUs, it is unlikely to think that each one packs hundreds of stream processors. 

While Zhaoxin's ZX C-1190 iGPU does not impress with OpenCL compute performance, it can still handle decoding and encoding of H.265/H.264 video at up to 4K as well as driving DisplayPort, HDMI, and D-Sub/VGA outputs. Since Zhaoxin's CPUs are designed primarily for office PCs and workhorse notebooks, its graphics capabilities are hardly something that its users will ever think about, as long as it can draw windows fast enough.  

Zhaoxin, a joint venture between Via Technologies and the Shanghai Municipal Government, is adding support for 'preferred cores' in their processors into the Linux kernel, mirroring the approach that Intel and AMD have used for years. The move is designed to improve the single-thread performance of Zhaoxin's CPUs by distinguishing cores that can reach higher frequencies than others, thus pinning latency-sensitive work into the fastest cores. 

General-purpose cores in multi-core processors tend to have different clock rate potentials: some of them can achieve higher operational frequencies, leading to a disparity among the cores and the necessity to bind certain workloads to these cores to get higher overall performance. This is why Zhaoxin has introduced a series of patches to the Linux kernel, as reported by Phoronix.

These patches enable the Linux kernel's scheduler to recognize and prioritize tasks on these high-performing cores by using the ACPI CPUFreq driver that relies on the cppc_get_highest_perf function in the CPPC driver. Once the driver detects the frequency of each core, it recognizes the highest frequency value as a priority indicator for the scheduler to set core priorities and favor these preferred cores.

However, there is a catch about these Zhaoxin patches: they lack specific details about the CPUs they are meant to support. Whether this preferred cores feature is currently available in Zhaoxin processors or is intended for future CPUs, such as the upcoming Kaixian KX-7000 based on the Century Avenue architecture, remains unclear. 

Meanwhile, Phoronix notes that Zhaoxin has recently put a lot of effort into patching the Linux kernel for its Yongfeng (CentaurHauls) microarchitecture, which powers the KaiSheng KH-40000 processors that are aimed primarily at data centers.

Assigning certain workloads to preferred cores makes a lot of sense for data center and consumer applications, so adding support for this feature to Linux is good news for those who plan to use Zhaoxin CPUs. It also emphasizes that Zhaoxin is gaining functionality akin to CPUs from AMD and Intel.

We've been watching information concerning China’s Glenfly Arise graphics cards emerge for at least the last year. Now, a self-confessed graphics card enthusiast and collector on Twitter has managed to acquire a Glenfly Arise 1020 sample from the PRC. The tests are ongoing, as refreshing Löschzwerg’s Twitter feed rewarded us with some new benchmark results. However, the first card in the west (maybe?) definitely has some issues.

Starting with the good news, there is an enthusiast outside of China testing one of these graphics cards in apps, games, and benchmarks that we are familiar with. That makes it easier to see how it stacks up against existing Nvidia, AMD, and Intel GPUs.

Löschzwerg shared some artfully taken photos of the simple looking Glenfly Arise 1020 earlier today. It has a sparsely populated PCB, and looks like it could fit with a low-profile bracket. The cooler appears to be a simple model designed for under 50W or so. It has just two connectors: an old VGA monitor connector, and a newer HDMI output.

In other words, it's very much the same card as we reported on this time last year, when images were shared by a Bilibili user based in China. That card looked very much like the reference model pictured above.

Initial testing of the Glenfly Arise 1020 appeared to disappoint Löschzwerg, as simply firing up the GPU-Z tool caused a system freeze. That's not a good start, but then GPU-Z does some low-level hardware accesses and if it tries to pull data from an unknown GPU, strange things can happen. The Project PhysX benchmark also fell flat on its face using this Shanghai Zhaoxin Semiconductor Company produced graphics card.

More luck was had with an old 3DMark graphics benchmarking tool, specifically 3DMark 11. Löschzwerg commented that in successfully running this benchmark, the Glenfly Arise 1020 demonstrated an advantage over the Moore Threads MTT S80.

In 3DMark 11, the Chinese graphics card, which we think uses mid-2000s era S3/Via GPU technology, scored as follows: Entry – E1980, Performance - P1081, and Extreme X275. What can we find for comparable reference points in the history of AMD / ATI / Nvidia graphics cards?

Looking back at a roundup of single-slot graphics cards on Tom’s Hardware dated June 2011, all of the models easily outpaced this Chinese GPU. We are talking about graphics cards such as the AMD Radeon HD 6850, Nvidia GeForce GTS 450, and Nvidia GeForce GT 440. That’s a shame, and to add insult to injury, ancient GPUs like the Radeon HD 7970 and GeForce GTX 680 appear to be about 10X faster than the Glenfly Arise 1020 using 3DMark 11 Extreme scores as a comparison metric.

Looking back at some Bilibili sourced tests, we saw evidence that the Glenfly Arise 1020 was about 60X slower than the relatively modern GeForce GTX 1060 (Geekbench OpenCL). It may not be quite that bad, but it's certainly not going to keep up with the best graphics cards.

Noting the above poor performance in last year’s tests and those by Löschzwerg today, we hope that some driver and OS wrangling could allow the new Chinese graphics card to improve over time. We know that it's pretty limited with its four CUs (256 shaders) mated with 2GB VRAM on a 64-bit bus, but these benchmark results are worse than expected.

Last time we mused whether the delivery of a working or improved quality Windows driver, or testing the Chinese hardware on a Chinese OS like the Tongxin/Galaxy Kirin operating system, would make a difference. However, we don't expect any massive changes in the coming months, especially considering the base hardware specs.

These early results remain interesting to see, and we hope Löschzwerg makes some good progress, alongside tech enthusiasts in China, in trying to demonstrate the full potential of the Glenfly Arise 1020. TL;DR: Making GPUs is hard.

The number of GPU startups in China is extraordinary as the country tries to gain AI prowess as well as semiconductor sovereignty, according to a new report from Jon Peddie Research. In addition, the number of GPU makers grew worldwide in recent years as demand for artificial intelligence (AI), high-performance computing (HPC), and graphics processing increased at a rather unprecedented rate. When it comes to discrete graphics for PCs, AMD and Nvidia maintain lead, whereas Intel is trying to catch up. 

18 GPU Developers

Tens of companies developed graphics cards and discrete graphics processors in the 1980s and the 1990s, but cut-throat competition for the highest performance in 3D games drove the vast majority of them out of business. By 2010, only AMD and Nvidia could offer competitive standalone GPUs for gaming and compute, whereas others focused either on integrated GPUs or GPU IP.

The mid-2010s found the number of China-based PC GPU developers increasing rapidly, fueled by the country's push for tech self-sufficiency as well as the advent of AI and HPC as high-tech megatrends. 

In total, there are 18 companies developing and producing GPUs, according to Jon Peddie Research. There are two companies that develop SoC-bound GPUs primarily with smartphones and notebooks in mind, there are six GPU IP providers, and there are 11 GPU developers focused on GPUs for PCs and datacenters, including AMD, Intel, and Nvidia, which design graphics cards that end up in our list of the best graphics cards.

In fact, if we added other China-based companies like Biren Technology and Tianshu Zhixin to the list, there would be even more GPU designers. However, Biren and Tianshu Zhixin are solely focused on AI and HPC for now, so JPR does not consider them GPU developers. 

China Wants GPUs

Being the world's second largest economy, China inevitably competes against the U.S. and other well-developed countries in terms of pretty much everything, including technology. China did a lot to lure engineers from around the world and make it worthwhile to establish various chip design startups in the country. In fact, hundreds of new IC design houses emerge in China every year. They develop all kinds of things from tiny sensors to complicated communication chips, thus enabling the country's self-sufficiency from Western suppliers.  

But to really jump on the AI and HPC bandwagon, China needs CPUs, GPUs, and special-purpose accelerators. When it comes to computing, it is impossible for Chinese companies to leave behind long-time CPU and GPU market leaders any time soon. Yet, it is arguably easier and perhaps more fruitful to develop and produce a decent GPU than try to build a competitive CPU.  

"AI training was the big motivator [for Chinese GPU companies], and avoidance of Nvidia's high prices, and (maybe mostly) China's desire for self-sufficiency," said Jon Peddie, the head of JPR. 

GPUs are inherently parallel, which means there are loads of compute units inside that can be used for redundancy, which makes it easier to get a GPU up and running (assuming that per transistor costs are relatively low and overall yields are decent). Also, since GPUs are fundamentally parallel, it is easier to parallelize them in scale-out manner. Keeping in mind that China-based SMIC does not have production nodes as advanced as those of TSMC, this way of performance scaling looks good enough. In fact, even if Chinese GPU developers lose access to TSMC's advanced nodes (N7 and below), at least some of them could still produce simpler GPU designs at SMIC and address AI/HPC and/or gaming/entertainment market. 

From China's perspective as a country, AI and HPC-capable GPUs may be arguably more important than CPUs too since AI and HPC can enable all-new applications, such as autonomous vehicles and smart cities as well as advanced conventional arms. The U.S. government of course restricts exports of supercomputer-bound CPUs and GPUs to China in a bid to slowdown or even constrain development of advanced weapons of mass destructions, but a fairly sophisticated AI-capable GPU can enable an autonomous killer drone, and drone swarms represent a formidable force, for instance. 

GPU Microarchitecture Is Relatively Easy, Hardware Design Is Expensive

Meanwhile, it should be noted that while there are a bunch of GPU developers, only two can actually build competitive discrete GPUs for PCs. That is perhaps because it is relatively easy to develop a GPU architecture, but it is truly hard to properly implement it and to design proper drivers. 

CPU and GPU microarchitectures are essentially at the intersection of science and art. They are sets of sophisticated algorithms that can be developed by rather small groups of engineers, but they might take years to develop, says Peddie.  

"[Microarchitectures] get done on napkins and white boards," said Peddie. "[As for costs] if it is just the architects themselves, that [team] can be as low as one person to maybe three – four. [But] architecture of any type, buildings, rocket ships, networks or processors is a complicated chess game. Trying to anticipate where the manufacturing process and standards will be five years away, where the cost-performance tradeoffs are, what features to add and what to drop or ignore is very tricky and time-consuming work. […] The architects spend a lot of time in their head running what-if scenarios — what if we made the cache 25% bigger, what if we had 6,000 FPUs, should we do a PCIe 5.0 I/O will it be out in time." 

Since microarchitectures can take years to develop and they require talented designers, in a world where time-to-market is everything, many companies license an off-the-shelf microarchitecture or even a silicon-proven GPU IP from companies like Arm or Imagination Technologies. For example, Innosilicon — a contract developer of chips and physical IP — licenses GPU microarchitecture IP from Imagination for its Fantasy GPUs. There is another China-based GPU developer, which uses a PowerVR architecture from Imagination. Meanwhile, Zhaoxin uses a highly reiterated GPU microarchitecture it acquired from Via Technologies, which inherited it from S3 Graphics. 

The cost of developing a microarchitecture may vary, but it is relatively low when compared to the costs of a physical implementation of a modern high-end GPUs. 

For years, Apple and Intel, both companies with plenty of engineering talent, relied on Img for their GPU designs (Apple still does to a certain extent). MediaTek and other smaller SoC suppliers rely on Arm.  Qualcomm used ATI/AMD for an extended period, and Samsung uses AMD after several years of trying to design their own graphics engine.

Two of the new Chinese companies have hired ex AMD and Nvidia architects to start their GPU companies, and another two use Img. Time to market and learning the skills of being an architect, what to worry about, and how to find a fix is a very time consuming process.  

"If you can go to a company that already has a design and have been designing for a long time, you can save a boatload of time and money – and time to market is everything," said the head of Jon Peddie Research. "There are just so many gotchas. Not every GPU designed by AMD or Nvidia has been a winner. [But] a good design lasts a couple of generations with tweaks." 

Hardware implementation and software development are prohibitively expensive with new production nodes. International Business Times estimates that the design costs for a fairly complex device made using 5nm-class technology exceeds $540 million. These costs will triple at 3nm. 

"If you include layout and floor plan, simulation, verification, and drivers then the [GPU developer] costs and time skyrocket," explained Peddie. " The hardware design and layout is pretty straight forward: get one trace wrong and you can spend months tracking it down." 

There are just a few companies in the world that can develop a chip featuring the complexity of modern gaming or compute GPUs from AMD and Nvidia (46 billion — 80 billion transistors), yet China-based Biren could do something similar with its BR104 and BR100 devices (we speculate that the BR104 packs some 38.5 billion transistors).  

Thoughts 

Despite prohibitive costs, eight out of the 11 PC/datacenter GPU designers are from China, which speaks for itself. Perhaps we won't see a competitive discrete gaming GPU from anyone except huge American companies in the near future. That's partly because its hard and time consuming to develop a GPU, and to a large degree it requires a prohibitively expensive hardware implementation for these high-complexity GPUs. Whether or not China can field competitive entrants remains to be seen, but any failure won't stem from a lack of trying. 

Zhaoxin, a China-based CPU developer with an x86 license, has yet to formally introduce its next-generation KaiSheng KH-40000 processors with up to 16 cores for datacenters. However, it has already started to submit benchmark results to the Geekbench 5 database. The new CPUs show noticeable microarchitecture-related performance improvements over their predecessors but can barely catch up with modern CPUs from AMD and Intel.

Mysterious CPUs

Zhaoxin, co-owned by Via Technologies and the Shanghai Municipal Government, has been gradually leveraging microarchitectures designed by Via (or rather by Centaur) since the mid-2010s, and its upcoming KaiSheng KH-40000 series processors for datacenters are based on the CentaurHauls microarchitecture that some claim resembles Intel's Haswell microarchitecture from 2013.  

The KaiSheng KH-40000/16 and KaiSheng KH-40000/12 CPUs run at 2.20 GHz, have 16 and 12 cores, and are equipped with 32MB and 24MB of L3 cache, respectively. In addition, the 16-core model seems to feature simultaneous multithreading technology (SMT), so it can process up to 32 threads concurrently, assuming that Geekbench 5 correctly reads its capabilities. Based on specifications of Zhaoxin's KaiSheng KH-40000/16 and KaiSheng KH-40000/12 published in the Geekbench 5 database, these CPUs look very similar to Centaur's never-released CHA processor unearthed earlier this year.  

There are differences though: CHA had eight cores, did not support SMT, and was architected for TSMC's N16 node, whereas KaiSheng KH-40000 has up to 16 cores, seems to feature SMT, and is believed to be designed for TSMC's N7 fabrication process. Furthermore, processor IDs of both KH-40000 CPUs read 'CentaurHauls Family 7 Model 11 Stepping 3' (1, 2), whereas the processor ID of Centaur's CHA is 'CentaurHauls Family 6 Model 71 Stepping 2,' so the CPUs in question use different silicon.  

What is odd, though, is that both CHA and KH-4000 operate at 2.20 GHz, so if we did not know CPU IDs, we could speculate that the model KH-4000/16 uses two eight-core CHA dies produced on TSMC's N16 node and glued together using an interconnect.

Mediocre Performance

For Zhaoxin, CentaurHauls should be a significant microarchitectural advancement from its LuJiazui microarchitecture from 2019. Furthermore, the improved core count should make KaiSheng KH-40000 CPUs more competitive on the server market. So, let's look at the performance numbers submitted by the CPU developer.

When it comes to single-threaded performance, Zhaoxin's (or Centaur's) CentaurHaul microarchitecture significantly outpaces the company's previous generation LuJiazui microarchitecture both in integer (by 22%) and floating point (by 75%) workloads even though the new CPU operates at 2.20 GHz. In contrast, the older one works at 2.70 GHz. The FPU performance uplift seems rather dramatic, but one should remember that we are dealing with a synthetic benchmark.

While the new microarchitecture is significantly better than the preceding one, KaiSheng KH-40000 CPUs with 12 and 16 cores cannot compete against any modern CPUs. Moreover, their single-threaded performance is even lower than that of ill-fated AMD's Bulldozer/Piledriver architecture from mid-2012.

As for multi-thread performance, we see a rather odd advantage that Zhaoxin's 16-core KaiSheng KH-40000/16 with SMT has over 12-core KaiSheng KH-40000/12 CPU. While, in theory, the 16C/32T chip can process 2.66 times more threads than its 12C/12T brethren (and we have never seen this kind of SMT efficiency from any well-known CPU microarchitecture so far), its actual performance advantage is higher than even hypothetical 2.66X (2.69X in integer, 2.84X in float). As we are dealing with a situation when one CPU only has four more cores than its rival, yet its performance is almost three times higher, we believe that there are factors beyond the number of cores that have such an effect on performance. 

Keeping in mind that Windows 10/11 does not always work optimally with schedulers of unfamiliar multi-core CPUs, we believe that the 12-core KaiSheng KH-40000/12 CPU results obtained on Windows 10 Pro do not reflect its true potential. 

Yet, even under Windows 10 Pro and without SMT, CentaurHoals is substantially faster than LuJiazui in multi-threaded integer (by 40%) and multi-threaded floating point (78%) workloads. The problem is that absolute performance numbers demonstrated by both KaiSheng KH-40000 and Centaur CHA CPUs are deficient by today's standards. 

Interestingly, multi-threaded performance numbers demonstrated by Zhaoxin's 12-core KaiSheng KH-40000/12 under Windows and without SMT are comparable to AMD's FX-8350 processor (four modules, eight threads), which the company once marketed as an eight-core CPU. We can hardly call the performance of a decade-old processor competitive by today's standards, at least in Geekbench 5, which is not the best benchmark.

Some Thoughts

While 12-core and 16-core configurations seem okay for desktops and entry-level servers, 12 and 16 cores from Zhaoxin do not deliver performance comparable to that of 12-core or 16-core processors from AMD and Intel. Under Windows and judging only by Geekbench 5 scores, Zhaoxin seems to be a decade behind AMD and Intel regarding performance. Even if Zhaoxin enables SMT on its upcoming CentaurHoals-based CPUs (for client and server applications) and Windows 'learns' how to properly use those cores, KaiSheng KH-40000/16 will still be two times slower than 2021 processors from AMD and Intel with the same core count.

Innosilicon has introduced its second-generation standalone graphics processor, the Fenghua II (aka Fantasy II). The new GPU has a power consumption between 4W and 15W, so it does not require either auxiliary PCIe power or active cooling, which is why it can target entry-level desktops and notebooks.  

Despite expectations, Innosilicon's second-generation discrete GPU does not offer higher performance than its predecessor. In fact, with raw compute performance of 1.5 FP32 TFLOPS/12 INT8 TOPS, the Fantasy II is actually more than three times slower than single-chip Fantasy I graphics card (which offers up to 5 FP32 TFLOPS/up to 25 INT8 TOPS performance), according to 163.com (via @Loeschzwerg_3DC) .The indisputable advantage of the new graphics card is its power consumption that ranges between 4W and 15W as opposed to 20W to 50W for single-chip Fantasy I. Meanwhile, performance-per-watt of the novelty remained at the same level with its bigger sibling.

Innosilicon's Fantasy I and Fantasy II Graphics Cards

Given the compute performance of the Fantasy II GPU, it will compete against Nvidia's GeForce GTX 1630 graphics board that uses the TU117 GPU originally introduced in April 2019. In fact, even AMD's Radeon RX 6400 might outperform the Fantasy II, based on its compute capabilities (and something tells us that this one will never find itself in our list of the best graphics cards available today). Meanwhile, Innosilicon does not publish performance results of its new graphics chip in real games. The only thing it says is that it hits 6500 in the outdated GLMark2 benchmark, according to ITHome. 

Innosilicon does not disclose which architecture powers its Fantasy II GPU, though we may speculate that the new graphics chip uses the same ImgTec's PowerVR architecture which is used for the Fantasy I GPU. That said, the new graphics processor supports DirectX, Vulkan, OpenGL, OpenCL, and even OpenGL ES application programming interfaces (APIs). Meanwhile, the new graphics processor has a built-in RISC-V security core with physical unclonable function (PUF). 

The Fantasy II GPU can support 2GB, 4GB, 8GB LPDDR4/4X/5/5X memory configurations with an up to 10 GT/s data transfer speed. The chip supports a DisplayPort/eDP 1.4, two HDMI 2.0, LVDS, and D-Sub/VGA outputs, though there is no word how many display pipelines it has (i.e., how many monitors it can support simultaneously). As for host interface, the chip uses a PCIe 3.0 x8 bus. 

While Innosilicon's Fantasy II is definitely not the best entry-level GPU around, it has another advantage over discrete graphics products from AMD, Intel and Nvidia (in addition to its low power consumption): it supports the vast majority of Linux distributions (including those exclusively used in China), such as CentOS, Kirin, KylinOS, Tongxin, UOS and Ubuntu. Furthermore, it supports all Chinese CPU platforms (something that GPUs from leading vendors do not support), such as Loongson, Huawei Kunpeng, Hygon Dhyana, Shenwei (Sunway) and Zhaoxin Feiteng. Furthermore, Innosilicon plans to release Windows 10 drivers for its Fantasy-series GPUs.

Images of a Glenfly Arise-GT10C0 discrete graphics card, and its exposed GPU, have been shared on Chinese social media platform Bilibili by user Hoki. This is a very basic graphics card by modern measures, with a boxy and utilitarian half-shroud cooler, and outputs limited to just VGA and HDMI. Moreover, the Bilibili post indicates that software support is still shaky, at least in Windows 10. So it's extremely unlikely that AMD or Nvidia need to worry about losing a spot on our best graphics card list anytime soon. 

The image above shows the Glenfly GPU, exposed by the removal of the half shroud cooler assembly. The inscriptions on the GPU aren't very revealing but confirm the GPU is part of the Arise–GT family, model 10C0. Other codes on the chip don't reveal anything to us. The model number is interesting though, because a couple weeks ago we saw some Geekbench 5 benchmark results for what was recognized as a Glenfly Arise 1020. It is hard to know if this is a different member of the Glenfly Arise discrete graphics cards family, or if the name difference is due to Geekbench sysinfo reporting quirks. Some Chinese tech sites today are talking as if the previous Geekbench info is from this exact same card.

To recap, the Geekbenched sample was tested in a Windows 10 system with a Zhaoxin 2.7GHz octacore CPU and 8GB of RAM. It was reported as coming with 24 compute units, a GPU clock of 608 MHz, and 2GB of VRAM. In OpenCL testing it scored 579, which is on a par with half-decade old smartphone GPUs. In comparison with a familiar PC GPU, like the classic Nvidia GeForce GTX 1060, the Glenfly was about 60x slower in the Geekbench OpenCL tests.

The new Bilibili post about the Glenfly Arise-GT10C0 indicates it has a lot in common with the benchmarked sample. It looks a lot like the Glenfly reference design included in our previous article, as well as using a similar cooler, with the same outputs. The new post about this card says the GT10C0 has 2GB of VRAM, too.

If the hardware specs are underwhelming, it's also worth pointing out that Glenfly hasn't offered up a driver. Still, the Bilibili poster managed to run Windows 10 and an app by the name of 'Haiguang C3138.' In Windows 10, the tester found there were glitches in HDMI output. Moreover, when attempting to play back local video files, most of them didn't show an image, says the Bilibili user.

The above-noted results are admittedly poor, but there may be potential for a lot better performance if a working Windows 10 driver is delivered. Capable software drivers are essential to graphics card performance, especially any kind of 3D acceleration. Just try and run your PC with the graphics driver uninstalled and see what happens.

We might see more of this Glenfly graphics card again shortly. The Chinese social media user says they will try and test it under the Tongxin/Galaxy Kirin operating system. Perhaps these subsequent tests will show improved potential.

A mysterious GPU from China has been spotted in the Geekbench online results browser. The benchmarking app was used to run tests on a system that appears to be mostly based on technology from Shanghai Zhaoxin Semiconductor Company. It has a Zhaoxin KaiXian CPU, Zhaoxin motherboard, and a "Zhaoxin Glenfly GPU." That last one is a new name.

Twitter's Benchleaks surfaced this OpenCL result, which can give some indication of GPU performance. We wouldn't put to much weight on the current score, however, as 579 for Geekbench OpenCL is quite terrible. For comparison, even the lowly GTX 1650 scores around 38,000, while an RTX 3070 scores in the 135,000 range.

The Chinese PC system sounds modern and capable enough on first impressions. Its CPU has 8C/8T and runs at approximately 2.7 GHz and is accompanied by a modest 8GB of DDR4 RAM. It has a discrete GPU called the Glenfly Arise 1020, with 24 compute units (even though that's an AMD GPU term) and 2GB of VRAM. These figures don't match up very well with the poor OpenCL score in Geekbench 5.

The Benchleaks comparison with an RTX 3070 seems pretty nonsensical, so we thought we would look through the full table of OpenCL scores maintained by Geekbench. One has to scroll down the page a long way to see scores in the 579 region. The only GPUs in this area of the table are vintage AMD iGPUs like the Radeon HD 6480G and Radeon HD 6410D from 2011. You can also find Arm-designed mobile GPUs like the Mali-G52 from 2018, and the Mali-T860 from 2016.

Better Contemporary Comparisons

Those old low-end iGPUs and Arm GPUs might not be very good comparisons either, for people living in 2022 and wondering about the Zhaoxin Glenfly Arise 1020 GPU's performance compared to contemporary PCs. So, looking at something more modern and relatable to PC enthusiasts, we see that the Zhaoxin GPU is about 20x slower than the AMD Radeon Vega 11 graphics, which were integrated into the AMD Ryzen 5 2400G and Ryzen 5 3400G APUs. If that isn't a mainstream enough reference, the legendary GeForce GTX 1060 6GB was about 60 times faster than the Zhaoxin Glenfly Arise 1020 GPU in these tests.

Is the Zhaoxin Glenfly Arise 1020 GPU an Entry Level Member of the Family?

Of course, the "Glenfly Arise 1020" GPU might not be Zhaoxin's best option. The '1020' could be a low-power minimum spec member of the GPU family. Last June we reported on some information and images which provided some insight into a Zhaoxin GlenFly discrete GPU at sub-75W that was spotted running the Unigine Heaven benchmark (a 2009 DX11 test). We don't know if that was the same GPU as the one in the Geekbench OpenCL tests, but it appeared to be more capable.

In the Unigine Heaven tests shared last year, the Chinese graphics card managed 32fps on an FHD monitor. Unfortunately, we don't know the benchmark quality settings used, or if it was upscaling to 1080p or running at native resolution.

A perhaps more likely problem is that OpenCL testing with Geekbench can be highly variable, particularly with early drivers. Using it to gauge GPU performance can lead to very skewed results, even with comparisons between GPUs using the same architecture. Properly tuned drivers might boost performance an order of magnitude or more, but without direct access to the hardware and drivers, this is all we have for now. We'll withhold any final judgement until hardware and drivers become publicly available, and consider this early look more of a proof of existence than a reliable measurement of potential performance.

Update 05/19 5:05 pm PT: A representative from Zhaoxin has confirmed to Tom's Hardware that the company only focuses on business in China. Zhaoxin has no intention thus far to sell its products to other countries. Zhaoxin has also sent out announcements to its partners to emphasize the company's policy and to ensure that they understand Zhaoxin's policy.

Original story:

Now that AMD and Intel do not officially sell their processors in Russia, the country has to find replacements to not find itself in the stone age one day. This week Dannie, a motherboard maker with offices in Russia, China, Lithuania, and Turkey, introduced its new motherboard based on Zhaoxin's system-on-chip. While this chip can barely compete against modern CPUs from AMD and Intel, it is compatible with the vast majority of apps and OSes, so it can indeed replace these processors.

Dannie's MBX-Z60A micro-ATX motherboard features Zhaoxin's eight-core KaiXian KX-6640MA system-on-chip featuring the LuJiaZui microarchitecture, a 4MB of L2 cache, and a frequency of 2.10 GHz – 2.70 GHz. In addition, the SoC features a built-in graphics processor, 16 PCIe 3.0 lanes, a built-on SATA controller, an integrated USB 3.0 controller, and other essential I/O technologies. Since the chip caters to desktops and laptops, its maximum thermal design power does not exceed 25W.

As for specifications of the MBX-Z60A motherboard, it carries two memory slots for DDR4 memory and has two PCIe x16 slots for add-in-boards, one M.2-2280 slot for SSDs, an M.2-2230 slot for Wi-Fi/Bluetooth adapters, and three SATA ports, reports Habr. External I/O connections are concerned; the platform has USB ports, display outputs (a DisplayPort, an HDMI, and a VGA/D-Sub), a GbE connector, 3.5-mm audio input/output, and even PS/2 ports.

From a performance point of view, the KaiXian KX-6640MA is slower than the KaiXian KX-U6780A processor that sits higher in the product stack and which we tested a couple of years ago. Still, the lower-end one is probably cheaper too, which is essential for many clients, particularly those who only run office applications and do not need serious performance anyway.

One of the advantages that processors developed by Zhaoxin, a joint venture between Via Technologies and the Shanghai Municipal Government, have over other CPUs designed in China because they feature the x86 instruction set architecture. Therefore, they are compatible with dozens of operating systems (including Windows) and tens of thousands of apps. As a result, while processors developed by Zhaoxin are tangibly slower when compared to CPUs from AMD and Intel, they have still been deemed an alternative because of compatibility.

In fact, in the recent quarters, we saw the growing adoption of Zhaoxin's CPUs by international brands amid a tight supply of low-end SoCs by AMD and Intel. Qnap introduced a NAS based on a Zhaoxin CPU about a year ago. Then Lenovo chose one of such chips for its thin-and-light machine aimed at the Chinese market and designed for government and government-controlled companies that need to use software intended solely in China.

Will Dannie's MBX-Z60A motherboard relies on Zhaoxin's eight-core KaiXian KX-6640MA SoC to help Russia reduce the consequences of international sanctions and lack of AMD and Intel CPUs? For some applications, probably yes. Dannie can produce 'tens of thousands of motherboards per month,' which should be enough to meet demand from various government institutions for a while. For Zhaoxin, additional sales are a benefit. But for any performance-demanding workloads, Zhaoxin's SoCs will not be able to replace CPUs from AMD and Intel.

The latest version of the CPU-Z hardware diagnostic and benchmarking software suite adds support for multiple yet-to-be-released processors and graphics processors, including those from AMD, Intel, and even the first discrete GPU from Chinese CPU developer Zhaoxin.  

Support by CPU-Z shows that certain CPUs and GPUs are indeed incoming; their respective vendors have supplied developers of the program with the information required to enable the detection of appropriate hardware. Typically, support by utilities like CPU-Z and HWInfo is an indicator that the development of certain upcoming products is near the finish line. 

Among the pieces of hardware now supported by CPU-Z version 2.01 for Windows are the following parts:

• AMD Rembrandt and Raphael APUs (RDNA 2).
• AMD Mendocino APU (Zen 2 + RDNA 2).
• Preliminary support for Intel Arc 3/5/7 (DG2).
• Preliminary support for Intel Raptor Lake (13th gen Core).
• Zhaoxin's Glenfly Arise-GT10C0 GPU.

AMD's Ryzen 6000-series 'Rembrandt' APUs set to use Zen 3+ cores are designed primarily for notebooks and are expected to hit the market shortly. By contrast, AMD's Ryzen 7000-series 'Raphael' are the company's all-new processors featuring the Zen 4 microarchitecture, are aimed at desktops, and are scheduled to arrive later this year. AMD's Mendocino is a somewhat less well-known name, but this one is a Zen 2-based mobile APU with an RDNA 2-powered integrated GPU targeted at gamers. Interestingly, readers with a good memory might remember Intel using the Mendocino codename for its 2nd Generation Celeron processors in 1998 – 1999. 

Intel's Arc Alchemist graphics processors are due to be released in high volumes in Q2 – Q3 for both laptops and desktops, so it is about time for Intel to add its support to diagnostic software. Meanwhile, Raptor Lake is expected to be released in Q4, yet Intel's partners are already testing this CPU, so they need CPU-Z. 

Unfortunately, we do not know much about Zhaoxin's GlenFly Aris-GT10C0 graphics chip. However, we understand that this part is a DirectX 11.1 or DirectX 12-capable GPU made using TSMC's 28nm fabrication process and featuring a 70W thermal design power. Perhaps, Zhaoxin is starting to ship this chip to interested parties and needs to add its support to CPU-Z. 

In addition, the latest version of CPU-Z supports numerous already existing or shortly incoming processors:

• Intel Core i9-12900T, Core i5-12600T (35W).
• Intel Atom x6427FE, x6425RE, x6425E, x6414RE, x6413E, x6212RE, x6211E, x6200FE (Elkhart Lake, FCBGA1493).
• Intel Pentium J6425, N6415 (Elkhart Lake, FCBGA1493).
• Intel Celeron J6413, N6211 (Elkhart Lake, FCBGA1493).
• AMD Ryzen 7 5700X, Ryzen 5 5600/5500.
• AMD Ryzen 3 5300GE, Ryzen 3 PRO 5350GE, Ryzen 5 PRO 5650GE, Ryzen 7 PRO 5750GE (Cezanne).
• AMD Ryzen 9 6980HX, 6900HX, Ryzen 7 6800H, Ryzen 5 6600H (45W).
• AMD Ryzen 9 6980HS, 6900HS, Ryzen 7 6800HS, Ryzen 5 6600HS (35W).
• AMD Ryzen 7 6800U, Ryzen 5 6600U (15-28W).
• AMD Ryzen 7 5825U, Ryzen 5 5625U, Ryzen 3 5425U (15W).
• AMD Radeon RX 6850M XT GPU (Navi 22).
• AMD RX 6800S, RX 6700S, RX 6650M, RX 6650M XT GPUs (Navi 23).
• NVIDIA GeForce RTX 3090 Ti GPU (GA102-350, 450W).
• Improved information accuracy when core isolation is enabled.
• Improved validation process for high clock submissions (>6GHz).

The Fenghua No.1 GPU recently passed an important milestone for mass adoption in China. The GPU, created by Xindong Technology and Innosilicon, was initially announced in November 2021. We received some interesting but not entirely revealing performance indicators a month later. This month the GPU was certified for running stably and offering "outstanding performance on the Tongxin UOS operating system," reports ITHome.

Tongxin UOS is quite important in China as it was developed as part of a government initiative to usurp Windows. It is based upon Debian Linux and is adapted to support the latest homegrown silicon. So the Fenghua No.1 GPU certification is a significant step, alongside the implementation of support for Chinese-designed semiconductor products such as Zhaoxin CPUs and GPUs, processors from the likes of Sunway and Loongson.

The GPU sounds like it is definitely up to par for providing a 'Chinese' solution for graphics acceleration and other GPU tasks. We put the provenance in quotes as our previous investigations indicated that Fenghua No. 1 is based upon Imagination Technologies' PowerVR architecture, though we can't be 100% certain about this aspect of the GPU at this time.

The new GPU completed "comprehensive in-depth adaptation certification" in Tongxin UOS. Moreover, it could be China's first domestic " 4K-level high-definition desktop and high-performance server-level graphics card," to have been certified. It is claimed to exhibit "excellent performance, power consumption, reliability and compatibility, and fully supports mainstream graphics frameworks such as OpenGL/ OpenGL ES/ OpenCL/ Vulkan to achieve performance breakthroughs," says the Chinese source publication.

We hope to see systems featuring the new GPU available for sale soon, and some third-party benchmarks give us insight into the nature of this desktop GPU.

Before signing off, it would be amiss not to remind readers of the specs of the first Fenghua No. 1 graphics cards. In brief, we hear that the PowerVR architecture GPU is built on a 12nm process in China, and it supports up to 19Gbps GDDR6/6X. SKUs are being configured with 4, 8 and 16GB of VRAM, and outputs include HDMI 2.1, DP 1.4, and VGA with multiple monitor support. One of our December reports on the Fenghua No.1 GPU indicated that there would be a version of the card with dual GPUs, effectively doubling up all the specs such as FP32 and INT8 performance.

More GPU rivals are always welcome, and it would be a remarkable chain of events if the Fenghua No. 1 becomes popular in China. However, we aren't sure we need another company aiming for the near-RTX 3060 space.

Rather than wait for China to rescue the affordable GPU market, we already have hopes of a significant correction coming, as GPU pricing seems to be taking a nose-dive. While you wait for your own GPU pricing sweet spot to be reached, check out our Best Graphics Cards for Gaming in 2022 guide.

Lenovo has introduced its new ultra-portable notebook based on Zhaoxin's KX-6600MA4 processor and leverages a custom Lenovo system-on-chip (SoC). The Kaitian N7 resembles the recently announced Yoga Slim 7i laptop and has numerous features expected from products in this class. 

The Lenovo Kaitian N7 is powered by a quad-core Zhaoxin KaiXian KX-6600MA4 CPU based on the LuJiaZui microarchitecture made using TSMC's 16 nm fabrication technology. The processor is paired with 16GB of DDR4 memory and the system is equipped with a 512GB NVMe-compliant SSD with a PCIe interface. It is noteworthy that the chip appears to run pretty hot as it is cooled using a dual-fan, dual heat pipe cooling system. As for software, the laptop runs a custom OS developed in China and naturally comes with a BIOS designed in Tianxia.  

Having tested an eight-core Zhaoxin KaiXian KX-U6780A processor, we doubt that the KX-6600MA4 is a performance champion. However, if Lenovo wanted to build a laptop equipped solely with hardware and software designed in China, it probably did not have much choice.  

The Kaitian N7 comes in a 13.3-inch chassis and is equipped with a 14-inch display with thin bezels and a 2.2K resolution, reports eZone. Like other machines of this class, it weighs around 1.29 kilograms measures up to 14.6 mm thick. As for connectivity, it includes Wi-Fi 6 + Bluetooth 5, HDMI output, two USB-C ports, a USB Type-A connector, and a TRRS audio connector for headsets. For added biometric security, it comes with a fingerprint reader. 

Meanwhile, the manufacturer equipped the Kaitian N7 with a custom Lenovo SoC that handles system-specific things like light sensing, cover opening, fast charging, power bank mode, power-saving control and mute control, reports ApoTheTech. In addition, the SoC offloads some tasks from the OS and ensures optimal work of all components. 

Unfortunately, even a powerful controller cannot compensate for inefficient CPU and the OS. The operating system boots in 14 seconds, which is pretty long by modern standards. Furthermore, even when equipped with a 61 Wh battery, the Kaitian N7 can only work for up to six hours. 

There is no word about the pricing or availability window of Lenovo's Kaitian N7, but we would expect it to hit the Chinese market in the coming weeks or months. Meanwhile, it does not look like Lenovo will bring this model to other markets. 

This is not the first Lenovo system to use a Zhaoxin processor. Late last year, the company introduced a thin-client/UCFF desktop PC based on a Zhaoxin chip.

News outlet Nikkei reported that China plans to create a special committee to collaborate with big-name foreign chipmakers, such as Intel and AMD. With this new move, China aims to establish a domestic chip supply chain to circumvent the sanctions that the U.S. has imposed on the country.

The organization is called the "cross-border semiconductor work committee." If Nikkei's information is accurate, China may launch the committee in the first half of this year. The Ministry of Commerce will oversee this special committee in conjunction with the Ministry of Industry and Information Technology.

According to Nikkei's sources, the committee's objective is to fortify the relationship between Chinese and foreign companies. The end game is to build a chip supply chain by securing semiconductor technologies from Japan, Europe, and the U.S. In addition, the committee wants to woo overseas companies into establishing development and manufacturing sites in China with the promise of cooperation with the local governments and funding.

Nikkei's acquired documents allegedly revealed that China targets Intel, AMD, and Infineon Technologies of Germany. In addition, an industrial group from the Netherlands, including ASML, is also on the list. The publication's sources claimed that a few companies had expressed their desire to jump on China's initiative. However, Intel and the Dutch group refused to comment when Nikkei contacted them.

It's not the first time China has set out to forge a partnership with a foreign chipmaker. For example, AMD had struck a joint venture with China via the Tianjin Haiguang Advanced Technology Investment Co. Ltd. (THATIC) to license some of its x86 and SoC IP. Then there's also the emergence of Zhaoxin, a joint venture between Via Technologies and the Chinese government, to produce x86 chips.

Semiconductor Manufacturing International Co. (SMIC), China's number one chipmaker, Advanced Micro-Fabrication Equipment Inc. China (AMEC), a semiconductor equipment maker and smartphone giant Xiaomi will reportedly participate in the program. The other participants include Tsinghua University, Peking University, and the Chinese Academy of Sciences.

In 2015, China announced the "Made in China 2025" $20 billion campaign to bolster the domestic semiconductor industry. The plan was to achieve 70% tech self-sufficiency by 2025. However, things aren't going as planned, according to IC Insights. The U.S. firm estimated that China had reached 16% in 2020, and even China's optimistic numbers had it at 30%.

The U.S. has already blocked the major Chinese players, including Huawei and SMIC, from obtaining U.S.-designed equipment, materials, and software. In addition, Chinese processor developers Phytium and Sunway and several quantum computing entities are also on the blacklist.

China's latest act could serve as a stepping stone for achieving its goals. However, other countries are also offering subsidies and incentives to attract foreign investment in the semiconductor industry so that China will have stiff competition. Additionally, IP theft continues to be a concern, such as a Micron - UMC fiasco, so foreign companies will be reluctant to bring their technology over to Chinese soil.

Via Technologies sold its Centaur Technology x86 CPU development team to Intel in early November and the company is now auctioning off some of the equipment from Centaur's site in Austin, Texas. Essentially, this means that the company is shutting down the site and ending chip development operations in the U.S. 

"20,000 square feet of lab space, featuring over 1200 computer test stations and thousands of components," reads the auction description at Silicon Valley Disposition (via Andreas Schilling). "Servers, electronic test & measurement, Cadence supercomputer, microscopes, burn-in chambers, wafer probers, facility support, and much more." 

Selling off lab space, chip development, testing and debugging equipment means that neither Via Technologies nor Intel intend to keep the chip development site. For Intel (which would have to buy tools from Via separately as it only paid for developers and probably some IP), it makes more sense to add developers from Centaur to its existing teams rather than to keep them as a separate entity. For Via, it does not make any sense to keep expensive chip development in the USA.  

Centaur Technology was (and technically still is) a developer of x86 processors based in Austin, Texas. After becoming a subsidiary of Via Technologies in 1999, the company developed a number of microarchitectures, which later were used for CPUs designed by Zhaoxin (a joint venture between Via Technologies and the Shanghai Municipal Government), but never gained any tangible market share. 

At some point, Via understood that it might be lucrative to sell off its Centaur development team (for $125 million) and preserve its right to develop/build/sell x86 processors, retain its CPU-related patents and keep some engineers. Meanwhile, it does not look like it will continue to design CPUs in the U.S., but rather focus on its Zhaoxin joint venture. 

When Xindong/Innosilicon introduced its Fenghua 1 (Fantasy 1) discrete graphics processing unit (GPU) in mid-November, the company unveiled a rather impressive set of features for the GPU. This week, Innosilicon published performance specifications for the chip. While the Fenghua 1 cannot compete against the fastest offerings in our best graphics cards list, and likely lands quite a way down from the top of the GPU benchmarks hierarchy, it aims to provide comparable performance modern mid-range GPUs from AMD and Nvidia.

The Xindong/Innosilicon's Fenghua 1 (Fantasy 1) GPU is reportedly based on Imagination Technologies' PowerVR architecture. While we can only speculate about the exact microarchitecture used, the graphics processors are quite capable as they support contemporary application programming interfaces for graphics and compute, including DirectX, Vulkan, OpenGL, OpenCL, OpenGL ES, Caffe 1.0, TensorFlow 1.1.2, and ONNX. Since this is a PowerVR-based GPU, it comes with Android, Linux, and Windows software stacks. The GPU is made using a 12nm fabrication process.  

There are two Fantasy 1 graphics cards at present: the single-chip Type A and the dual-chip Type B board. The single-chip Fantasy 1 boasts compute performance of around 5 FP32 TFLOPS for graphics and 25 INT8 TOPS for AI/ML, which is comparable to the (theoretical) performance of Nvidia's GeForce RTX 2060 GPU. Meanwhile, the dual-chip Fantasy 1 doubles that with FP32 compute performance of approximately 10 FP32 TFLOPS and 50 INT8 TOPS for AI/ML. This is slightly higher compared to performance numbers offered by Nvidia's GeForce RTX 3060.

Note that those figures represent theoretical compute performance, and as we've seen from AMD and Nvidia in the past, scaling via dual-GPU designs can double compute while creating a host of other hurdles when it comes to real-time graphics processing. Besides driver support, dual GPUs typically have to contain two copies of everything in memory — one copy for each GPU and its direct attached VRAM. Syncing data for frames between the GPUs further complicates things, often requiring game-specific support.

Xindong/Innosilicon's Fantasy 1 Graphics Cards

The Fantasy 1 Type A card can be equipped with 4GB, 8GB or 16GB of GDDR6 or GDDR6X memory, and the Fantasy 1 Type B card should double that with support for up to 32GB of memory. Both graphics cards support a PCIe Gen4 host interface as well as DisplayPort 1.4, eDP 1.4, and HDMI 2.1 interfaces.

While the raw compute performance of the Fantasy 1 Type B looks impressive, it likely won't directly translate over to games. The developer never mentions multi-GPU technologies akin to AMD's CrossFire or Nvidia's SLI — neither of which are properly supported on modern GPUs in contemporary games. It looks like the Fantasy 1 Type B is aimed mostly at datacenters. Keeping in mind that the GPU fully supports GPU virtualization as well as PCIe SR-IOV, GPU computing in datacenters and virtual desktop infrastructure (VDI) are among the applications it was designed for.

Perhaps the most surprising part about the Fantasy 1 GPU is power consumption. The typical power consumption of one Fantasy 1 Type A card in 'a multi-channel cloud environment' is supposedly about 50W, which is considerably lower than the TDP of Nvidia's GeForce RTX 2060 (TU106). Obviously, Xindong/Innosilicon's claims have to be independently tested, but they do sound impressive.

Xindong/Innosilicon is currently sampling its Fantasy 1 graphics cards with interested parties. However, it is unclear when these GPUs will be available commercially. Besides availability and performance, we would also need pricing details to determine whether they can compete with the likes of AMD Radeon, Nvidia GeForce, and Intel Arc — or more likely, against the datacenter variants of those GPUs.

FinalWire just released the latest 6.60 version of AIDA64, its diagnostic and benchmarking software. The latest version brings in preliminary support for Intel's next-generation Raptor Lake processors as well as Nvidia's yet-to-be-announced top-of-the-range graphics processor for mobile PCs, essentially confirming the green company's plans to launch the rumored GeForce RTX 3080 Ti laptop GPU. We don't have detailed specifications for either of those yet, but assuming you manage to get your hands on an early sample, AIDA64 will apparently know what it is.

Enhanced CPU Support 

FinalWire's AIDA64 v6.60 update introduces multiple innovations, including AVX-512 and AVX2 accelerated benchmarks for Intel's Alder Lake (which do not officially support AVX-512). It also has support for Raptor Lake processors and preliminary support for the next generation Meteor Lake CPUs that are due to land about two years from now. In addition, the new version of the tool includes an AVX accelerated benchmark for various Zhaoxin and Hygon C86 Mukti/Dhyana processors.

But while it is nice to see enhanced support for Intel's Alder Lake and Raptor Lake CPUs as well as Chinese CPUs, the confirmation of Nvidia's GeForce RTX 3080 Ti for laptops looks particularly interesting. 

Nvidia's GA102 and RTX 3080 Ti Coming to Laptops

It's been rumored that Nvidia plans to introduce new Ti-branded GeForce RTX GPUs for desktops with more VRAM. Now it looks as though there will also be some new GeForce RTX 30-series Ti products for laptops in the near future. According to the press release, what's even more exciting is that a new GeForce RTX 3080 Ti will be coming to laptops.

The notes state, "GPU details for nVIDIA GeForce RTX 2060 12GB and GeForce RTX 3080 Ti Laptop." What they don't say is whether that GPU will be based on Nvidia's GA104 GPU (with up to 6144 CUDA cores) or if it will use GA102 GPU (with up to 10752 CUDA cores). However, Nvidia already has the GA104-powered GeForce RTX 3080 for laptops with 6144 CUDA cores, so using the same GPU with higher clocks or more VRAM may not make a lot of sense. What seems more likely is that Nvidia will use the considerably larger and more power hungry GA102 in a configuration appropriate for high-end gaming notebooks.

At present, Nvidia's GA102 is used exclusively for desktop graphics cards: the GeForce RTX 3090, GeForce RTX 3080 Ti, and GeForce RTX 3080. While those all require 320W or more power, Nvidia could drop the clocks and maybe even switch to GDDR6 memory (instead of GDDR6X) to fit the chip into a mobile profile. The biggest notebooks top out at around 150W of power for the GPU, and such a configuration should be possible. It would also unlock performance levels previously not available for mobile PCs. Of course that would impact battery life, but for desktop replacement notebooks that shouldn't matter too much.

Nvidia yet has to confirm its plans to launch the GeForce RTX 3080 Ti for laptops, so take this information with a grain of salt. However, AIDA64 isn't normally prone to leaking false information, so we strongly suspect we'll see the GPU in notebooks in the coming months.

After the fall of Matrox Graphics, S3 Graphics, and XGI in the mid-2000s, no companies have competently challenged AMD and Nvidia in the field of discrete graphics cards for PCs. But with the rise of special-purpose-datacenter GPUs, mining GPUs, and gaming, numerous rivals for the two mighty GPU designers emerged. For example, this week Chinese company Xindong Technology announced its Fenghua GPU aimed at gaming PCs and datacenters. 

Xindong may not be a household name, but it seems to be a very ambitious GPU developer from China that wants to compete against considerably bigger players. The company recently successfully tested its first Fenghua No. 1 graphics processing unit developed entirely in-house and then designed by Innosilicon, a contract chip designer. 

The Fenghua No. 1 seems to be a multi-chiplet GPU with a GDDR6X-based memory subsystem that supports all modern (and legacy) application programming interfaces for graphics and compute, including DirectX, Vulkan, OpenGL, OpenCL, and even OpenGL ES. It can also work with Android, Linux, and Windows. The chip uses a PCIe Gen4 interface to connect to the host and can output graphics using DisplayPort, eDP 1.4, and HDMI 2.1 interfaces, just like today's best graphics cards. 

Xindong yet has to disclose performance levels one can expect from its GPU, but it says it plans to use its GPU for desktops, cloud games, cloud mobile games, Xinchuang desktop, and workstations. This positioning seems very self-confident since it is relatively easy to make GPUs process select workloads in the datacenter. Still, it is much harder to make them work flawlessly across dozens of games and applications used by gamers at home.  

Furthermore, multi-GPU rendering (which is probably what Xingdong means by implementing a multi-chiplet GPU) is very tricky with today's games and rendering techniques. Neither AMD nor Nvidia support CrossFireX and SLI technologies on their most recent gaming GPUs. 

At present, we have no idea whether Xindong's Fenghua GPU will be a viable competitor for AMD's Radeon and Nvidia's GeForce. Still, at least there is one more rival (in addition to Intel, Phytium, Zhaoxin, and a number of others that design datacenter and HPC GPUs) for these two graphics giants.

It is important to note that the Fenghua No. 1 is far from the first custom GPU designed by contract chip developer Innosilicon. Last year the company demonstrated a datacenter GPU, and we also know that the company has some Ethereum accelerators. 

Via Technologies announced plans to offload part of its Centaur Technology x86 CPU subsidiary to Intel for $125 million. The Taipei, Taiwan-based company will retain its right to develop, build, and sell x86 processors, its CPU-related patents, and some engineers. However, it will allow some of its engineers from Austin, Texas, to join Intel. As a result, the CPU giant will get another team of experienced engineers.

"Intel will recruit some of Centaur's employees to join Intel with certain covenants from the Company for Intel's recruitment," a statement by Via Technologies with TWSE reads. "As consideration, Intel will pay Centaur $125 million. […] Closing of the transaction is contingent on certain conditions in the agreement to be satisfied. Payment will be made in full upon closing."

The announcement of the deal (first noticed by AnandTech) raises more questions than it gives answers. The agreement involves the personal wishes of Centaur employees, but does not include any fixed assets or intangible assets, said Chen Baohui, CFO of Via Technologies, at a press conference held in Taipei, Taiwan, reports UDN. He also stressed that the transaction could not be considered a sale of Centaur or the team (even though Centaur's website has been taken down). In fact, the company did not even specify the actual matter of the transaction or its timing. 

Centaur Technology, a 100% Via Technologies subsidiary, has an interesting history. The company was founded in 1995 to build cheap x86 CPUs and got its funding from Integrated Device Technology (IDT), which had its own fabs back then. IDT released the first Centaur-designed processor under the WinChip brand in 1997 and the second WinChip CPU in 1998. The chips featured an in-order execution microarchitecture with a low-performance floating point unit (FPU). However, IDT's WinChips never became popular because Intel had CPUs with out-of-order microarchitectures and Intel's chips also featured a high-performance FPU. To that end, IDT canceled the third iteration of WinChip and sold Centaur Technology to Via Technologies in 1999.  

Later on, Centaur designed Via C3, Via C7, and Via Nano/Isaiah (a 64-bit superscalar out-of-order) processors sold by Via. After Isaiah, Centaur Technology developed its 'Haswell-like' CNS core that supported AVX-512instructions (using a 256-bit SIMD) and was meant to be used in the CHA processor with a built-in AI accelerator. Meanwhile, CHA has never materialized.  

One interesting thing to note is that the Nano/Isaiah microarchitecture was adopted and refined by Zhaoxin, a joint venture between VIA Technologies and the Shanghai Municipal Government. Zhaoxin continues to develop x86 CPUs and has rather solid plans for the future. It can access advanced manufacturing capacities and therefore can release relatively competitive CPUs, so Via's presence on x86 market will certainly continue, albeit under a different brand. 

This is not the first time that Intel has acquired assets from Via Technologies. Back in 2015, the company bought some assets of Via Telecom (a developer of entry-level smartphone SoCs, baseband processors, and modems).

A small-form-factor PC featuring the Chinese-made Zhaoxin KX-6640MA processor, a reminder that AMD and Intel aren’t the only game in town any more, has appeared for sale (and is sold out) on Shopee Indonesia, where it was spotted by CNX Software.

The KX-6640MA is a quad-core X86 processor that runs at 2.1 GHz base clock and 2.6 GHz boost clock with a 4MB cache, a 25W TDP and C-960 integrated graphics that support 4K output. Alongside this sits 8GB of DDR4, upgradable to 64GB, a pair of NVMe M.2 slots, one with a 256GB SSD already in it, and a 1TB HDD. 

Video is handled by a pair of HDMI 2.0 ports, and there are two gigabit ethernet sockets too, along with WI-Fi 5 and Bluetooth 4. There are six USB 3.0 ports, all of them type-A, plus headphones and mic ports and a power brick. Compatible operating systems aren’t discussed, but there’s something that looks a lot like Windows 10 in the shopping site's images. The price converts to around $270.

Zhaoxin processors haven't been able to compete with the best CPUs, sitting below i3s and Pentiums in our testing. This doesn’t mean it’s not good enough, if the price is right, to carve out a niche for itself. And of course, it was probably made with the home market in mind. 

And while it’s interesting to see something like this for sale outside the Chinese market, if you’re thinking about ordering one when they come back into stock you might want to pause: CNX reports that requests for a review sample were rebuffed by Beelink as “they had stopped production due to a lack of supply for the CPU”. However, we found this Zhaoxin KX-6640MA-based mini PC for sale for $260, and apparently in stock, on Newegg, for anyone wanting to break out of the Intel / AMD hegemony.

Qnap has quietly introduced its six-bay TVS-675 NAS powered by an eight-core Chinese x86 CPU designed by Zhaoxin, a joint venture between Via Technologies and the Shanghai Municipal Government. The NAS for small businesses (SMBs) is one of the first (if not the first) commercial devices set to be available in the U.S. and Europe, with the KaiXian KX-U6580 processor developed primarily for the Chinese market. 

Zhaoxin was established in 2013 to build x86 processors in China for the local market and cut reliance on processors designed by AMD and Intel, but retain compatibility with the vast library of x86 software. The initial generations of Zhaoxin's products weren't successful because of utterly low performance, and the latest KaiXian KX-6000-series is also not a performance champion in general-purpose tasks (according to our review). But it looks like Qnap found the eight-core 2.50 GHz KaiXian KX-U6580 (with Zhaoxin C-960 graphics) good enough for its entry-level NAS.  

Usually, Qnap's entry-level network area storage devices use Arm-based SoCs or Intel's Celeron SoCs powered by the company’s low-power architecture. NAS manage files, run light applications, or encode videos, so these inexpensive SoCs are good enough for their workloads. Zhaoxin's KaiXian KX-6000 can certainly run light apps, but as our review revealed, its media encoding performance is very low by today's standards, possibly because it lacks certain hardware encoders applicable to our test scenarios under Windows, whereas its general-purpose architecture is not exactly suited for such tasks. Keeping in mind that Qnap's TVS-675 NAS is aimed primarily at SMBs that hardly need it for media encoding, this drawback of the KaiXian KX-6000 probably won't affect users of this NAS. Furthermore, Qnap says that the SoC does support H.264/H.265 transcoding with its OS.

Speaking of the Qnap TVS-675, this NAS promises to provide SMBs high capacity, high performance, expandability, and rather fast network connectivity. The NAS has six hot-swappable 3.5-inch bays for high-capacity SATA HDDs, two SO-DIMM slots for DDR4 memory modules (the system ships with 8GB, but you can install up to 64GB), two M.2-2280 slots for SSDs supporting a PCIe 3.0 x1 or SATA interface, and two PCIe 3.0 x4 slots for SSDs or network cards. The unit also has two 2.5 GbE ports, two USB 3.2 Gen 1 Type-A connectors, two USB 3.2 Gen 2 Type-A ports, and one HDMI 2.0 output.  

Qnap equipped the TVS-675 with three fans to cool down the PC-oriented SoC, six HDDs, two SSDs, and expansion cards. The company claims that the fans produce 23 dB of noise (though it does not disclose how it measured this level). 

The Qnap TVS-675 ships with a choice of either QTS or the enterprise-grade ZFS QuTS operating system. QTS supports all the capabilities expected from an entry-level NAS, including, Qsync cross-platform file sharing (for Apple, Windows, IBM AIX, and Linux machines), Qtier caching technology that automatically moves frequently-used file/data to SSDs, snapshots, RAID modes, virtual JBOD, FTP server, HTTP access, and so on. The NAS also supports QNAP-certified first-party and third-party apps.

Qnap hasn't disclosed pricing yet, though this is a business-oriented machine with rich features and will be priced accordingly.

Zhaoxin, a joint venture between China-based Shanghai SASAC (which controls an 80% stake) and Taiwan-based CPU developer Via Technologies (which owns the remaining 20% stake). Last year it unveiled plans to launch a standalone graphics processor and it appears that Zhaoxin managed to produce one.

GlenFly Technology, a graphics subsidiary of Zhaoxin that is a formal assignee of Zhaoxin's graphics patents, recently published (thanks to @Loeschzwerg_3DC for the link) an image of a graphics card that is running the Unigine Heaven benchmark on a 22-inch Full-HD LCD at 32 frames per second (FPS). Considering the fact that GlenFly is not supposed to be running third-party hardware, we can assume that the board is indeed based on a Zhaoxin GPU.  

Last year Zhaoxin promised a standalone graphics processor supporting a DirectX 11.1 or DirectX 12-level feature set with a 70W thermal design power implemented using TSMC's 28 nm fabrication process. Such a GPU is not going to set performance records by today's standards, but it can certainly serve entry-level desktops. Until fairly recently, Via Technologies used S3 Graphics-developed GPU cores for basic processors, so using the same IP for something else is a possibility. 

The depicted card features a low-profile form-factor, does not have any auxiliary PCIe power connectors (i.e., its power consumption is below 75W supplied by a PCIe x16 slot), uses a modest cooler, and has an HDMI and a D-Sub display output. 

The Unigine Heaven is a 2009 benchmark that relies on DirectX 11-level capabilities and is designed primarily for stability testing. Modern graphics cards can easily hit hundreds of FPS in this test, but since we are not aware of GlenFly's test settings, we will refrain from making any conclusions about performance of the GPU.  

There is a claim based a post in a Chinese forum that the portrayed GPU is Zhaoxin's Ji Rui 3000 that has been cancelled, though we could not verify it. Keeping in mind that the image was published in 2021 and usually old benchmarks are run on GPUs during their bring up process, it might be too early to say that the chip has been cancelled. 

Neither Zhaoxin nor GlenFly Technology have ever formally announced a discrete GPU, so it is impossible to draw any conclusions about the graphics card pictured by GlenFly. Meanwhile, it is pretty evident that the two companies have actually built a standalone GPU for desktop PCs. 

The number of graphics processing units (GPU) developers grew to 19 last year as multiple companies are preparing for looming artificial intelligence (AI), high-performance computing (HPC), and edge computing megatrends, according to an observation by GraphicSpeak, a Jon Peddie Research (JPR) publication. The question is, how many of the started GPU projects will actually come to fruition? 

Dozens of companies used to develop discrete graphics processors (dGPUs) in the 1980s and the 1990s, but their number shrank to less than half a dozen by the early 2000s and to two (or three, depending on how you count) later in the decade. By the early 2010s, companies like Arm, Imagination Technologies, DMP, Vivante (now part of VeriSilicon), and Think Silicon (an Applied Materials company) focused purely on GPU IP development (except for DMP). In contrast, companies like Apple and Qualcomm started to design their own integrated graphics processors (iGPUs). 

The advent of compute on graphics processing units for AI, HPC, and other highly-parallel workloads, as well as the emergence of multiple new device categories, have spurred significant interest in graphics in recent years. This inspired China-based Jingjia Micro to develop and release the country's first discrete GPU in 2014 and Intel to return to discrete GPU development in 2017. 

By late 2019, the number of dGPU, iGPU, and GPU IP developers grew to 11, according to Graphic Speak. Meanwhile, if Tianshu Zhixin (established in 2018) with its Big Island GPGPU chip is included, it grows to 12.

But the interest in GPUs and the ongoing trade war between the U.S. and China inspired many other chip designers to start developing their own GPUs. At least eight new companies announced their GPU plans in 2020. Some of these companies are from China, and some others plan to adopt the open RISC-V architecture. The majority of new entrants are looking to build datacenter-grade solutions rather than consumer GPUs, and companies like Phytium will inevitably address client PCs with their graphics technology, too.

Developing GPUs is a very tricky task. These processors tend to be extremely complex in terms of transistor count, and they also feature distinct subsystems and a multifaceted software stack. It remains to be seen how many of these new GPU projects will actually be a success. Still, it is pretty obvious that with Intel and several new entrants, AMD and Nvidia will face considerably more competition in the next couple of years than they have in the last couple of decades.

Via Technologies and Shanghai Zhaoxin Semiconductor Co., a joint venture between Via and Shanghai Municipal Government, have announced that they have entered into an agreement under which Via will transfer some of its IP to Zhaoxin. The intellectual property will enable Zhaoxin to continue the development of its x86 PC platforms. 

Once a successful supplier of chipsets and low-power CPUs, Via Technologies refocused from PC products to various emerging applications for industrial, transportation, and smart city markets in recent years. Yet Via Group did not abandon PCs completely as it still has some low-end CPUs and chipsets and also holds a 14.75% stake in Shanghai Zhaoxin, a developer of CPUs and chipsets.  

Because Via Group has an x86 license it obtained back in 1999 (through its acquisitions of Cyrix and Centaur Technology) and co-owns Zhaoxin, the latter can develop x86 CPUs. Originally, Zhaoxin designed CPUs based on the Via Isaiah microarchitecture, but eventually, it introduced its own microarchitectures (which are likely evolutions of the Isaiah). Apparently, now Zhaxin needs more intellectual property from Via Group to keep developing its platforms. According to Via Technologies's filings with the Taiwan Stock Exchange, the company plans to transfer certain CPU and chipsets-related IP to Zhaoxin. The transaction does not include patents, and Via will be able to keep using its IP going forward.   

"The company and its 100% owned subsidiaries Viabase and Viatech sell some chipset products related intellectual property rights (excluding patent rights) to Shanghai Zhaoxin Semiconductor in which Via has indirect shareholding of 14.75%," a statement by Via Technologies reads. […] In addition, the subsidiary Viabase sold some CPU products related intellectual property rights (excluding patent rights) to Shanghai Zhaoxin Semiconductor." 

Neither Via Technologies nor Zhaoxin disclosed which IP is to be transferred, keeping in mind that Zhaoxin already has access to Via's most advanced Isaiah x86 microarchitecture. Typically, under the terms of agreement with Intel, x86 licensees (which includes such companies as AMD and Via) cannot sell the license but can enable their subsidiaries to develop and produce x86 processors. Meanwhile, it is possible to obtain an x86 license by acquiring one of the existing x86 licensees. That said, while Via Technologies will indeed provide certain x86-related IP to its subsidiaries, this is not a transfer of an x86 license. As for chipset technologies, Zhaoxin has used graphics cores developed by Via's subsidiary S3 Graphics for years.

"After the transaction, the company and its subsidiaries can still continue their current business," the statement added. "The transaction is a kind of asset activation, which can increase revenue without affecting future business development." 

Via Group will receive payments from Zhaoxin in two installments in 2021. Total proceedings are expected to be about $197 million after all the taxes are paid. 

Sources: Taiwan Stock Exchange, DigiTimes, Economic Daily

China-based Zhaoxin is best known for its homegrown x86 processors, but according to a report from cnTechPost, the company recently teased a new low-cost 70W discrete GPU in a video posted to its newly-redesigned website. The video has since been removed, but the website grabbed a snip of the company's roadmap, which lists the new stand-alone 70W GPU that will be fabbed using TSMC's 28nm process. 

Zhaoxin is best known for its CPUs, like the KaiXin CPU we recently tested, that are designed entirely in China. The company is backed by Chinese government-controlled interests, with the state-owned Shanghai SASAC owning 80% of the company while VIA, which provides the x86 license, owns 20% of the company. 

The company didn't reveal many details of the GPU, though we know it will slot into a relatively low 70W TDP and come with TSMC's 28nm process. The GPU will purportedly come to market later this year. The company's use of the 28nm process obviously lags the industry norm, with stalwarts like AMD and Nvidia already moving on to TSMC's 7nm tech, but that could be due to the threat of potential US sanctions that could prevent TSMC from providing 16nm and smaller processes to its China customers on the grounds that the newer processes are based on US IP.

It's unclear how Zhaoxin would be able to produce a full-fledged and competitive GPU without infringing on graphics IP from AMD, Nvidia, or Intel. However, as we covered in our KaiXan review, Zhaoxin's parent company VIA has access (via a previous acquisition) to graphics IP from S3 graphics. 

Zhaoxin also has integrated graphics units for its CPU platform, though the GPUs are embedded in the chipset. Our tests found these graphics units to be woefully inadequate compared to the integrated graphics on competing CPUs, even Intel's lackluster UHD 630, but extending that architecture out to a larger die with a higher TDP envelope would obviously improve performance. However, we wouldn't expect the new graphics cards to eclipse even older AMD and Nvidia graphics cards. 

Zhaoxin's existing iGPUs support DX11, OpenCL 1.1, and OpenGL 3.2 and features hardware-accelerated video encoding and decoding, but details are scarce and GPU monitoring applications couldn't scrape any further details on the architectural components. The integrated graphics support DisplayPort, eDP, HDMI and VGA interfaces and can simultaneously output to two screens at a 4K resolution. 

Given the 70W TDP envelope, it's conceivable that these cards will be destined for low-cost desktop systems, thus providing China a possible alternative if the US-China tradewar intensifies. 

We've reached out to Zhaoxin and will update as necessary. 

Korean news outlet The Elec has received word from its industry sources that Huawei will be taking a crack at the server graphics card market this year. If true, Huawei will have its work cut out for the tech giant since the company will have to compete against veterans, such as Nvidia and AMD, and eventually, newcomer Intel.

According to the report, Huawei will setup up shop at Korea under the newly created Cloud and AI Business Group. Huawei Korea is presently comprised of the Carrier, Enterprise and Consumer divisions. The Cloud and AI Business Group will report directly to the Enterprise division.

Huawei has reportedly already begun building up its roster. The Elec believes that the corporation is trying to attract current and former talent from Nvidia to help pursue its goals. Huawei obviously has no lack of resources or infrastructure. With the right amount of talent around the company, Huawei could turn out to be a worthy rival.

Last year, Huawei launched the Ascend 910 AI chip that it claims is capable of delivering up to 256 TFLOPS of half-precision (FP16) performance, which would be twice as fast as Nvidia's Tesla V100, and to 512 TOPS of INT8 compute performance at 310W. Huawei's silicon is even one step ahead of Nvidia in terms of manufacturing process: The Ascend 910 is based on TSMC's N7+ process node, while the Tesla V100's GV100 die uses the 12nm process. Huawei's Ascend 910 would eventually make its way into the brand's own Atlas 300 PCIe 4.0 accelerator card.

Huawei's desire of stepping into the server graphics card market isn't a coincidence either. We suspect China's ambitious "3-5-2" initiative to replace all foreign hardware and software in public and government institutions by 2022 probably played a big role in Huawei's strategic move. China already has its domestic companies and other joint ventures producing homemade operating systems, motherboards, and processors, such as Zhaoxin's latest KaiXian KX-U6780A. 

Sourcing chips might prove to be a problem for Huawei as a recent report reveals that the U.S. is in preparation to prevent TSMC from doing business with the Chinese tech giant. If that happens, Huawei will probably have to find comfort in other arms, such as Samsung or China's own SMIC.

Graphics cards might be the last piece to China's self-sufficiency, and ultimately that's where Huawei comes in. If the company successfully produces a competitive server-class graphics card, Huawei might expand its offerings to the consumer market, too.

China-based Zhaoxin Semiconductor, which roughly translates to 'million core,' has fielded a new eight-core KaiXian KX-U6780A processor based on its mysterious LuJiaZui microarchitecture and fabbed on TSMC's 16nm FinFET process. Chips based on the architecture bring a new level of performance to China's roster of homegrown chips that span from desktop PCs to servers, but the KX-6000 series is destined for gaming rigs, PCs, and office machines. Today we have the chip in for testing on Zhaoxin's HX002EH1 demo board to determine if it can rank among our list of best gaming CPUs.

Zhaoxin definitely isn't a household name, but it's one of the very few companies that designs bespoke x86 processors. That means it competes with the likes of dominant chip producers AMD and Intel, which is surprising given the patent-protected x86 instruction set and strict licensing agreements that have long narrowed the field of x86 chip producers to the Intel, AMD, and VIA triumvirate. Details are scarce, but the fabless Zhaoxin consists of a joint venture between the Shanghai Municipal Government, which invested 80% of the company's initial capital, and Taiwanese VIA Technologies, which provided 20%. This partnership structure purportedly allows Chinese government-controlled interests to design x86 processors while staying within the legal boundaries of licensing agreements. 

The end result is a processor designed specifically for the China market that Zhaoxin says reaches the same level of performance as Intel's seventh-gen Core i5-7400. While that's certainly not overly impressive given that Intel's seventh-gen processors debuted in early 2017, it marks a big step forward for Zhaoxin, which claims its new LuJiaZui microarchitecture provides a 1.5X improvement in instructions per cycle (IPC) throughput over the preceding WuDaoKou architecture. It also comes with several new features, like Zhaoxin's first integrated graphics engine. The KX-6000 line is a step towards a larger goal: It serves as the intermediary processor before the company's next-gen 7nm chips that will come with DDR5 memory and the PCIe 4.0 interface as the company aims for process parity with the incumbents in 2021. 

But disrupting the existing duopoly is no easy task. Let's take a quick look at what we know about the Zhaoxin KX-6000 series, then dive into our test results. 

The Road to China-Designed Chips

China is by far the world's largest importer of processors, the majority of which are exported as finished products. However, the country has long sought independence from western influences on its economy. China only produces roughly 16% of the silicon it uses in-country, and only half of that production is controlled by Chinese interests. As such, China's reliance on western CPU production presents a strategic liability for the country's economy in the event of a blockade or tariffs, not to mention the implications of potentially nefarious backdoors built into the chips. It's also no secret that today's wars are won with the high-powered chips found in everything from fighter jets to battleships and submarines.

As such, China began an initiative to build its own processors back in the early 80's that eventually evolved into the "Made in China 2025" campaign that has the goal of producing 70% of the chips used in China by 2025. This initiative began long before the U.S.-China trade war unfolded, and the U.S. has long blocked China's attempts at gaining the know-how and equipment to fabricate chips through acquisitions and mergers, a situation that has only intensified recently. Chinese foundry SMIC and others continue to advance, but those foundry's will need solid chip designs when, and if, they can jump to leading-edge nodes. 

Undeterred by numerous blocked acquisitions and an inability to buy key tooling for its own fabs, China spread its tentacles to several public/private partnerships to foster indigenous designs. In 2016, AMD forged a partnership with Hygon, a company funded by China-controlled interests, to produce Dyhana processors based on AMD's Zen architecture. The resulting THATIC joint venture was eventually blacklisted by the U.S. in the name of national security as the trade war unfolded. 

Fueled by government investments and preferential treatment in the market, other native Chinese companies have joined the fray, but with non-x86 architectures. Huawei has developed the Kunpeng 920 Armv8 processors for its line of motherboards for desktop PC OEMs and Phytium Technologies is shipping ARM chips. That's important given the rise of Windows on ARM, and Chinese company Loongson also continues to advance its MIPS designs that rely on x86 emulation.

But there's no substitute for the ubiquity of a native x86-64 processor, and Zhaoxin has plenty of uptake. Both Lenovo and HP have fielded Zhaoxin-based systems for the Chinese market, with Lenovo offering its Kaitian desktop PCs, Zhaoyang FF03 laptops, and server products. HP also has a line of products, including its HP 268 Pro G1 MT that comes powered by the KX-U6780A processor. 

Having the support of these large OEMs, not to mention the constellation of smaller builders inside China, is important as the country now looks to enforce its new "3-5-2" edict that comes as a result of the trade war. This policy stipulates that all government and public institutions must switch to 100% China-native hardware and software by 2022, with 30% of foreign gear replaced in the first year, 50% replaced the next, and the remainder eliminated in the third year of the program. That means the Zhaoxin KX-U6780A stands a good chance of finding some measure of success relatively quickly. 

Zhaoxin KX-6000 Series Specifications

Zhaoxin isn't a newcomer to chip design, though. The company was founded in 2013 and its first CPU core hailed from Centaur, a company that VIA Technologies purchased in 1999. Zhaoxin's first-gen ZhangJiang cores are largely thought to be built on a lightly-modified version of Centaur's Isaiah (also called CN) microarchitecture. Zhaoxin followed that with its own WuDaoKau architecture, which represented a full-on redesign of Isaiah that has now evolved into today's x86-64 LuiJiaZui microarchitecture.

The first four iterations of the Zhoaxin chips were fabbed on HLMC's 28nm process, meaning they were entirely China-built processors. Zhoaxin's KX-6000 series moves to TSMC's 16nm process that's produced in Taiwan, and the next generation of chips will come with DDR5 and PCIe 4.0 built on a 7nm process. It's yet to be seen if Zhaoxin will use SMIC, a rising Chinese foundry with plans for 7nm, or TSMC for the KX-7000 series chips. The latter is far more likely. 

Unfortunately we know very little about the LuiJiaZui architecture. We do know that the processor has a superscalar out-of-order multi-issue architecture, but the company has not shared block diagrams or further details. The architecture supports the AVX (256-bit width) and SSE 4.2 instruction sets, along with VMX virtualization (compatible with Intel's VT-X). In the past, Zhoaxin's chips relied on an external graphics processor housed in its ZX-100S chipset, but the KX-6000 marks the debut of the company's own integrated graphics engine. 

The 70W chips come as a soldered-down HFCBGA package that measures 35mm x 35mm, so you'll have to buy a motherboard along with the chip (they aren't socketed). The KX-6000 series comes in both four- and eight-core variants. 

The flagship KX-U6880A comes with eight cores that can operate at up to 3.0 GHz, an increase of 1GHz over the previous-gen models. Our KX-U6780A model ticks at 2.7 GHz, and, like the flagship, comes with eight cores and eight threads. The KX-6000 series comes with eight 32KB slices of L1 data and instruction caches, and 8MB of L2 cache carved up into two 4MB slices. The L1 and L2 caches are 8-way set associative, and the chip doesn't have any L3 cache.

The chips are somewhat rudimentary in regards to two key technologies: Boost frequencies and simultaneous multi-threading (SMT). Unlike Intel and AMD, the KX-6000 chips have no SMT/Hyper-Threading, which leaves some performance on the table in threaded workloads. The chips also don't have boost capabilities, so if you select the Windows high-power or balanced power profile they operate at 2.7 GHz regardless of workload, electrical, or thermal conditions (aside from a normal temperature throttling mechanism for self-protection). That means better cooling will not unlock higher performance, and overclocking is also not on the menu.

Using the power saving feature does enable P-State transitions as shown in our boost testing above, but this is designed more for power saving as opposed to extending performance beyond the 70W power budget. Interestingly, the first-gen ZX-C chips did have boost capabilities, but this feature was apparently deprecated in newer versions of the architecture. This obviously has performance implications in lightly-threaded applications. The chip also supports C1-C4 C-States. 

Zhaoxin lists the chip with dual-channel DDR4-2666 support (no ECC), but our spartan development board doesn't allow manipulation of the memory frequency and timings, and it doesn't support XMP profiles. Those options might be enabled in shipping systems, but for now, the board merely uses the default SPD profile. That means you'll have to buy a pricey kit with a tailored profile for higher frequencies. Zhaoxin says it expects the chip to support DDR4-3200 in the future, but that is still under development.  

The processor exposes 16 lanes of PCIe 3.0 and has an integrated graphics engine with an undisclosed architecture (thought to be based on IP from VIA's partner, S3 Graphics) [EDIT: Zhaoxin later clarified that the graphics are a proprietary solution]. We do know it supports DX11, OpenCL 1.1, and OpenGL 3.2 and features hardware-accelerated video encoding and decoding, but details are scarce and GPU monitoring applications can't scrape any further details on the architectural components. The integrated graphics support DisplayPort, eDP, HDMI and VGA interfaces and can simultaneously output to two screens at a 4K resolution. 

Zhaoxin instructed us to use the default Windows 10 display drivers for the best performance, which yielded 15 fps in DOTA 2 at 1280x720 with the lowest quality settings. We also found HP's Zhaoxin Graphics DCH driver for the chip and gave it a spin. HP's driver is meant for the Windows 10 China Government Edition that is designed to meet the country's security and privacy standards (Microsoft basically stripped out all of its spyware for this special version of Windows for the Chinese government, and we don't have access to a copy). 

The driver installed fine on our Windows 10 Pro test image, resulting in the performance listed in the table above. We tested with both DDR4-2133 and DDR4-2666, but increased memory throughput didn't yield big performance improvements like we typically see with Intel and AMD iGPUs. The ~1.6 fps gain at 1280x720 and lack of improvement at 1920x1080 implies that the bottleneck lays elsewhere. In either case, these performance metrics significantly lag the Core i5-7400's UHD Graphics 630 engine that is largely considered useless for meaningful gaming. You're still best served going with AMD for integrated graphics performance. 

Dota 2 also loaded extremely slow on the Zhaoxin test system and was generally unresponsive for portions of our test run. As such, you'll need to pair the KX-U6780A with a discrete GPU for a decent gaming experience, not to mention to gain access to DX12 support. We have plenty of discrete GPU testing on the following pages. 

The processor supports Linux, Windows 10, and China's homemade "domestic operating system." China also has strict encryption technology requirements for domestically designed and produced chips, which consist of its own SM3 and SM4 encryption hashing functions (based on elliptic curves cryptography) defined by the Chinese National Standard. This is a similar implementation to the encryption used in the modified AMD EPYC designs that comprise the Hygon line of processors, with the obvious intention of avoiding prying western eyes through potentially backdoored encryption algorithms. The China Internet Network Information Center claims SM3 offers similar security and efficiency to SHA-256, while SM4 is similar to AES-128. The chips also support SHA-1 and SHA-256. 

On the security vulnerability front, the chips do require mitigations for Spectre Variant 2 but do not suffer from Meltdown. Zhaoxin is integrating in-silicon fixes for Spectre v2 in future architectures. We don't know the company's status on the laundry list of other mitigations that target processors with speculative execution features. 

Zhaoxin hasn't specified pricing for the KX-U6780A, though we have already seen systems on the market. Given the lack of pricing, and the fact that we're testing with a development board instead of a shipping motherboard, consider this a performance preview.

With those details out of the way, let's take a closer look. 

MORE: Best CPUs

MORE: CPU Benchmarks Hierarchy

MORE: All CPUs Content

Zhaoxin KX-6000 IPC and Performance Scaling

We prefer to measure instruction per cycle (IPC) throughput by locking all processors to the same frequency, typically matching the minimum base speed of the fastest processor to minimize the effect on cache and fabric timings that impact performance. However, the Zhaoxin KX-U6780A ticks at a mere 2.7 GHz that falls well below the minimum base speeds of comparable processors, and the spartan BIOS doesn't support modifying the multiplier. We also can't make adjustments to memory timings. As such, we tested with comparable processors without boost mechanisms or fixed the clock rates (A10-9700) to assure static frequencies, set the memory to the supported frequencies for each chip, and then normalized the results. This isn't our preferred method, but it is good enough for the task at hand. 

We assigned the Zhaoxin KX-U6780A as our baseline model, but it obviously lags the competing chips by a large margin. AMD's Bristol Ridge A10-9700 with the pre-Zen Excavator cores beats the Zhaoxin chip in every metric, while the 12nm Zen+ architecture on the Ryzen 3 3200G extends the lead further. The Ryzen 5 3600 with Zen 2 cements AMD's lead. You can see the current state of Zen 2's IPC performance with more expansive results here. 

Intel's Kaby Lake opens up a big lead over the Zhaoxin, representing the refreshed Skylake architecture's general IPC trend. Intel's current-gen Coffee Lake processors offer mostly identical IPC performance, so this is an accurate depiction of the current state of play for Intel, security mitigations included. Intel's stagnation on the microarchitectural front hurts versus AMD, but it has plenty of breathing room against Zhaoxin. 

Zhaoxin's key remit will be to improve IPC in the future through architectural enhancements and increased frequencies, but that isn't a straightforward proposition: Other aspects of the chip will also have to move forward in lockstep. 

The first chart in our album has the multi-threaded Cinebench score we attained with each processor (Multi-Core), along with that score divided by the number of cores (Multi-Threaded Per-Core Score). We also included the result of the Cinebench single thread test (Single Threaded Score). 

These heavily-threaded applications give us an idea of how well each workload scales on the respective architectures. Threading plays a role in boosting the per-core performance in each of these tests, but we're focusing simply on the performance of each physical core regardless of the number of cores. The KX-U6780A obviously suffers from poor per-core performance, but there may be other architectural issues at play that hinder scalability.

Pay attention to the per-core score we calculated from the multi-threaded result and the score from the single-threaded test. As you'll notice, chips with Hyper-Threading gain some performance over our calculations from the multi-thread test, often in the ~20% range, because now both threads are active on a single core. For chips that feature boost technology, they also get the added benefit of a higher single-core boost frequency, albeit typically for a short time considering the length of this test. It goes without saying, but the combination of those technologies would benefit Zhaoxin's processors.

However, there are other factors that can limit performance scalability. The Core i3-8100 doesn't have Hyper-Threading or boost technology, and you'll notice that it loses some performance, albeit not a drastic amount, when comparing the single-thread test results to our calculated per-core results from the multi-threaded test. These types of scaling losses can come from cache and fabric contention, a condition that can be exacerbated by bandwidth-consuming thread dependencies, so these factors must be accounted for during the design stages of the processor. You have to right-size the fabric for the job, and here we can that Zhaoxin's chip loses only four points between the two measurements. You'll see a similar trend with the POV-Ray tests. 

Scaling up per-core performance requires faster interconnects to handle inter-core traffic, not to mention access to memory and I/O devices. That's the key reason why both AMD and Intel pound the interconnect drums so frequently in marketing materials – it has a tremendous impact on workload scalability.

It's fully possible that the KX-U6780A's process node could clock higher, but striking the correct balance between per-core performance and chip interconnects could be best achieved by dialing back the frequency to match the interconnect saturation point, thus landing lower on the voltage/frequency curve and yielding better power efficiency and thermals. We won't know if that is a concern here until Zhaoxin shares more details of its architecture. 

We also included scaling tests with the V-Ray and Stockfish benchmarks, both of which scale very well and fully saturate the cores during operation. We don't have comparable single-threaded test results (there isn't a benchmark for that), but it does provide an interesting holistic view of how Zhaoxin relies upon more cores to compete with chips with far more efficient designs. 

Zhaoxin KX-U6780A Power Consumption

Measuring power consumption is always a tricky proposition, with different methodologies yielding different results. Intercepting power at the physical layer (i.e., measuring at the 8-pin connector) provides the most accurate measurements, but VRM inefficiencies lead to higher power draw measurements that don't match the actual power consumed by the processor. 

Many software utilities provide granular power logging features, but these reports can be inaccurate with some motherboards. However, the advantage of polling the sensor loop boils down to the fact that this technique measures the actual amount of power consumed by the processor itself. To merge the best of both worlds while still ensuring accuracy, we typically compare the power measurements at the physical layer from those we pull from the sensor loop to verify that the software output plausibly coincides. That technique enables fine-grained power testing that represents the real power consumption of the processor under test. 

Unfortunately, the Zhaoxin development board doesn't support sensor loop-based power logging, so we turned to Passmark's Inline PSU tester to measure the amount of power flowing into the 8-pin connector. This device measures in a pass-through mode with a high level of accuracy and has expansive logging capabilities, making it an excellent addition to our arsenal of power-testing tools. However, the measurements for the Khaoxin processor come directly from the 8-pin connector as opposed to the information from the sensor loop, so you'll have to account for VRM inefficiencies that can lop ~10 to ~15% off the power readings. 

Zhaoxin's design also complicates matters. We know little about the LiuJiaLolapoolza architecture, but the company informed us that while the chipset and graphics units are part of the single monolithic die underneath the heatspreader, these units pull power from a separate power domain that's fed through the 24-pin connector. The company uses a special motherboard to measure the total power draw of the processor, but we don't have access to that equipment.

Because of this power delivery arrangement we can't ascertain the real total power draw of the package (we have no way of knowing how much power flowing from the 24-pin goes to the processor, specifically), so you'll have to take these power measurements with a grain of salt. However, we did measure ~55W of power consumption through the 8-pin connector, and after accounting for VRM losses, we're looking at roughly the same power draw, if not slightly less, than AMD's A10-9700 that's fabbed on a 28nm process. 

Even with the somewhat unclear power results, we can see the power burden bestowed by the older 16nm process. This will certainly improve when the company moves to the 7nm process with the KX-7000 series, but unsurprisingly, the KX-U6780A isn't very power efficient compared to competing processors with smaller process nodes that yield lower power consumption and higher performance.   

Zhaoxin HX002EH1 Dev Board

Regular consumers will never see this reference validation board because it's designed for Zhaoxin's own internal dev work. We can see that the stock cooler topping the chip bears three heatpipes for efficient thermal dissipation, but the fan is loud and the BIOS doesn't offer any bells or whistles, like custom fan curves. Instead, the fan runs on its own based on load. 

The custom cooler mounts over the BGA package that sits adjacent to the 16-lane PCIe slot (lane width is x8, though). The board also houses one 4-lane and three single-lane PCIe slots, along with an old-school PCI slot. 

The four layer slab of PCB measures 244 mm x 305 mm, meaning it adheres to the ATX specification. As such, it also supports the standard ATX power connections, like a 24-pin and single 8-pin for power delivery. The 8-pin feeds a three-phase power delivery subsystem that comes with no additional cooling, such as a heatsink. That's not too much of a concern given the 70W TDP of the chip. 

The board sports a decent helping of connectivity that includes VGA, HDMI and DisplayPort outs along with the following accommodations: 

Four SATA 3.0 connectors
- One PCIe M.2
- One USB 3.1 Gen 2 port on one Type C connector
- One USB 3.1 Gen 2 port on one Type C pin header
- Two USB 3.1 Gen 1 ports on one Type A connector
- Two USB 3.1 Gen 1 ports on one pin header
- Two USB2.0 ports on one Type A connector
- Eight USB2.0 on x4 pin header
- Two UART ports
- One Audio Codec ALC662

The motherboard sports the ZX-200 IO expansion chip (6W chipset) that provides eight lanes of PCIe 2.0 and houses the in-built SATA and USB controller. The 40nm chip sports up to 11 USB ports, as listed above. Zhaoxin says this chipset is used for desktop PC's, all-in-ones, and laptops. 

The board is extremely spartan, and has the BIOS to match. You can't specify memory frequencies or timings, overclocking isn't permitted, and almost all of the features are handled automatically. We imagine that custom motherboards will come with more of the enthusiast-minded trimmings, though given the chip's capabilities, we wouldn't expect the fantastic RGB light shows and muscular power delivery cooling solutions we see on high end boards.  

MORE: Best CPUs

MORE: Intel & AMD Processor Hierarchy

MORE: All CPUs Content

Test Notes

We tested our game suite at both DDR4-2133 and DDR4-2666 and recorded performance deltas in a few titles. We've included DDR4-2133 results in the titles that experienced uplift beyond the normal standard deviation. That bodes well for Zhaoxin's future support for DDR4-3200, albeit with the stipulation that you'll need a pricey kit with the pre-defined DDR4-3200 SPD. 

VRMark and 3DMark

Synthetic gaming benchmarks often aren't indicative of real-world performance, but the 3DMark DX11 and DX12 tests measure the amount of raw computational horsepower exposed to the game engine. For now, most of today's game engines don't scale as linearly with additional compute resources, but these tests help us gauge how games could exploit processing resources as the engines become more sophisticated.

In the synthetic world of the Fire Strike benchmark, the Zhaoxin processor scales well and beats out nearly all of the dual-core competition, with the exception of the overclocked Ryzen 3 3000G. The Core i5-7400, which Zhaoxin has used as a comparison point, leads by ~8%, which isn't too bad of a showing for the newcomer. However, flipping over to the DX12 physics test finds the Zhaoxin processor barely edging out the AMD's Bristol Ridge A10-9700. The previous-gen Athlon processors, with a mere two cores to their name, also outperform the KX-U6780A. 

VRMark test prizes per-core performance (a mixture of frequency and IPC), exposing the KX-U6780A's weakness in lightly-threaded workloads. Even the Bristol Ridge chip, which hails from 2016, sails past the Khaoxin processor. Meanwhile, the Core i5-7400 nears the top of the chart. 

Civilization VI AI and Stockfish

Civilization VI prizes per-core performance, so it isn't surprising to see the Intel chips take the lead in this test of AI engine performance in a turn-based strategy game. The KX-U6780A isn't as competitive here as a result of its lower IPC and constrained clock frequencies, but it is in range of the Bristol Ridge A10-9700 with its Excavator cores. Socket AM4 soldiers on, though, with much better performance in successive AMD generations. 

Stockfish, an open-source chess engine, is designed to extract the utmost performance from many-core chips by scaling well up to 512 cores. That scalability meshes well with the KX-U6780A's eight-core design, but the four-core Intel i3 models continue to lead by a large margin. 

Ashes of the Singularity: Escalation

Ashes of the Singularity: Escalation loves cores and threads, but clock rates obviously play a role. That certainly doesn't help the eight-core KX-U6780A as it falls to beat all but the A10-9700. Even the dual-core quad-thread Athlon 200GE pulls ahead by an impressive 24%, though it's obvious this class of processor isn't the best fit for such a heavily-threaded workload. 

Civilization VI Graphics Test

Civilization VI is known for its preference for high clock rates, so it's no surprise to find the Zhaoxin chip grappling with Bristol Ridge yet again. We do see the slightest of gains when we switch from DDR4-2133 to DDR4-2666, so the step up to DDR4-3200 could provide a bit more performance in the future. We wouldn't expect miracles, though. 

Dawn of War III

The Warhammer 40,000 benchmark responds well to threading, but it's clear that Intel's clock speed and IPC advantage has an impact as the 8th- and 9th-gen chips take a solid lead. That isn't to discount the Ryzen 3 3200G, though, which musters a solid result given its price point. The Zhaoxin chip trails significantly.

Far Cry 5 

Far Cry 5 tells a now-familiar tale, with the pecking order remaining largely similar throughout the remainder of our tests. 

Final Fantasy XV

We run this test with the standard quality preset to sidestep the impact of a bug that causes the game engine to render off-screen objects with the higher-resolution setting. The Zhaoxin chip pulls up above the previous-gen dual-core Pentium G4560 by 4.9 fps, notching a much-needed win against an Intel chip. Those wins won't come often, though. 

Hitman 2

We experienced significant hitching during the Hitman 2 sequence with the Zhaoxin processor, but the A10-9700 also provided lackluster performance. 

Project Cars 2

Although Project CARS 2 is purportedly optimized for threading, clock rates obviously affect this title's frame rates. It's pretty clear that the KX-U6780A yields similar performance to the A10-9700 in most games. 

World of Tanks enCore

The World of Tanks benchmark doesn't hold any surprises, with the KX-U6780A again trailing the test pool. 

MORE: Best CPUs

MORE: Intel & AMD Processor Hierarchy

MORE: All CPUs Content

Test Notes

The firmware in the Zhoaxin-provided development board is certainly a spartan affair, so there are very few options. Overclocking is off the table, as are any type of memory manipulations. However, we did find an option to disable AVX, but we're told this feature is to address compatibility with some applications. We found that it did offer heightened performance in some workloads, and we've included those results in the relevant tests. 

Rendering

We expect heavily-threaded workloads to expose the best of the eight-core Zhaoxin's performance, but that only manifested in a few workloads – the majority of the tests find the processor lagging behind dual-core competitors. The chip does climb the ranks quite a bit more than it did in the gaming benchmarks, but it's clear that its single-threaded performance is lacking. 

For instance, the Zhaoxin chip trailed competing models tremendously in the single-threaded Cinebench test, but through the virtue of its extra cores it landed in the middle of the pack in the threaded Cinebench results.

Using the disable AVX option yielded some extra performance, to the tune of an extra 5.5% in the multi-threaded test and 5.8% in the single-threaded test, but that isn't enough to overcome even the dual-core Core i3-7100, let alone the modern quad-core chips. This could be a software optimization issue, but we see a similar trend in POV-Ray, though performance is considerably further behind the competing processors. 

The Zhaoxin chip manages to beat the Pentium lineup in the Blender benchmark, but falls to the newer Pentium models in the Corona and v-ray tests. 

Encoding

We typically use the Intel/Netflix-designed SVT-AV1 video encoder to test performance in a threaded encoding application, but the Zhaoxin processor repeatedly failed to execute the test properly. That leaves us with the single-threaded LAME and FLAC encoding test results, which both reveal a massive disparity in performance compared to competing chips from both Intel and AMD. This could be a lack of software optimization for the architecture, but it's clear that this single-threaded performance needs to improve drastically in future Zhaoxin architectures through both a combination of higher frequencies and microarchitectural advances.   

The threaded HandBrake x264 and x265 tests again reveal the Zhaoxin chip's subpar performance in AVX workloads. The x265 test uses a heavier distribution of AVX instructions than the x264 tests, but even AMD's aging Excavator cores in the A10-9700 outperform the KX-U6780A in both tests.  

Web Browser

Browsers tend to be impacted more by the recent security mitigations than other types of applications, so Intel has generally taken a haircut in these benchmarks of fully-patched systems. However, neither Intel or AMD face any meaningful competition here from the Zhaoxin processor. The results speak for themselves – the KX-U6780A isn't a contender in lightly-threaded applications. 

Office and Productivity

The Zhaoxin processor fails to impress in any of our office and productivity tests as it loses every benchmark by large margins. We expect these chips to find plenty of use in governmental offices, but that will come with a stiff performance penalty and generally sluggish performance in these types of applications. 

Compression, Decompression, Encryption, AVX

The 7zip and Zlib compression/decompression benchmarks rely heavily upon threading and work directly from system memory, thus avoiding the traditional storage bottleneck in these types of tasks. The KX-U6780A puts up much more impressive performance in these tests as it can utilize its memory throughput and cores to chew through the workloads, besting all but the quad-core Intel processors. AMD's dual-core Ryzen 3 3200G proves its mettle with its four threads as it notches a considerable lead against the Zhaoxin chip in both 7zip decompression and compression workloads. 

The heavily-threaded y-cruncher benchmark, which computes pi using the taxing AVX instruction set, finds the Zhaoxin chip getting a nice performance bump by toggling the AVX compatibility feature in the BIOS, but we see the same trend of woeful single-threaded performance offset by the heft of eight cores in the threaded workload. In either case, losing to the $50 dual-core Athlon 200GE in a threaded workload isn't a good look for an eight-core processor. 

China's state-defined encryption is a big impetus for its homegrown CPU initiatives, but there aren't any established SM3/SM4 tests available yet from the roster of ISVs that provide reputable test software. We're sending out requests for tests that reflect the advantages/disadvantages of those algorithms. Meanwhile, we can see that the chip provides strong performance in the AES, Hashing, and SHA3 benchmarks.

MORE: Best CPUs

MORE: Intel & AMD Processor Hierarchy

MORE: All CPUs Content

It's clear from our tests today that Zhaoxin still lags the leaders of the processor market, which comes as a side effect of both lagging process technology and architecture. The Zhaoxin KX-U6780A suffered at the hands of its relatively tame 2.7 GHz clock speed throughout our test suite, and we don't think the flagship's extra 300 MHz would significantly change the overall outcome. 

AMD and Intel have the advantage of decades of research and development, not to mention pioneering roles in the evolution of the x86 instruction set. That makes it extremely hard for any newcomer, x86 or otherwise, to establish a competitive product. We've often heard comparisons of the KX-6000 series to Intel's seventh-gen Core i5 series, but those are obviously borne of a very narrow selection of tests, if not a single test, used as a comparison point.

It's safe to say that Intel doesn't have to worry about meaningful competition from Zhaoxin yet, and the same goes for AMD. Even AMD's outdated Bristol Ridge chips with Excavator v2 cores were enough to stave off the challenger – we didn't even have to bring a Zen 2 processor into the test pool for AMD to enjoy a healthy performance lead. 

However, Intel's historical advantage of being an integrated device manufacturer (IDM) has become a liability as the company struggles to economically mass-produce its 10nm node. Now TSMC serves as the great equalizer in the processor market because nearly any chip designer can simply adopt its 7nm node and have a leading edge process that's competitive with Intel's 10nm. That's an advantage that bodes well for future Zhaoxin chips if it continues to leverage the Taiwanese foundry. The picture gets a bit murkier if Zhaoxin goes with Chinese fab SMIC for its 7nm KX-7000 processors, but that seems unlikely in the near term given the easy access to TSMC's mature nodes. However, anything is possible given the Chinese government's desire for a completely homegrown chip.

Zhaoxin plans to adopt the PCIe 4.0 and DDR5 interfaces with its next chips, but as we observed throughout our testing, the company will have to pair those features with other meaningful advances to make real headway. The higher clock speeds, lower power consumption, and increased density of the 7nm node will yield improvements, no doubt, but Zhaoxin needs to improve its microarchitecture.

The company has made alterations to Centaur's Isaiah architecture for its previous chips, but even though we don't know the finer details, Zhaoxin may need a clean-sheet design moving forward. Centaur, which largely operates independently inside of Via, does have a new incredibly promising CNS core architecture in the works. It's possible that Zhaoxin will adopt this architecture, or a derivative, for its KX-7000 series [EDIT: Zhaoxin later clarified that it will not use a CNS derivative]. 

For now, the Zhaoxin line of chips are unimpressive compared to AMD and Intel in the global market, but given the Chinese government's overriding desire to free itself from western influences on its economy, not to mention its chip-powered military endeavors, that might not mean much. China's sprawling initiative to rid its government offices of foreign software, hardware and encryption by the end of 2022 is an incredibly ambitious goal, and that means it will adopt and incentivize nearly any indigenous logic, regardless of performance. There are other alternatives: Huawei is adopting ARM architectures for a new line of desktop PC motherboards, but the ubiquity of the x86 instruction set is hard to ignore. 

As we've often seen, 'cheap and good-enough' almost always wins the bulk of the market. If Zhaoxin gets pricing right, it should be able to carve out a nice chunk of a suddenly larger Chinese PC market.

China's 3-5-2 initiative only applies to its governmental pursuits, though, so Intel and AMD will still be available to everyday Chinese consumers. At least for now. We're sure that the majority of enthusiasts and professionals will stick with Intel and AMD hardware given they offer more performance and efficiency, but Zhaoxin has backing that assures its future architectures will become more competitive. Some of the lackluster performance we recorded today could also be reminiscent of the challenges AMD faced with its new architectures, so broader adoption could yield some performance uplift via software optimizations.

Remember AMD's Phoenix-like rise from the relative ashes of the semiconductor market to the value and performance leader? I do. It only took one daring new architecture with a massive 52% IPC gain paired with a good-enough 14nm GlobalFoundries process, and perhaps a little bit of luck with Intel's delays on the 10nm node, to upset both the desktop PC and data center markets. 

That begs the question: Is Zhaoxin just one microarchitecture away from becoming a serious contender on the global stage, granting China success with its chip initiatives at the same time? Only time will tell, but it all starts with achieving process node parity. Zhaoxin plans to achieve that mark in 2021.  

MORE: Best CPUs

MORE: Intel & AMD Processor Hierarchy

MORE: All CPUs Content

Zhaoxin, based in Shanghai, has its domestically produced KX-6000-series processors ready for prime time. The x86 chips are already listed on Chinese e-commerce website Taobao and are scheduled to launch as soon as March of this year.

The KaiXian KX-U6780A is based on the LuJiaZui microarchitecture. It's rated with a 70W TDP (thermal design power) and sports eight cores and eight threads clocked at 2.7 GHz. The processor has 8MB of L2 cache but lacks L3 cache. TSMC produces the KX-U6780A for Zhaoxin with the foundry's 16nm process node.

Like your typical mainstream processor, the KX-U6780A has a dual-channel memory controller. The chip supports DirectX 11 graphics, modern interfaces, such as M.2, PCIe 3.0, SATA and USB 3.1, and instruction sets including SSE 4.2 and AVX.

The KX-U6780A comes wrapped in BGA packaging, but, thankfully, Shenzhen Cjoyin Electronics has developed the C1888 mini-ITX motherboard for the eight-core processor. 

KaiXian KX-U6780A + C1888 Combo

Chinese YouTube channel 二斤自制 got its hands on the KX-U6780A and C1888 combo. The C1888 in the video is an engineering sample, so the final specifications will differ a bit.

The C1888 motherboard currently features an old-school green PCB, but will also be available in white after its official release. Additionally, the manufacturer plans to add a power button and standardize the motherboard holes for mainstream CPU coolers.

The C1888 features two DDR4 SO-DIMMs that support DDR4-3200 RAM. It also provides an mSATA and SATA port for storage. 

Motherboard-maker Shenzhen Cjoyin claims that the C1888 supports the NVMe protocol and that it will add an M.2 slot to the finished product to accommodate high-speed SSDs.

The C1888 comes with an Embedded DisplayPort (eDP) and two COM interfaces. There's also a slot for a PCIe wireless network card. Similar to other mini-ITX motherboards, the C1888 only has one PCIe 3.0 expansion slot. Unfortunately, it's wired to operate at x8.

Connectivity options on the C1888 include one front USB 2.0 header, one HDMI port, one VGA port, two USB 2.0 ports, two USB 3.0 ports, two Gigabit Ethernet ports and three 3.5mm audio jacks.

According to the video, the KX-U6780A and C1888's maximum power consumption amounts to 90W. The motherboard draws power through a DC power connector and 4-pin power connector. However, the final product will drop the 4-pin power connector in favor of the standard 24-pin. The motherboard will retain the DC power connector for the convenience of small form factor (SFF) users.

KaiXian KX-U6780A Benchmarks

In order to measure the progress of China's domestic processors, the KX-U6780A is pitched against the previous KX-C4580. As a recap, the KX-U6780A has eight cores and eight threads running at 2.7 GHz, while the KX-C4580 only has four cores and four threads at 1.83 GHz.

Thanks to its four additional cores and higher clock speeds, the KX-U6780A showed some big performance improvements over the KX-C4580. In the multi-thread benchmarks, the eight-core chip outperformed the quad-core part by over 200%. In terms of single-core improvement, the KX-U6780A is up to 171% faster.

Based on these results, the KX-U6780A should be slower than an Intel Core i5-7600K. Intel's quad-core chip delivers up to 180% and 32% higher performance than the KX-U6780A on the single-and multi-thread tests with the CPU-Z benchmark, respectively.

The current generation of Zhaoxin chips are still admittedly behind the likes of Intel and AMD. Nevertheless, the Chinese chipmaker has ambitious plans to catch up to its rivals by 2021.

The tech YouTube channel also paired the KX-U6780A with an aftermarket RTX 2060 Super graphics card for a couple of gaming tests.

It should be emphasized that the test system is running on an engineering sample of the C1888 motherboard. As a result, the processor and graphics card usage are somewhat low because the motherboard limits the power of the PCIe 3.0 slot. The manufacturer will remove the restriction for the final product so that the graphics card will have more room to maximize its potential.

In any case, the results showed the KX-U6780A working with the RTX 2060 Super capable of maintaining 60 frames per second (fps) and above in the games tested. However, we would love to see how the couple holds up in the more processor-intensive games.

The final pricing for the KX-U6780A with the C1888 motherboard is unknown. The bundle is currently up for pre-order for 4,300 yuan ( ~$620) Taobao, and it's limited to 55 units.

MyDrivers reports that China's homegrown Zhaoxin processors are now operational with its self-developed Unity Operating System (UOS). It's no surprise that China wants to stop depending on U.S. technology, and the latest development is an important step towards the country's autonomy. The news come on the tail of new Chinese government restrictions that force Chinese institutions to replace all foreign-powered PC hardware and software with domestically-produced products over a three-year span. 

Tongxin Software, the developer behind UOS, has been able to get the KaiXian KX-6000 and KaiSheng KH-30000 processors to work on the desktop and server versions of UOS, respectively. The aforementioned chips are Zhaoxin's latest 16nm parts that sport eight cores with base clock speeds up to 3 GHz. The Chinese CPU maker has already laid out its ambitious plans for the company's next generation of processors.

Thanks to the new integrated display driver in UOS, the KaiXian KX-6000 processors offer improved video and audio performance. The chips also support 3D graphics and ultra-high-definition video decoding, among other features. On the server side, the company says the KaiSheng KH-30000 parts have exhibited excellent stability, and the chips should work fine inside file and database servers.

Tongxin Software will continue to work hand-in-hand with Zhaoxin to optimize and improve application performance with the chipmaker's processors. Tongxin Software is also committed to getting other Chinese-made chips, such as Loongson, FeiTeng, ShenWei, Kunpeng and Hygon, to work with UOS. In due time, the country hopes UOS will have matured enough to replace Microsoft's Windows operating system completely.

As reported by WikiChip Fuse this week, Chinese CPU maker Zhaoxin has outlined the designs for its next-generation KaiXian and KaiSheng processors. This reveal comes on the heels of a Bloomberg report this fall that China is pushing hard to decrease the country's dependency on U.S. technology.

Zhaoxin has two primary product lines. The KaiXian chips are designed for the consumer market, while the KaiSheng processors are tailored towards the server market. Zhaoxin's existing KaiXian KX-6000 and KaiSheng KH-30000 series are on the outdated 16nm process node. However, the Chinese company expects to reach parity with Intel and AMD by 2021.

KaiXian KX-7000: Marching to PCIe 4.0 and DDR5

Current KaiXian KX-6000 processors are based on the LuJiaZui microarchitecture and produced with TSMC's 16nm manufacturing process. The chips max out at eight cores and feature base clocks up to 3 GHz. However, the upcoming KX-7000 chips will seemingly employ a new processor microarchitecture, which Zhaoxin hasn't unveiled yet. 

TSMC is reportedly in charge of producing the KX-7000 chips for Zhaoxin with the foundry's 7nm process node. Going from 16nm to 7nm is a pretty big jump and should allow Zhaoxin to squeeze more megahertz into the KX-7000 chips. If the new microarchitecture is legit, Zhaoxin should be able to close the performance gap between it and Intel or AMD.

The 7nm KX-7000 parts are expected to come with a new iGPU (integrated graphics processing unit) that's compatible with DirectX 12, as well as support for the latest PCIe 4.0 interface, (which Intel chips don't support yet), and DDR5 RAM.

According to WikiChip, Zhaoxin has started working on sub-7nm designs for several mobile and desktop CPUs as well. The Chinese semiconductor company is keeping a tight lip on the details. However, the publication believes that Zhaoxin could hit TSMC up for its N5 or N5P manufacturing process.

KaiSheng KH-40000: Leaping to 32 Cores

Similar to Zhaoxin's consumer offerings, the existing KH-30000 chips feature the same LuJiaZui microarchitecture and 16nm node. They sport up to eight cores and base clocks up to 2.7 GHz. The KH-40000 series will be the direct replacement and debut with a new and improved microarchitecture. 

Sadly, the KH-40000 will be stuck on the 16nm process node. Nonetheless, Zhaoxin is promising to quadruple the core count from eight to 32 cores. This would effectively put the KH-40000 on the same ground as AMD's Threadripper chips - at least from a core perspective. It remains to be seen how the Zhaoxin's new microarchitecture fairs against its rivals.

The KH-40000 processors seemingly come with dual-socket support, so it will be possible to combine two of them for the possibility to have up to 64 cores in a single system. The 16nm chip's other attributes include continued use of DDR4 RAM and the PCIe 3.0 interface.

Late last year, the China-based processor designer Zhaoxin Semiconductor (jointly owned by the Shangai government and VIA Technologies, another semiconductor corporation) promised its upcoming octa-core CPUs based on the 16nm node from TSMC would be able to match Intel's quad-core i5 processors, and today is that day: the newly-announced KX-6000 CPUs are said to deliver performance on par with the Core i5-7400, yet supposedly pull off the feat at a mere 3 GHz. And they support Windows. While this level of performance may not seem very impressive, it is actually a very significant development for a few reasons.

Companies like VIA and Zhaoxin are eager to compete with Intel, AMD, IBM, and others because China wants its own processors, not only to compete within its own borders but also to compete abroad and reduce the country's exposure to outside influence. The international CPU scene is dominated by American companies like Intel, AMD, and IBM, among others. For both political reasons (such as self-determination and independence) and economic reasons, the prospect of China producing and exporting profitable and powerful processors is enticing.

It's also a significant technological achievement for Zhaoxin. Any kind of CPU that can compete with Intel on any field is impressive since Intel is at the top of the hierarchy in both size and technology. Intel's traditional rival AMD and the more server- and specialist-focused IBM are Intel's only real competition; it would certainly shake up the dynamic if Zhaoxin could field competitive processors for the x86 market. The Core i5-7400 isn't Intel's fastest desktop CPU by any means (not even when it launched in 2017), but it is one of the fastest desktop CPUs made by any company today. Even AMD's first-gen Ryzen CPUs struggled against it in single-threaded workloads.

Lenovo already uses previous-gen Zhoaxin processors in some notebooks, and benchmarks have popped up in public databases, meaning the chips are more than just a pipe dream. Zhaoxin is also similarly planning on launching competitive server CPUs (the KH-40000 series), but on the 7nm node from TSMC with PCIe 4.0 and DDR5 support. Zhaoxin hasn't said exactly when these CPUs would arrive, but it will be a little while since it is slated to use upcoming DDR5 memory.

Via’s joint venture in China is apparently hoping to have a wholly Chinese-developed and manufactured X86 SOC that is competitive with AMD’s current offerings by 2019.

The older of us may remember the days of owning a motherboard with a Via chipset. Along with Intel and AMD, the Taiwanese company is one of three that have a license to produce the X86 architecture. It’s not surprising then that powers in the Chinese government approached them to help develop the Chinese domestic semiconductor industry.

In 2013, Via entered a joint venture with the municipal government of Shanghai to found Zhaoxin, a fabless semiconductor company. Since then, it has been continuously developing its own X86 SOCs. Now in its fifth generation, Zhaoxin’s latest KX series of SOCs combine its home-grown X86 CPU with DDR4, PCI-E 3.0, USB 3.1, and SATA3 controllers. The chips are fabricated on a 28nm process and can operate in the 2GHz range. Zhaoxin’s current adopters include Lenovo, but its chips don’t make it out of the country because they’re largely used in products built for the Chinese government.

In a press conference reported on by eefocus, a Chinese website, Zhaoxin discussed its future product roadmap. We’re relying on Google Translate, but eefocus mentions that Zhaoxin also follows a “tick-tock” development cycle. Progressing from this year’s KX-5000 series, which is a “tock” cycle, Zhaoxin’s 2018 KX-6000 series will be the “tick” cycle. That chip will move to a 16nm process and is hoping to hit frequencies in the 3GHz range. The 2019 KX-7000 series will be an another “tock” cycle, which means an architecture change.

What’s special about the KX-7000 series is that Zhaoxin expects it to be competitive to AMD. We don’t know from the translation whether the text is referring to AMD’s current processors or AMD’s 2019 processors. It’s not explicitly stated, but we assume that the 7000 series will still be fabricated on the same 16nm process, so it’s unlikely to be competitive with AMD’s 2019 Zen 2 processors, which will be built on a 12nm process.

Another interesting note is that, in a picture from German website golem.de, the 7000 series is shown to have a DDR5 memory controller. We don’t know what expected performance gain there is from this or if any such a gain has been accounted for in the statement on being competitive with AMD. DDR5 isn’t out yet, but history has shown that faster memory doesn’t usually afford much performance gain  in workloads that aren't memory bound, so even reaching performance parity with Zen, which runs DDR4, will be quite a feat for Zhaoxin.

Other remarks by the company show that it is currently still focusing on being competitive by being cheaper. Its target market, for now, remains desktop and laptop computers destined for procurement by the Chinese government. China has always sought to have home-grown industry, both for economic development and national security. Especially with the involvement of the Chinese government, we won’t be surprised to see Zhaoxin taking a place beside China’s other fast-growing, home-grown chip developers.