Huawei's next-gen CPU could rival Apple's current best — Kirin CPU with Taishan V130 cores rumored to match Apple M3 performance
Another ambitious project of a blacklisted company
According to a leak from a Weibo user, Huawei's chip development unit is working on a next-generation Kirin processor for PCs in addition to releasing the HiSilicon Kirin 9010 smartphone application processor. That part will feature the company's next-generation general-purpose cores and a revamped GPU. Huawei reportedly hopes the new Kirin will compete against Apple's M3 in multi-thread workloads.
Huawei's upcoming HiSilicon Kirin processor is projected to feature eight general-purpose Arm cores, including four next-generation high-performance Taishan V130 cores (the Kirin 9000s uses Taishan V120 cores) and four energy-efficient cores. The system-on-chip is expected to integrate the company's next-generation Mailiang 920 GPU with 10 clusters, bringing a vast performance uplift compared to the Kirin 9000s, which features a quad-cluster Mailiang 910 GPU.
The new Kirin processor will reportedly support up to 32GB of memory, which suggests it will have a 128-bit interface for LPDDR5/LPDDR5X or DDR5 SDRAM.
In general, Huawei expects its next-generation HiSilicon Kirin SoC for client PCs to offer multi-thread performance close to Apple's M3 and graphics performance close to that of Apple M2. However, it is unclear how the next-gen Kirin processor for desktops and laptops will perform against Intel's 16-core Core Ultra 9 185H (6P, 8E, 2LP) platform that powers Huawei's latest Matebook X Pro laptop.
Undoubtedly, Huawei's next-generation Kirin processor for client PCs looks quite impressive. Whether its performance will be on par with Apple's M3 or M2 remains to be seen, but the tech giant's ambitions are evident.
The leak claims that Huawei's HiSilicon could release higher-performance Kirin SoCs in "Pro" and "Max" configurations, similar to Apple, with more general-purpose cores, revamped GPU, and a more comprehensive memory interface.
Given that Huawei does not control 8% to 10% of the global PC market (like Apple does), building higher-end flavors of its Kirin processors for desktops and laptops may not make sense. However, because China's objective is self-sufficiency in the semiconductor field, Huawei's high-performance Kirin processors could be instrumental in replacing higher-end parts from AMD and Intel from PCs used by various government agencies. From this point of view, it is strategically feasible for the company to pursue such an endeavor.
Stay On the Cutting Edge: Get the Tom's Hardware Newsletter
Get Tom's Hardware's best news and in-depth reviews, straight to your inbox.
The only question is whether Huawei's manufacturing partner SMIC will have enough leading-edge production capacity to make all the chips that Huawei needs.
Anton Shilov is a contributing writer at Tom’s Hardware. Over the past couple of decades, he has covered everything from CPUs and GPUs to supercomputers and from modern process technologies and latest fab tools to high-tech industry trends.
-
frogr Do you realy believe that SMIC's 7 nm node can match the power- performance curve of TSMC's 3 nm node? Or perhaps their microprocessor design is that much better than Apple's M3 design.Reply -
aetherwing
TBF, the original source didnt mention about power efficiency. and there are some other rumors implying that 5nm node is well under way. anyway, time will tell.frogr said:Do you realy believe that SMIC's 7 nm node can match the power- performance curve of TSMC's 3 nm node? Or perhaps their microprocessor design is that much better than Apple's M3 design. -
The Historical Fidelity
Nope, in fact it’s not even as good as TSMC N7, I read somewhere that it’s equivalent to TSMC N12frogr said:Do you realy believe that SMIC's 7 nm node can match the power- performance curve of TSMC's 3 nm node? Or perhaps their microprocessor design is that much better than Apple's M3 design. -
The Historical Fidelity
TBF, considering the Kirin 9000 saw significant performance regression when switching from TSMC to SMIC, I doubt SMIC 5nm will level the playing field.aetherwing said:TBF, the original source didnt mention about power efficiency. and there are some other rumors implying that 5nm node is well under way. anyway, time will tell. -
aetherwing
significant performance regression? i am sure if you are trolling or not, but they have almost same performance in test runs and benchmarksThe Historical Fidelity said:TBF, considering the Kirin 9000 saw significant performance regression when switching from TSMC to SMIC, I doubt SMIC 5nm will level the playing field. -
The Historical Fidelity
Yes, the kirin 9000 (SMIC) has a 25% lower CPU and iGPU clock frequency vs kirin 9000 (TSMC) while using more electricity. All the benchmarks by reputable reviews show this performance regression.aetherwing said:significant performance regression? i am sure if you are trolling or not, but they have almost same performance in test runs and benchmarks -
Jahmalachi
Where you get that from? I swear any news is taken as fact as long as it’s anti-chinaThe Historical Fidelity said:Nope, in fact it’s not even as good as TSMC N7, I read somewhere that it’s equivalent to TSMC N12
kirin 9000 tsmc 5nm vs Kirin 9000s smic 7nm why are you surprised it’s less efficient?The Historical Fidelity said:Yes, the kirin 9000 (SMIC) has a 25% lower CPU and iGPU clock frequency vs kirin 9000 (TSMC) while using more electricity. All the benchmarks by reputable reviews show this performance regression. -
The Historical Fidelity
SMIC 7nm is compared to TSMC N12 due to the frequency-power curve. If it takes more power to produce 25% slower clock rates, that yields a frequency-power curve that is worse than TSMC N7 and on par with TSMC N12.Jahmalachi said:Where you get that from? I swear any news is taken as fact as long as it’s anti-china
kirin 9000 tsmc 5nm vs Kirin 9000s smic 7nm why are you surprised it’s less efficient?
P.S. I see you are new to these forums so I will kindly inform you to remain civil when engaging other members. -
gg83
I'd like a trustworthy source for that statement. Only benchmarks prove nothing about actual performance.aetherwing said:significant performance regression? i am sure if you are trolling or not, but they have almost same performance in test runs and benchmarks