Chinese semiconductor industry

Status
Not open for further replies.

gelgoog

Lieutenant General
Registered Member
You need a lot less chips for base stations that you would need to make smartphones. But eventually the supply will run out and they are stuck with what chip supply they do have. I think Huawei will need to start making their own chips soon.

China had 1.43 million 5G base stations total last year. Compare that with amount of smartphones.

China built 654,000 base stations in 2021. But they shipped 266 million 5G phones.
While a base station does use a lot more semiconductors than a smartphone, it is probably like an order of magnitude more semiconductors per unit. Not two and a half orders of magnitude.
 
Last edited:

KYli

Brigadier
You need a lot less chips for base stations that you would need to make smartphones. But eventually the supply will run out and they are stuck with what chip supply they do have. I think Huawei will need to start making their own chips soon.

China had 1.43 million 5G base stations total last year. Compare that with amount of smartphones.

China built 654,000 base stations in 2021. But they shipped 266 million 5G phones.
While a base station does use a lot more semiconductors than a smartphone, it is probably like an order of magnitude more semiconductors per unit. Not two and a half orders of magnitude.
From what I heard, Huawei has enough base stations chips to last at least until 2024.

Please, Log in or Register to view URLs content!
 

antiterror13

Brigadier
It's called the loongson. Apparently it's quite disliked on chinese socmed and a lot of people say it's useless

Can you give us the link that people says Loongson is useless ?

I thought it was quite good, especially 5000 series. My understanding is that most Govt desktop use Loongson also top Chinese Supercomputers use it as well, including Beidou sats

Also latest Loongson includes "LoongArch" Chinese ISA

I believe Loongson will get better and better in fast pace in this environment
 

tokenanalyst

Brigadier
Registered Member
You need a lot less chips for base stations that you would need to make smartphones.
Weird, i always though that a base stations would required more chips than an smartphone. I do know that the reason smartphones SOC require the latest node process is because smaller gates can switch with less power, so you can pack more in an SOC and no decrease on battery life. Base stations in other hand power is not an "big" issue, so Huawei is betting in compound semiconductors like GaAS, GaN, SiC, advanced packaging, photonics and software algorithms to get more performance out of their products without the need of smaller process nodes. In other words Huawei is trying to innovate the sh*t out their situation until China catch up in semiconductor manufacturing.

Please, Log in or Register to view URLs content!
Please, Log in or Register to view URLs content!
Please, Log in or Register to view URLs content!
Please, Log in or Register to view URLs content!
Please, Log in or Register to view URLs content!
Please, Log in or Register to view URLs content!
Please, Log in or Register to view URLs content!

1649542412516.png1649544466902.png
 

hvpc

Junior Member
Registered Member
Also base station don't need 7nm chips, 14nm even 28nm is totally fine
Indeed. A lot of chips go into a basestation than a mobile phone: FPGA or ASIC, memory, converters (inverter, rectifier), power amplifiers, power management IC, RF, analog, RF filters, wireless communication, etc. chips. I'm sure I'm missing some additional chips in the list I provided.

Huawei has easy access to mostly all except their own designed ASIC chip which was on 14nm and/or 7nm. @antiterror13 is correct that 28nm could be used, but I think Huawei/Hisilicon will have to spend more money to design their own 28nm ASIC or use general 28nm FPGA out on the market. It does look like Huawei is trying to come up with something specific and optimal to their needs through heterogeneous integration of multiple 28nm chips for when they run out of their stock of 14/7nm ASIC chips. Of course this method would not have the best power vs performance, which is an important KPI, but this should still be good enough in my opinion.

Another key enabler of 5G basestation is GaN power amplifier which has wider bandwidth and thermal performance. This is why China has been very big in promoting third generation compound semiconductor development in China. From what I've seen, China is leading or on-par with rest of the world on GaN & SiC chips manufacturing capability. So we are okay on this front.

So, ASIC chip and maybe transceiver switches are the only areas that may hinder Huawei's long term dominance in 5G network.
 

daifo

Captain
Registered Member
Can you give us the link that people says Loongson is useless ?

I thought it was quite good, especially 5000 series. My understanding is that most Govt desktop use Loongson also top Chinese Supercomputers use it as well, including Beidou sats

Also latest Loongson includes "LoongArch" Chinese ISA

I believe Loongson will get better and better in fast pace in this environment

I think i remember looking at benchmarks and the earlier 5000 series was running like a 8-10 year intel x86. I am guessing not many binaries are compile for mips/loongson , hence more trouble trying to get even open source software to run. x86 emulation? Prob running 1/2 speed. Deepin seemed to be really buggy the last time i installed it, prob adds to the misery lol

I am sure it "works" but besides the nationalist or hobbyist, i doubt any regular users will be praising it
 
Status
Not open for further replies.
Top