nice, this was speculated a while back when they launched IPO bid and a significant portion of the capital from that was said to go into this second Wuxi fab.
Chinese semiconductor industry
- Thread starter Hendrik_2000
- Start date
- Status
Loongson won two award from the securities and fund industry. Looks like the Finance industry has been pushed to use as much of the domestic CPUs as possible and Loongson should be gaining more orders here.
From this context, seems like they've completed the working for fully domesticating the hardware stack and software ecosystem for the Financial industries.
Yes, I also thought about it.
There are 2 Arf lens system, one, possibly 0.75NA, for 90nm resolution and another, maybe 0.85NA, for 65nm resolution. We don't know when they decided to skip the dry version with the 0.85NA and move it directly to immersion.
IRCD has a SMEE prototype since some time already, so maybe it was sent before they made this decision.
So maybe they sent to IRCD a SMEE 0.85NA in dry setup....and only later they decided to skip the production of this model and move directly to immersion, and IRCD ended up with a prototype of the Arf dry 0.85NA. that will remain a single sample.
This would explain how IRCD is able to process 55nm nodes.
Update on this
Hygon 3 is definitely on sale now and is enough to satisfy the market demand. Hygon 4 will be significantly improved according to this.
Chinese CPU market is less than 20% domestic, so a lot of room for improvement.
From the Chinese social media. Phase 2 will be 83k wpm (phase 1 is going to be 95k wpm this year). Phase 1 lines did 90-65/55nm process. Phase 2 will elevate to 65/55-40nm process. Higher level of process will require more expensive tooling.
Looks like this is forced because auto chips are moving to more advanced nodes with 40nm MCU, CMOS image sensor, Display drive IC for consumer electronics and auto and such. Moving to more advanced process will allow for access to larger market. Auto grade MPU/MCU are moving from 90nm to more advanced process, so they need to do so too.
If construction works starts soon, then production can start in 2024 (based on the 17 months construction -> production time of phase 1). With IC market recovering later this year, 2024 could be a good year to add capacity.
yeah, I don't know if Hygon can compete long term since they are pretty restricted on the licensing part. I think they've been the biggest domestic player due to the maturity of their technology. But it seems like their upside is quite limited. Long term, I think it's going to just be RISC-V and LoongArch in China. Even ARM has better future than x86.
The big jump in cost will come if they switch from a dry ArF capable node to an immersion ArF node. For IoT and auto you need tons of very cheap yet long lasting and rugged chips. I don't think you can price them that low if you have to use immersion ArF since the equipment cost is 5x higher and the process cost is N times higher due to new resist requirements and process inefficiencies introduced due to the water compatibility requirement (i.e. waterproof resist topcoats). The ruggedness is also an issue, particularly the thermal stability.
Note the post says that the only reason that they're migrating from 90 nm to 65-45 nm is due to shortages in 90 nm capacity, not because they want to!
Interesting comments here from various Chinese GPU makers. No mention of Moore Threads for some reason, even though they are becoming quite successful.
So, there is always the concern that Biren just got turned off by TSMC due to the October surprise. My guess is that they most likely got around it by lowering their transmission speed. It seems like they've gone radio silent for the past 3 months to not attract more attention for themselves. Same with several other Chinese GPU makers who still have some work to do before their GPGPU is ready with the full software stack to be used by cloud services. At this point, I really hope they've contracted with TSMC to have a full stock of their GPUs in the even that they get even more cut off. Until then, a lot of their work will be involved with getting their GPUs tested, validated and integrated with all the major partners they've signed up with like China Mobile, Inspur, Ping An and Baidu.
So, there is always the concern that Biren just got turned off by TSMC due to the October surprise. My guess is that they most likely got around it by lowering their transmission speed. It seems like they've gone radio silent for the past 3 months to not attract more attention for themselves. Same with several other Chinese GPU makers who still have some work to do before their GPGPU is ready with the full software stack to be used by cloud services. At this point, I really hope they've contracted with TSMC to have a full stock of their GPUs in the even that they get even more cut off. Until then, a lot of their work will be involved with getting their GPUs tested, validated and integrated with all the major partners they've signed up with like China Mobile, Inspur, Ping An and Baidu.
啥情况，90nm光刻机是能做55nm的吗？
can the 90nm lithography machine do 55nm?
可以。65nm和55nm用的是相同的设备，这2个节点上光刻真正能满足的是互连层，而对于小于互连层的更小尺度的线宽则需要配合光刻胶、材料、刻蚀等一些特殊的工艺手段巧妙实现
Can. 65nm and 55nm use the same equipment, the lithography on these two nodes can really satisfy the interconnection layer, and for the smaller-scale line width than the interconnection layer, it needs to cooperate with photoresist, material, etch and other special craft means to achieve ingeniously.
- Status