Chipmakers are crippling products to evade US China ban
(2022/11/08)
- Reference: 1667919610
- News link: https://www.theregister.co.uk/2022/11/08/chipmakers_alter_products_china_ban/
- Source link:
Systems that once contained NVIDIA and TSMC chips, which are now restricted by the US government, are popping up this week with slower specs to meet US export controls to China and evade the hassles of obtaining special licenses.
Chinese server maker Inspur was [1]spotted swapping out the its A100 Graphic Processing Unit (GPU) in its [2]NF5688MG [PDF] for an A800 in [3]updated promotional material.
H3C, the China-based Tsinghua Group and Hewlett Packard mashup, are also suddenly featuring the A800 on [4]select servers.
[5]
According to [6]specs provided online by Chinese server-maker Omnisky, which also uses the chip, it remains eerily similar to the [7]A100 , [PDF] but with a 40GB PCIe option available in the A800, in addition to a 80GB PCIe and 80GB SXM available in both models.
[8]
[9]
The A800 has a degraded transfer rate of 400GBps across all three variations, as opposed to the 600GBps in the A100, thus falling in line with US law that restrict 600GBps and above being shipped to the Middle Kingdom. The Register asked Nvidia about the A800 and will update if substantial information arises.
[10]China chip imports down 12.4% as tech trade war with US intensifies
[11]US bars Nvidia and AMD from selling AI-centric accelerators to China and Russia
[12]China's drive for efficient datacenters has made liquid cooling mainstream
[13]Intel fires up internal foundry model to make its own chips
An Nvidia spokesperson told The Reg :
The Nvidia A800 GPU, which went into production in Q3 2022, is another alternative product to the Nvidia A100 GPU for customers in China. The A800 meets the US government's clear test for reduced export control and cannot be programmed to exceed it.
Nvidia’s NVIDIA A100 Tensor Core GPU, the A100x, which adds a data processing unit (SmartNIC) to the A100, and the H100 GPU that Nvidia claims will enhance natural language processing apps, were all put under a new license requirement as worries swarmed that they would end up in Chinese or Russian military hands.
Nvidia [14]warned in a late August SEC filing that new US export license requirements could hinder the development of the H100 and support for A100 customers and potentially require some operations be moved out of China. The company also said its Q3 2022 forecast could be rendered inaccurate as it included $400 million in sales to China that might not happen.
Meanwhile, GPUs still in development and produced by Taiwan Semiconductor Manufacturing Company (TSMC) under contract for Alibaba and startup Biren Technology also [15]reportedly have reduced processor speeds.
[16]
China's domestic chip facilities still have a long way to go to be globally competitive. However, Biren did have a frontrunner in the making, but amid uncertainty about its potential legality, TSMC had [17]reportedly put the brakes on the silicon.
Biren’s BR100 was pitched as an Nvidia H100 rival. The drag of the chip’s transfer rate from 640GBps to 576GBps, at least in its [18]marketing material if not in real life, would allow TSMC to produce the GPU without facing US scrutiny. ®
Get our [19]Tech Resources
[1] https://www.reuters.com/technology/exclusive-nvidia-offers-new-advanced-chip-china-that-meets-us-export-controls-2022-11-08/
[2] https://www.inspur.com/lcjtww/resource/cms/2022/08/2022082418465643998.pdf
[3] https://www.inspur.com/lcjtww/2526894/2526897/2526898/2575255/index.html
[4] http://www-waf.h3c.com/cn/Products_And_Solution/Server/H3C/Products/RackServer/Products_Series/Application_Optimization_GPU/R5300_G5/
[5] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_onprem/systems&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=2&c=2Y2qLKvUq6tsyCqGTWIU@mgAAAAs&t=ct%3Dns%26unitnum%3D2%26raptor%3Dcondor%26pos%3Dtop%26test%3D0
[6] http://www.rthpc.com/cpzx/419.html
[7] https://www.nvidia.com/content/dam/en-zz/Solutions/Data-Center/a100/pdf/nvidia-a100-datasheet-nvidia-us-2188504-web.pdf
[8] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_onprem/systems&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44Y2qLKvUq6tsyCqGTWIU@mgAAAAs&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[9] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_onprem/systems&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33Y2qLKvUq6tsyCqGTWIU@mgAAAAs&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0
[10] https://www.theregister.com/2022/10/24/china_chip_imports_down_124/
[11] https://www.theregister.com/2022/09/01/us_stops_nvidia_and_amd/
[12] https://www.theregister.com/2022/10/31/china_liquid_cooling_demand_surges/
[13] https://www.theregister.com/2022/10/12/intel_sets_up_internal_foundry/
[14] https://www.theregister.com/2022/09/01/us_stops_nvidia_and_amd/
[15] https://www.ft.com/content/7df13a5e-84e8-44af-b0d3-3e3efa6a8671
[16] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_onprem/systems&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44Y2qLKvUq6tsyCqGTWIU@mgAAAAs&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[17] https://www.theregister.com/2022/10/24/china_chip_imports_down_124/
[18] https://www.semianalysis.com/p/how-chinas-biren-is-attempting-to
[19] https://whitepapers.theregister.com/
Chinese server maker Inspur was [1]spotted swapping out the its A100 Graphic Processing Unit (GPU) in its [2]NF5688MG [PDF] for an A800 in [3]updated promotional material.
H3C, the China-based Tsinghua Group and Hewlett Packard mashup, are also suddenly featuring the A800 on [4]select servers.
[5]
According to [6]specs provided online by Chinese server-maker Omnisky, which also uses the chip, it remains eerily similar to the [7]A100 , [PDF] but with a 40GB PCIe option available in the A800, in addition to a 80GB PCIe and 80GB SXM available in both models.
[8]
[9]
The A800 has a degraded transfer rate of 400GBps across all three variations, as opposed to the 600GBps in the A100, thus falling in line with US law that restrict 600GBps and above being shipped to the Middle Kingdom. The Register asked Nvidia about the A800 and will update if substantial information arises.
[10]China chip imports down 12.4% as tech trade war with US intensifies
[11]US bars Nvidia and AMD from selling AI-centric accelerators to China and Russia
[12]China's drive for efficient datacenters has made liquid cooling mainstream
[13]Intel fires up internal foundry model to make its own chips
An Nvidia spokesperson told The Reg :
The Nvidia A800 GPU, which went into production in Q3 2022, is another alternative product to the Nvidia A100 GPU for customers in China. The A800 meets the US government's clear test for reduced export control and cannot be programmed to exceed it.
Nvidia’s NVIDIA A100 Tensor Core GPU, the A100x, which adds a data processing unit (SmartNIC) to the A100, and the H100 GPU that Nvidia claims will enhance natural language processing apps, were all put under a new license requirement as worries swarmed that they would end up in Chinese or Russian military hands.
Nvidia [14]warned in a late August SEC filing that new US export license requirements could hinder the development of the H100 and support for A100 customers and potentially require some operations be moved out of China. The company also said its Q3 2022 forecast could be rendered inaccurate as it included $400 million in sales to China that might not happen.
Meanwhile, GPUs still in development and produced by Taiwan Semiconductor Manufacturing Company (TSMC) under contract for Alibaba and startup Biren Technology also [15]reportedly have reduced processor speeds.
[16]
China's domestic chip facilities still have a long way to go to be globally competitive. However, Biren did have a frontrunner in the making, but amid uncertainty about its potential legality, TSMC had [17]reportedly put the brakes on the silicon.
Biren’s BR100 was pitched as an Nvidia H100 rival. The drag of the chip’s transfer rate from 640GBps to 576GBps, at least in its [18]marketing material if not in real life, would allow TSMC to produce the GPU without facing US scrutiny. ®
Get our [19]Tech Resources
[1] https://www.reuters.com/technology/exclusive-nvidia-offers-new-advanced-chip-china-that-meets-us-export-controls-2022-11-08/
[2] https://www.inspur.com/lcjtww/resource/cms/2022/08/2022082418465643998.pdf
[3] https://www.inspur.com/lcjtww/2526894/2526897/2526898/2575255/index.html
[4] http://www-waf.h3c.com/cn/Products_And_Solution/Server/H3C/Products/RackServer/Products_Series/Application_Optimization_GPU/R5300_G5/
[5] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_onprem/systems&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=2&c=2Y2qLKvUq6tsyCqGTWIU@mgAAAAs&t=ct%3Dns%26unitnum%3D2%26raptor%3Dcondor%26pos%3Dtop%26test%3D0
[6] http://www.rthpc.com/cpzx/419.html
[7] https://www.nvidia.com/content/dam/en-zz/Solutions/Data-Center/a100/pdf/nvidia-a100-datasheet-nvidia-us-2188504-web.pdf
[8] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_onprem/systems&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44Y2qLKvUq6tsyCqGTWIU@mgAAAAs&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[9] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_onprem/systems&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=3&c=33Y2qLKvUq6tsyCqGTWIU@mgAAAAs&t=ct%3Dns%26unitnum%3D3%26raptor%3Deagle%26pos%3Dmid%26test%3D0
[10] https://www.theregister.com/2022/10/24/china_chip_imports_down_124/
[11] https://www.theregister.com/2022/09/01/us_stops_nvidia_and_amd/
[12] https://www.theregister.com/2022/10/31/china_liquid_cooling_demand_surges/
[13] https://www.theregister.com/2022/10/12/intel_sets_up_internal_foundry/
[14] https://www.theregister.com/2022/09/01/us_stops_nvidia_and_amd/
[15] https://www.ft.com/content/7df13a5e-84e8-44af-b0d3-3e3efa6a8671
[16] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_onprem/systems&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44Y2qLKvUq6tsyCqGTWIU@mgAAAAs&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[17] https://www.theregister.com/2022/10/24/china_chip_imports_down_124/
[18] https://www.semianalysis.com/p/how-chinas-biren-is-attempting-to
[19] https://whitepapers.theregister.com/
USA Vs. China... not really.
badflorist
Who makes the money in this love spat? The USA and China aren't going to war or this crap so that only leaves the money to blame.... but who makes it? I see prices rising all over the place forcing people to up end their lives but I'm not seeing a reason beyond money and paranoia fueling this. It may be paranoia fueling money that's fueling paranoia, a sort of circular moral panic, but there needs to be clear reasons behind this "China Bad" if people are to care about any of it, and I care about none of it.
I remember when
Anonymous Coward
I remember once upon a time when certain US export restrictions were imposed but meanwhile there were official distributors of US products in China.
$JOB didn't ship the hardware because Export Rules but the customer in a restricted country still got the hardware via other means . We installed our software on their new shiny and collected the check.
I vaguely recall how easy it is to turn an MR-15
Anonymous Coward
into an AR-15.
#justsayin'
Halt
StrangerHereMyself
I wonder how long before the U.S. puts a halt to this so-called crippling. I can already see the Chinese simply "overclocking" these GPU's to get identical performance to the model it was derived from.
What the U.S. had in mind was that the export of these chips would completely stop and only lower-tiered models would be exported, not vendors relabeling their products to thwart the sanctions.
I do wonder...
...whether those speeds are artificially low and can be unlocked with firmware once they've reached their destination.