Biren Technology Unveils BR100 7 nm HPC GPU with 77 Billion Transistors

[Uskompuf]

Chinese company Biren Technology has recently unveiled the Biren BR100 HPC GPU during their Biren Explore Summit 2022 event. The Biren BR100 features an in-house chiplet architecture with 77 billion transistors and is manufactured on a 7 nm process using TSMC's 2.5D CoWoS packaging technology. The card is equipped with 300 MB of onboard cache alongside 64 GB of HBM2E memory running at 2.3 TFLOPs. This combination delivers performance above that of the NVIDIA Ampere A100 achieving 1024 TFLOPs in 16-bit floating point operations.

The company also announced the BR104 which features a monolithic design and should offer approximately half the performance of the BR100 at a TDP of 300 W. The Biren BR104 will be available as a standard PCIe card while the BR100 will come in the form of an OAM compatible board with a custom tower cooler. The pricing and availability information for these cards is currently unknown.



View at TechPowerUp Main Site | Source

 

[DeathtoGnomes]

Reviews reviews reviews will tell its story, not some PR.

 

[Daven]

Since they didn’t release a transistor or transistor block level diagram, there is absolutely no way to know if there is anything under the heat spreader. Most of the specs are just exact powers of 2. It would be nice to have more products like this on the market but I’m still skeptical of extreme hardware that just all of a sudden comes out of China. I would love to be proven wrong.

 

[TheoneandonlyMrK]

Is this the Third new Chinese GPU?

 

[Denver]

Reviews reviews reviews will tell its story, not some PR.

It will never happen. If the product exists, it will be limited to the Chinese domestic market for obvious reasons.

 

[FreedomEclipse]

It will never happen. If the product exists, it will be limited to the Chinese domestic market for obvious reasons.

Aliexpress would like to have a word.

 

[stimpy88]

When I read press releases like this, my first thought is who did they copy/rip off. My second thought is does it actually work.

 

[dj-electric]

I think that something in the terminology of GPUs has to be changed.
If a GPU does not do graphics processing, but is just a giant parallel proccesing node unit, maybe we can come up with a more appropriate name.

 

[defaultluser]

When I read press releases like this, my first thought is who did they copy/rip off. My second thought is does it actually work.

or maybe used imagination?

\https://www.hardwaretimes.com/this-...next-gen-rtx-4090-and-a-chiplet-architecture/

 

[Vya Domus]

I think that something in the terminology of GPUs has to be changed.
If a GPU does not do graphics processing, but is just a giant parallel proccesing node unit, maybe we can come up with a more appropriate name.

There used to be a name for something like that in the 80s and 90s, they were called VPUs (vector processing units), they were basically analogous to what GPUs are used for today in datacenters. They fell out of favor because regular CPUs started to integrate vector instructions as well.

 

[Steevo]

There used to be a name for something like that in the 80s and 90s, they were called VPUs (vector processing units), they were basically analogous to what GPUs are used for today in datacenters. They fell out of favor because regular CPUs started to integrate vector instructions as well.

The fusion of CISC and RISC is already starting and with both major GPU players architecture becoming so closely related it’s almost like we will see our first full system on a chip that can run everything efficiently in our lifetimes.

 

[konga]

I think some people here are confused on what this is meant to be. It's a datacenter HPC GPU. There will be no reviews for it, just like there's no reviews for Nvidia and AMD's datacenter GPUs (in mainstream tech media at least). And when was the last time you saw one of those on ali express? We may end up with very little information about this thing even if it's 100% real and has good performance because it'll likely stay in chinese datacenters exclusively.

 

[Metroid]

Can it run crysis?

 

[ModEl4]

This thing should be huge (32768 shader processors at +1.95GHz or more likely 40960 shader processors at +1560MHz?) no wonder it's chiplets based.

 

[Minus Infinity]

Imagine how many faces per second this will scan for Xi. I'll bet it'll be used for all sorts of nefarious reasons

 

[konga]

Imagine how many faces per second this will scan for Xi. I'll bet it'll be used for all sorts of nefarious reasons

Wait until you hear what Nvidia and AMD's hardware is going to do for every other government. Nvidia isn't even trying to hide it, if you watch their presentations about the omniverse and their ultimate goal to create a virtual clone of the entire world that can be monitored.

 

[DeathtoGnomes]

.Imagine how many faces per second this will scan for all the countries that use face scanning. I'll bet it'll be used for all sorts of nefarious reasons

FTFY. Just about every country has some level of ability to face scan, every traffic cam is a spy cam. And if they can scan your face, they likely have a file started on you.