Monday, April 5th 2021
Tianshu Zhixin Big Island GPU is a 37 TeraFLOP FP32 Computing Monster
Tianshu Zhixin, a Chinese startup company dedicated to designing advanced processors for accelerating various kinds of tasks, has officially entered the production of its latest GPGPU design. Called "Big Island" GPU, it is the company's entry into the GPU market, currently dominated by AMD, NVIDIA, and soon Intel. So what is so special about Tianshu Zhixin's Big Island GPU, making it so important? Firstly, it represents China's attempt of independence from the outside processor suppliers, ensuring maximum security at all times. Secondly, it is an interesting feat to enter a market that is controlled by big players and attempt to grab a piece of that cake. To be successful, the GPU needs to represent a great design.
And great it is, at least on paper. The specifications list that Big Island is currently being manufactured on TSMC's 7 nm node using CoWoS packaging technology, enabling the die to feature over 24 billion transistors. When it comes to performance, the company claims that the GPU is capable of crunching 37 TeraFLOPs of single-precision FP32 data. At FP16/BF16 half-precision, the chip is capable of outputting 147 TeraFLOPs. When it comes to integer performance, it can achieve 317, 147, and 295 TOPS in INT32, INT16, and IN8 respectively. There is no data on double-precision floating-point numbers, so the chip is optimized for single-precision workloads. There is also 32 GB of HBM2 memory present, and it has 1.2 TB of bandwidth. If we compare the chip to the competing offers like NVIDIA A100 or AMD MI100, the new Big Island GPU outperforms both at single-precision FP32 compute tasks, for which it is designed.
Pictures of possible solutions follow.
Source:
via VideoCardz
And great it is, at least on paper. The specifications list that Big Island is currently being manufactured on TSMC's 7 nm node using CoWoS packaging technology, enabling the die to feature over 24 billion transistors. When it comes to performance, the company claims that the GPU is capable of crunching 37 TeraFLOPs of single-precision FP32 data. At FP16/BF16 half-precision, the chip is capable of outputting 147 TeraFLOPs. When it comes to integer performance, it can achieve 317, 147, and 295 TOPS in INT32, INT16, and IN8 respectively. There is no data on double-precision floating-point numbers, so the chip is optimized for single-precision workloads. There is also 32 GB of HBM2 memory present, and it has 1.2 TB of bandwidth. If we compare the chip to the competing offers like NVIDIA A100 or AMD MI100, the new Big Island GPU outperforms both at single-precision FP32 compute tasks, for which it is designed.
41 Comments on Tianshu Zhixin Big Island GPU is a 37 TeraFLOP FP32 Computing Monster
China will -or at least try to- annex Taiwan. It is just a matter if time. All the efforts to became less dependent on the US and it's allies are to shake down the economic stranglehold. That's why China is trying to tie itself to Russia and India. Big enough markets to balance whatever embargos America throws at them.
This will be a long and ugly dance though...
Longer answer? It depends how long will China behave like it does. Currently, there is absolutely no reason to like them even a bit. Its same as hatred towards other group of people, its not "phobia" when it has some merit. People love to sugarcoat it or just accuse others from being non-tolerant or something "phobic".
Currently, China IS the enemy. Accusing someone of sinophobia when it comes to Taiwan or Hongkong is same as accusing someone of teutophobia when it comes to Nazis.
They about as good guys as Islamic state, maybe China is lesser evil, but evil still.
When I see an article like this and want to head over to the discussion section, I expect discussions about the product itself not how the countries' government is the spawn of satan. And believe me, the people here with the usual sentiments towards China also share a bit, if not a lot of hatred for Chinese people, not just the government, seen it way too much for it to not be the case. My point about sinophobia remains.
We don't need none of that here. Not trying to backseat mod. Just plain tired of the seeing the same crap over and over again.
This is my last comment on the matter, I have said what needed to be said. Unwatching this thread as it'll devolve into chaos regardless...
What do we know about this product except for what is in this snippet of news.
What is the architecture lineage, what is the software base, does this have any graphics potential at all or is it a pure number cruncher, are there other products in the family or is this a one off, is this built for a specific supercomputer, will this be released to the general public, will the software base they are using be released to the general public, is the software base open or closed source.?
There is much we do not know.
Nowhere does anything hint at any sort of change when you make more mining GPUs available. It does not do a thing for supply of GPUs in other channels. I know, its comfortable to think otherwise, but get real. Mining is a bottomless pit as long as the projected ROI is favorable.
But mining is not the key issue causing these shortages at large.
The worst part is companies are not going to massively ramp production to meet demand for the same reason internet providers aren't investing in better infrastructure. They believe this is just a temporary side effect of the pandemic. While I agree to a point, I think the end result is that things will never completely go back to the way they were before. Some of the people who found a new distraction or reinvigorated an old passion for gaming are going to keep on gaming and spending money on gaming hardware. Just like some jobs that shifted to work from home will stay work from home and many, many other things will keep internet bandwidth demands high.