Traditionally, the GPU market has been dominated by the two graphics giants– Nvidia and AMD. While the argument continues on which one is better, there’s no doubt that both makers enjoy competitive rapport with their customers. However, the Big Island 7nm GPU from China is all set to disrupt this two-brand industry. Will it succeed in toppling the two titans? Let’s find out!
Today, we’ll give you a full rundown of this new and highly-anticipated graphics technology. From its background and model to its features and future prospects, here’s all you need to know about Big Island 7nm.
Background and Development
China has long tried to break into tech markets, regardless of niche and industry. With the recent tussles and tensions with the US, the Asian manufacturing giant is keen on treading new grounds in technology.
The graphics processor industry, right now, runs at the beck and call of Nvidia and AMD (Both US companies). The gaming industry continues to be among the fastest growing and most lucrative sectors. So, it’s no surprise that a Chinese company would want a piece of the market.
The company in question is Shanghai Tianshu Intellectual Semiconductor. Commonly called Tianshu Zhixin, the company had been working on its first home-grown graphics processor for some years. By March-end, everyone was aware of the launch of Big Island 7nm GPU from China.
Overview of the Big Island 7nm GPU from China
The Big Island 7nm GPU from China is essentially a General Purpose Graphics Processing Unit (GPGPU). So, if it does break into the market, the Nvidia and AMD rivalry will see a new dynamic added to it.
According to Shanghai Tianshu, Big Island is designed to match the highest graphics technology available today. Also, they claim it will outperform current industry leaders on many parameters. These parameters include price, performance, power consumption, etc.
Specifications of Big Island 7nm GPU from China
The BI GPU rests on the Shanghai Tianshu’s custom 7nm process node. It is housed within a 2.5D CoWos packaging. Additionally, the silicon interposer measures 65nm and has over 24 billion transistors. These BI specs are impressive, to say the least!
There are also features that match the styles followed by AMD and Nvidia. For instance, Big Island’s GPGPU has an HBM2 memory of 32GB and runs with a bandwidth of 1.2 TB/second. It also supports compatibility with the PCle 4.0 standard. As far as FP32 workload goes, BI can handle up to 37 TFLOPs.
Nvidia and AMD are known for their efficient architecture – Ampere and cDNA, respectively. But with Big Island, the precise architecture is currently under wraps. But we should get a good idea once the hardware reaches retail shelves and online stores. But given China’s current push for domestic markets, it may not happen immediately.
Big Island Vs. Nvidia Vs. AMD: What you need to know
If you’re familiar with Nvidia and AMD technology, the specifications above should already give you a good idea about how BI matches up. But for the sake of clarity, we’ll make a quick comparison of some core parameters.
To keep the differences consistent, we’ve taken the AMD Instinct MI100 and Nvidia’s A100 for this exercise.
For starters, Big Island’s transistor count (24B) is barely half of what Nvidia (54B) and AMD (50B) offer today. But processing speed is not restricted only to transistor count. Also, BI may release subsequent models that have more transistors. So, we’ll have to consider other aspects without calling it straight from this feature alone.
Big Island’s Cores are listed as TBC, while AMD (7680) and Nvidia (6912) both hover around the 7000-7500 mark. Before getting to test Big Island’s cores, you can assume that AMD has the upper hand on this front.
Memory and Bandwidth
Big Island’s HBM2 memory matches AMD’s memory of 32GB. It also has an identical bandwidth of 1.2TB/s. With the A100 offering a whopping 80GB of HBM2 and bandwidth of 1.6TB/s, Nvidia is clearly ahead here.
FP32 and BFLOAT16 comparison
The FP32 Computation reveals varying results for all three GPUs. The AMD scores 23 TFLOPs which goes up to 46 TFLOPs for Matrix. The Nvidia A100 scores a minimum of 19.5 TFLOPs which can shoot skywards to 312 TFLOPs for Sparsity. Since Big Island clocks around 37 TFOPs, it can indeed compete with both AMD and Nvidia’s averages.
The results are similar when we consider the BFLOAT16 Computation. AMD bottoms out around 92 TFLOPs, while Nvidia’s lower limit is around 312 TFLOPS. Nvidia’s BFLOAT16 score shoots up once again to 624 TFLOPs for Sparsity. Big Island averages around 147 TFLOPs, which has the potential to be an alternative for both Nvidia and AMD.
Thermal Design Power (TDP)
The TDP (Thermal Design Power) for all three GPUs are on very competitive levels. Nvidia’s A100 is slightly behind with 400 watts, while both Big Island and AMD have TDPs of about 300 watts. So, Big Island’s power consumption is certainly market-competitive at the moment.
Finally, an important factor to consider in comparison is the price. However, the Big Island 7nm GPU from China is still launching and yet to make it to international markets. So, we don’t have price information for this new GPU. Tianshu Zhixin has announced that Big Island will have a better price vs. performance ratio than existing systems. But until it reaches our desks, no conclusive observations can be made right now.
Nvidia and AMD have long held the mantle as premier manufacturers of graphics processors. To avoid the monopoly of one brand over the market, getting a third player in the industry is a welcome change.
However, Big Island will have to overcome specific challenges from both Nvidia and AMD as well as market forces. The global shortage of chips and components is a temporary but real obstacle that the semiconductor industry faces right now. So, it remains to be seen if Big Island will prove its mettle in this competitive industry. For now, we’ll have to stick to AMD and Nvidia, but keep our sights out for Big Island’s entry.