The second domestic full-featured GPU in half a year, comes with the first gaming graphics card! Moore thread core making is very fast, Zhang Jianzhong is so "yellow"
Jin Lei comes from Aofei Temple
Qubits | Public account QbitAI
He's coming, he's coming.
Holding a GPU , he is wearing a black leather jacket , jeans, yellow skin, long hair, and black-framed glasses that cannot block his tough-guy aura.
But he is not Nvidia's Huang.
He is Lao Zhang, who was once one of Lao Huang's right-hand men——
Zhang Jianzhong , James, previously served as the top leader of NVIDIA China for 15 years.
He is now better known as the founder and CEO of the domestic GPU company Moore Thread .
The GPU that has just been warmed up is the second one that he has put into mass production and launched this year .
It has only been half a year .
Not only that, there are more tags attached to this GPU:
Domestic , full-featured , the world's first , mid-to-high-end ...
And based on it, Zhang Jianzhong released the first mid-to-high-end gaming graphics card in China .
To be honest, it is hard to imagine that such "speed" was created by a company that has only been established for 2 years.
This inevitably leads to questions:
While achieving speed, have the quality and performance also kept up?
We might as well take a look together.
The world's first full-featured GPU supporting PCIe interface
Moore Thread’s second domestic full-featured GPU is named “MT-Chunxiao” .
Zhang Jianzhong also released it as the first product.
It is understood that Chunxiao integrates 22 billion transistors, has built-in 4096 MUSA architecture general computing cores and 128 tensor computing cores, and can support computing precisions such as FP32, FP16 and INT8.
Other key parameters are as follows:
-
GPU core frequency: 1.8GHz
-
FP32 computing power: 14.4 TFLOPS
-
INT8 computing power: 57.6 TOPS
-
Video memory bandwidth: 448GB/s
-
- Video memory type: GDDR6
Zhang Jianzhong also mentioned at the scene that Chunxiao has unlocked a "world first":
Because it is the only GPU in the industry that supports the PCIe Gen5 interface.
(Many manufacturers have reached a consensus that PCIe Gen5 will be the key development direction of consumer and enterprise storage devices in the future.)
So what is the difference between the two and the "Su Di" GPU released by Moore Threads half a year ago ?
Zhang Jianzhong said that Chunxiao has achieved a comprehensive upgrade in the four major engines of Moore's Thread GPU:
-
The performance of modern graphics rendering engines is improved by up to 3-5 times
-
AI computing acceleration engine performance increased by up to 4 times
-
Intelligent multimedia engine performance increased by up to 4 times
-
Physical simulation engine performance increased by up to 2.5 times
As for why Moore's Threads has to innovate its own GPU in just half a year, we asked Zhang Jianzhong about this question:
Our first full-featured GPU Sudi is actually a mid- to low-end processor and can meet the needs of domestic domestic applications in the GPU industry.
But for most mainstream users, they still expect higher-performance GPUs, so we quickly released Chunxiao to satisfy high-end gamers and more users’ graphics and computing needs.
In this way, our products can cover all high, medium and low-end users.
Speaking of games, Zhang Jianzhong released another "first in China" product based on the Chunxiao GPU.
The first gaming graphics card in China
In fact, Moore Threads also released a graphics card product MTT S60 based on Sudi half a year ago .
However, the "place of use" of this graphics card seems to be more oriented towards industry, that is, the B-end.
This time, the graphics card MTT S80 built based on Chunxiao is the kind that can be touched by more people - the first gaming graphics card in China .
At the scene, Zhang Jianzhong also used a more interesting word to describe it: "national tide" .
From a performance perspective, its 4096 programmable MUSA cores can provide 14.4TFLOPS of single-precision floating point computing power at a main frequency of 1.8GHz.
Similar to Chunxiao, MTT S80 is also the industry's first graphics card product equipped with a PCIe Gen5 interface:
Combined with 16GB GDDR6 large-capacity high-speed video memory, supplemented by 8K ultra-high-definition and 1080P 360Hz high refresh rate display output capabilities, it can bring a great experience to gamers.
It seems that "all talk and no practice" is not the style of Moore's thread conference. Just like last time , Zhang Jianzhong also directly applied the effect on the spot.
For example, it has been adapted to "Diablo 3" in the Windows environment , and this game has relatively high requirements for graphics card performance.
With the blessing of MTT S80, even if the whole process is in 4K high-definition quality, the FPS can be maintained at around 60 (the higher the FPS, the smoother the picture) .
In addition, Zhang Jianzhong also demonstrated the "Need for Speed" that is loved by racing game enthusiasts . The effect under MTT S80 can be said to be quite smooth:
It is understood that the MTT S80 currently has a built-in MUSA DirectX Driver module in the Windows driver and has completed the adaptation of dozens of mainstream games.
More importantly, Zhang Jianzhong said that this graphics card will be available for sale in limited quantities on Double Eleven Day .
So you can look forward to the actual effect after you get it and whether the price is good.
New full-featured server GPU products
In terms of server products, Moore Threads has also updated the MTT S3000 .
Similarly, it is also based on the MUSA architecture and Chunxiao GPU, and its computing power can cover the complete MUSA software stack of graphics rendering, video processing, and deep learning.
Supported scenarios include AI reasoning and training, cloud gaming, cloud rendering, video cloud, digital twins, digital content creation, etc.
In terms of performance, MTT S3000 contains 4096 MUSA stream processing cores and 128 dedicated tensor computing cores, with a transistor scale of 22 billion.
Its operating frequency is 1.9GHz, and the memory bit width is 256bit. It is equipped with 32GB GDDR6 memory and has a bandwidth of 448GB/s. It can support a variety of calculation precisions such as FP32, FP16, and INT8, among which the FP32 computing power can reach 15.2TFLOPS.
In Zhang Jianzhong’s view, ecological collaboration is crucial to the advancement of AI applications .
Therefore, MTT S3000 is also compatible with multiple mainstream deep learning frameworks such as PyTorch, TensorFlow, Baidu PaddlePaddle , Jittor , and can accelerate dozens of AI models such as Transformer, CNN, and RNN.
Moreover, MTT S3000 can be said to have "advanced with the times", and the recently popular AI paintings: Disco Diffusion and Stable Diffusion are also able to hold on.
More than just a hardware update
In addition to the above-mentioned hardware products, looking at the normal press conference, "the combination of software and hardware" is also a major feature of Moore Thread.
This is actually not difficult to understand. During our communication with Zhang Jianzhong, he also revealed the reason:
The software ecosystem is the key to promoting the popularization of GPU computing.
The GPU research and development system is very complex. Only with hardware can software be developed. After the software runs on the hardware, more optimization is needed. After optimization, defects will be found in the architecture, and the hardware must be optimized again.
Hardware and software are in a process of mutual iteration and continuous improvement.
To this end, Moore Thread takes the MUSA architecture as the core this time, and Moore Thread has released a complete MUSA software stack .
Its purpose is to serve the majority of developers and end users.
In addition, Moore Thread has made corresponding new actions in GPU cloud native, metaverse, and AIGC.
One More Thing
Let’s focus on Zhang Jianzhong’s outfit this time:
This leather jacket, well, is really interesting.
-over-
"AIGC Live Series - Application Implementation and Commercialization" Live Broadcast Registration
Apart from the technology boom, how does AIGC actually come to fruition? And how to achieve long-term business realization?
From 7 to 8 pm on November 3-4, Qubit Think Tank will join hands with leading domestic AIGC companies and investment institutions to answer your questions. Scan the QR code to make an appointment~
Click here to follow me
Featured Posts