GTX170 So this dawn unified computing architecture will have a huge impact on the future development of human science and technology, and will also have a profound impact on the computer industry. "
After hearing Lin Chen's words, people just nodded slightly in understanding.
“Well, you may not understand the power and greatness of this Dawn unified computing architecture.
Considering that what we are going to release is a graphics card, we will not introduce the Sugon unified computing architecture in other fields.
So let's take the graphics card as an example. Let me introduce the powerful significance of this architecture in the graphics card. "
While Lin Chen was talking, the background of the projection screen behind him switched to a picture of a bunch of chips arranged in an orderly manner.
"As shown in the figure, the unified computing architecture is a parallel computing architecture. Its main feature is that it allows multiple GPU chips to perform parallel computing."
Lin Chen paused for a moment when he said this, and then continued:
“There is currently an SLI technology on the market called Graphics Crossfire, which allows two or more graphics cards to work together.
However, this form cannot actually take advantage of the power of multiple graphics cards. The actual increase in performance of one more graphics card does not exceed 50%, and there is a lot of waste of performance.
Our Sugon unified computing architecture is different, it can perfectly allow multiple GPU graphics chips to cooperate.
There will not be a black technology technology with huge performance waste like the graphics card crossfire SLI technology.
With this technology, it will become a reality for dozens, hundreds or even thousands of GPU graphics chips to work together.
You can imagine how powerful the smoothness and game quality that can be achieved if dozens, hundreds, and thousands of graphics cards are integrated together. "
Hearing Lin Chen's words, people couldn't help but imagine the terrifying performance that would be achieved if dozens, hundreds, and thousands of graphics cards were gathered together.
Then their eyes couldn't help but become a little crazy, because this prospect was so wonderful!
Not to mention the terrifying performance of dozens, hundreds, and thousands of graphics cards brought together, the actual performance of just a few graphics cards can drive them crazy!
Under this circumstance, people at this time finally realized how powerful Sugon's unified computing architecture was and how great this technology was.
In response, a faint smile appeared on Lin Chen's lips, and then he said with a smile:
“As everyone has guessed, our Sugon graphics card is an independent graphics card developed using Sugon’s unified computing architecture.
Among them, the Sugon GTX170 graphics card is mainly composed of 240 stream processors using SIMT technology, 32 texture pickup units, 80 raster processing units, and finally 2G ultra-large video memory.
Through cooperation under the Sugon unified computing architecture, they finally exerted terrifyingly powerful graphics display performance!
One thing worth mentioning is that the GTX170 chip transistor count of the Sugon GTX170 graphics card is as high as 1.4 billion transistors.
This number of transistors is equivalent to the number of transistors of 8 Sugon R7 Ryzen processors! "
Hearing this, the nearly 20,000 spectators, guest reporters, and tens of millions of netizens on the Internet who were watching the press conference through live video all showed stunned expressions.
What did they hear? A graphics card chip actually has 1.4 billion transistors. Is this too many transistors?
This is absolutely crazy!
Of course, after being surprised, they were even more eager to know how powerful the Sugon GTX170 graphics card was.
However, when people's eyes lit up, Invida CEO Lao Huang's face looked quite serious.
"Damn it, they actually developed the CUDA architecture too!"
At the same time as he said this, Invida President Lao Huang looked quite ugly.
There is nothing more tragic and depressing than having to spend a huge amount of money to develop core technology, only to find that others have finished the product before the product is completed.
You must know that their company, Yinweida, has been developing the CUDA architecture since 2004. As a result, the development of the CUDA architecture has not yet been completed. Sugon Technology has already launched the finished product.
This depressing mood almost made Invida CEO Lao Huang want to vomit blood. Of course, what made him vomit blood even more was that through Lin Chen's explanation, he already knew how terrifying this Sugon GTX170 graphics card was!
Not to mention other things, just the 1.4 billion transistors can make people's hearts tremble.
Because there is currently no graphics card chip on the market with 1.4 billion transistors.
With so many transistors, no matter how inefficient the Sugon GTX170 graphics card is designed, this graphics card will still have terrifying performance.
Of course, this Sugon GTX170 graphics card is definitely not a design waste, because the architecture of this graphics card is Sugon Unified Computing Architecture, which is similar to the CUDA architecture.
Sugon's unified computing architecture can certainly fully unleash the potential of those 1.4 billion transistors.
Therefore, the graphics card developed by Sugon Technology is simply the best, and it is simply a super artifact that can kill any opponent in an instant!
Thinking of this, Invida President Lao Huang's face became even more ugly.
However, if Lao Huang knew at this time that the Suguang GTX170 graphics card architecture in front of him was actually a copy of the graphics card architecture of the previous Invida GTX280 series, then Lao Huang would probably be so angry that he died on the spot.
In fact, what Sugon Technology copied at this time was the GTX280 graphics card architecture of the previous Invida. The number of stream processors, texture pickup units, and raster processing units were all the same.
Even the number of transistors is 1.4 billion. The difference is that Sugon Technology was kind-hearted and gave 2G of super large video memory.
In addition, Sugon Technology's GTX170 graphics card chip area will be much larger than the previous GTX280 graphics card chip.
After all, one uses 90 nanometers and the other uses 65 nanometers. There is indeed a big gap in the process technology between the two parties.
As for whether it is possible to plug 1.4 billion transistors into a chip using a 90nm process technology, it will certainly work as long as the heat dissipation problem is solved!
You must know that the previous Invida GeForce RTX 4090 graphics card used an 8-nanometer process technology. The number of transistors on it was as high as 76 billion transistors. That is the real monster!
Of course, there are actually stronger ones. For example, using the 5-nanometer process technology in 2023, .D's MI300 graphics card will have 146 billion transistors.
As long as the heat dissipation is done well, the mere number of 1.4 billion transistors will naturally not be a problem.
Moreover, this is a parallel computing architecture, and the computing tasks are evenly distributed among multiple GPU chips. The heat accumulation problem of this chip is not as serious as that of the CPU processor.
Of course, using a 90-nanometer process to pack 1.4 billion transistors is not without its disadvantages. The disadvantage is to solve the heat dissipation problem.
Compared with other graphics cards of the same era, and even compared to the previous GTX280 graphics card, the Sugon graphics card looks a bit "huge and hideous".
There is no way. As the chip becomes larger, the requirements for heat dissipation are naturally higher. At this time, it is natural to install two high-power fans and multiple heat pipes for rapid heat dissipation. (End of chapter)
